def extractnumber_de(text): """ This function prepares the given text for parsing by making numbers consistent, getting rid of contractions, etc. Args: text (str): the string to normalize Returns: (int) or (float): The value of extracted number undefined articles cannot be suppressed in German: 'ein Pferd' means 'one horse' and 'a horse' """ aWords = text.split() aWords = [ word for word in aWords if word not in ["der", "die", "das", "des", "den", "dem"] ] and_pass = False valPreAnd = False val = False count = 0 while count < len(aWords): word = aWords[count] if is_numeric(word): # if word.isdigit(): # doesn't work with decimals val = float(word) elif isFractional_de(word): val = isFractional_de(word) elif isOrdinal_de(word): val = isOrdinal_de(word) else: if word in de_numbers: val = de_numbers[word] if count < (len(aWords) - 1): wordNext = aWords[count + 1] else: wordNext = "" valNext = isFractional_de(wordNext) if valNext: val = val * valNext aWords[count + 1] = "" if not val: # look for fractions like "2/3" aPieces = word.split('/') # if (len(aPieces) == 2 and is_numeric(aPieces[0]) # and is_numeric(aPieces[1])): if look_for_fractions(aPieces): val = float(aPieces[0]) / float(aPieces[1]) elif and_pass: # added to value, quit here val = valPreAnd break else: count += 1 continue aWords[count] = "" if and_pass: aWords[count - 1] = '' # remove "and" val += valPreAnd elif count + 1 < len(aWords) and aWords[count + 1] == 'und': and_pass = True valPreAnd = val val = False count += 2 continue elif count + 2 < len(aWords) and aWords[count + 2] == 'und': and_pass = True valPreAnd = val val = False count += 3 continue break if not val: return False return val
def extractnumber_sv(text): """ This function prepares the given text for parsing by making numbers consistent, getting rid of contractions, etc. Args: text (str): the string to normalize Returns: (int) or (float): The value of extracted number """ aWords = text.split() and_pass = False valPreAnd = False val = False count = 0 while count < len(aWords): word = aWords[count] if is_numeric(word): val = float(word) elif word == "första": val = 1 elif word == "andra": val = 2 elif word == "tredje": val = 3 elif word == "fjärde": val = 4 elif word == "femte": val = 5 elif word == "sjätte": val = 6 elif is_fractional_sv(word): val = is_fractional_sv(word) else: if word == "en": val = 1 if word == "ett": val = 1 elif word == "två": val = 2 elif word == "tre": val = 3 elif word == "fyra": val = 4 elif word == "fem": val = 5 elif word == "sex": val = 6 elif word == "sju": val = 7 elif word == "åtta": val = 8 elif word == "nio": val = 9 elif word == "tio": val = 10 if val: if count < (len(aWords) - 1): wordNext = aWords[count + 1] else: wordNext = "" valNext = is_fractional_sv(wordNext) if valNext: val = val * valNext aWords[count + 1] = "" if not val: # look for fractions like "2/3" aPieces = word.split('/') if look_for_fractions(aPieces): val = float(aPieces[0]) / float(aPieces[1]) elif and_pass: # added to value, quit here val = valPreAnd break else: count += 1 continue aWords[count] = "" if and_pass: aWords[count - 1] = '' # remove "och" val += valPreAnd elif count + 1 < len(aWords) and aWords[count + 1] == 'och': and_pass = True valPreAnd = val val = False count += 2 continue elif count + 2 < len(aWords) and aWords[count + 2] == 'och': and_pass = True valPreAnd = val val = False count += 3 continue break if not val: return False return val
def extract_datetime_de(string, currentDate, default_time): def clean_string(s): """ cleans the input string of unneeded punctuation and capitalization among other things. 'am' is a preposition, so cannot currently be used for 12 hour date format """ s = s.lower().replace('?', '').replace('.', '').replace(',', '') \ .replace(' der ', ' ').replace(' den ', ' ').replace(' an ', ' ').replace( ' am ', ' ') \ .replace(' auf ', ' ').replace(' um ', ' ') wordList = s.split() for idx, word in enumerate(wordList): if isOrdinal_de(word) is not False: word = str(isOrdinal_de(word)) wordList[idx] = word return wordList def date_found(): return found or \ ( datestr != "" or timeStr != "" or yearOffset != 0 or monthOffset != 0 or dayOffset is True or hrOffset != 0 or hrAbs or minOffset != 0 or minAbs or secOffset != 0 ) if string == "" or not currentDate: return None found = False daySpecified = False dayOffset = False monthOffset = 0 yearOffset = 0 dateNow = currentDate today = dateNow.strftime("%w") currentYear = dateNow.strftime("%Y") fromFlag = False datestr = "" hasYear = False timeQualifier = "" timeQualifiersList = [ u'früh', 'morgens', 'vormittag', 'vormittags', 'nachmittag', 'nachmittags', 'abend', 'abends', 'nachts' ] markers = ['in', 'am', 'gegen', 'bis', u'für'] days = [ 'montag', 'dienstag', 'mittwoch', 'donnerstag', 'freitag', 'samstag', 'sonntag' ] months = [ 'januar', 'februar', u'märz', 'april', 'mai', 'juni', 'juli', 'august', 'september', 'october', 'november', 'dezember' ] monthsShort = [ 'jan', 'feb', u'mär', 'apr', 'mai', 'juni', 'juli', 'aug', 'sept', 'oct', 'nov', 'dez' ] validFollowups = days + months + monthsShort validFollowups.append("heute") validFollowups.append("morgen") validFollowups.append(u"nächste") validFollowups.append(u"nächster") validFollowups.append(u"nächstes") validFollowups.append(u"nächsten") validFollowups.append(u"nächstem") validFollowups.append("letzte") validFollowups.append("letzter") validFollowups.append("letztes") validFollowups.append("letzten") validFollowups.append("letztem") validFollowups.append("jetzt") words = clean_string(string) for idx, word in enumerate(words): if word == "": continue wordPrevPrev = words[idx - 2] if idx > 1 else "" wordPrev = words[idx - 1] if idx > 0 else "" wordNext = words[idx + 1] if idx + 1 < len(words) else "" wordNextNext = words[idx + 2] if idx + 2 < len(words) else "" # this isn't in clean string because I don't want to save back to words if word != 'morgen' and word != u'übermorgen': if word[-2:] == "en": word = word[:-2] # remove en if word != 'heute': if word[-1:] == "e": word = word[:-1] # remove plural for most nouns start = idx used = 0 # save timequalifier for later if word in timeQualifiersList: timeQualifier = word # parse today, tomorrow, day after tomorrow elif word == "heute" and not fromFlag: dayOffset = 0 used += 1 elif word == "morgen" and not fromFlag and wordPrev != "am" and \ wordPrev not in days: # morgen means tomorrow if not "am # Morgen" and not [day of the week] morgen dayOffset = 1 used += 1 elif word == u"übermorgen" and not fromFlag: dayOffset = 2 used += 1 # parse 5 days, 10 weeks, last week, next week elif word == "tag" or word == "tage": if wordPrev[0].isdigit(): dayOffset += int(wordPrev) start -= 1 used = 2 elif word == "woch" and not fromFlag: if wordPrev[0].isdigit(): dayOffset += int(wordPrev) * 7 start -= 1 used = 2 elif wordPrev[:6] == u"nächst": dayOffset = 7 start -= 1 used = 2 elif wordPrev[:5] == "letzt": dayOffset = -7 start -= 1 used = 2 # parse 10 months, next month, last month elif word == "monat" and not fromFlag: if wordPrev[0].isdigit(): monthOffset = int(wordPrev) start -= 1 used = 2 elif wordPrev[:6] == u"nächst": monthOffset = 1 start -= 1 used = 2 elif wordPrev[:5] == "letzt": monthOffset = -1 start -= 1 used = 2 # parse 5 years, next year, last year elif word == "jahr" and not fromFlag: if wordPrev[0].isdigit(): yearOffset = int(wordPrev) start -= 1 used = 2 elif wordPrev[:6] == u"nächst": yearOffset = 1 start -= 1 used = 2 elif wordPrev[:6] == u"nächst": yearOffset = -1 start -= 1 used = 2 # parse Monday, Tuesday, etc., and next Monday, # last Tuesday, etc. elif word in days and not fromFlag: d = days.index(word) dayOffset = (d + 1) - int(today) used = 1 if dayOffset < 0: dayOffset += 7 if wordNext == "morgen": # morgen means morning if preceded by # the day of the week words[idx + 1] = u"früh" if wordPrev[:6] == u"nächst": dayOffset += 7 used += 1 start -= 1 elif wordPrev[:5] == "letzt": dayOffset -= 7 used += 1 start -= 1 # parse 15 of July, June 20th, Feb 18, 19 of February elif word in months or word in monthsShort and not fromFlag: try: m = months.index(word) except ValueError: m = monthsShort.index(word) used += 1 datestr = months[m] if wordPrev and (wordPrev[0].isdigit() or (wordPrev == "of" and wordPrevPrev[0].isdigit())): if wordPrev == "of" and wordPrevPrev[0].isdigit(): datestr += " " + words[idx - 2] used += 1 start -= 1 else: datestr += " " + wordPrev start -= 1 used += 1 if wordNext and wordNext[0].isdigit(): datestr += " " + wordNext used += 1 hasYear = True else: hasYear = False elif wordNext and wordNext[0].isdigit(): datestr += " " + wordNext used += 1 if wordNextNext and wordNextNext[0].isdigit(): datestr += " " + wordNextNext used += 1 hasYear = True else: hasYear = False # parse 5 days from tomorrow, 10 weeks from next thursday, # 2 months from July if ( word == "von" or word == "nach" or word == "ab") and wordNext \ in validFollowups: used = 2 fromFlag = True if wordNext == "morgen" and wordPrev != "am" and \ wordPrev not in days: # morgen means tomorrow if not "am # Morgen" and not [day of the week] morgen: dayOffset += 1 elif wordNext in days: d = days.index(wordNext) tmpOffset = (d + 1) - int(today) used = 2 if tmpOffset < 0: tmpOffset += 7 dayOffset += tmpOffset elif wordNextNext and wordNextNext in days: d = days.index(wordNextNext) tmpOffset = (d + 1) - int(today) used = 3 if wordNext[:6] == u"nächst": tmpOffset += 7 used += 1 start -= 1 elif wordNext[:5] == "letzt": tmpOffset -= 7 used += 1 start -= 1 dayOffset += tmpOffset if used > 0: if start - 1 > 0 and words[start - 1].startswith("diese"): start -= 1 used += 1 for i in range(0, used): words[i + start] = "" if start - 1 >= 0 and words[start - 1] in markers: words[start - 1] = "" found = True daySpecified = True # parse time timeStr = "" hrOffset = 0 minOffset = 0 secOffset = 0 hrAbs = None minAbs = None for idx, word in enumerate(words): if word == "": continue wordPrevPrev = words[idx - 2] if idx > 1 else "" wordPrev = words[idx - 1] if idx > 0 else "" wordNext = words[idx + 1] if idx + 1 < len(words) else "" wordNextNext = words[idx + 2] if idx + 2 < len(words) else "" wordNextNextNext = words[idx + 3] if idx + 3 < len(words) else "" wordNextNextNextNext = words[idx + 4] if idx + 4 < len(words) else "" # parse noon, midnight, morning, afternoon, evening used = 0 if word[:6] == "mittag": hrAbs = 12 used += 1 elif word[:11] == "mitternacht": hrAbs = 0 used += 1 elif word == "morgens" or (wordPrev == "am" and word == "morgen") or word == u"früh": if not hrAbs: hrAbs = 8 used += 1 elif word[:10] == "nachmittag": if not hrAbs: hrAbs = 15 used += 1 elif word[:5] == "abend": if not hrAbs: hrAbs = 19 used += 1 # parse half an hour, quarter hour elif word == "stunde" and \ (wordPrev in markers or wordPrevPrev in markers): if wordPrev[:4] == "halb": minOffset = 30 elif wordPrev == "viertel": minOffset = 15 elif wordPrev == "dreiviertel": minOffset = 45 else: hrOffset = 1 if wordPrevPrev in markers: words[idx - 2] = "" words[idx - 1] = "" used += 1 hrAbs = -1 minAbs = -1 # parse 5:00 am, 12:00 p.m., etc elif word[0].isdigit(): isTime = True strHH = "" strMM = "" remainder = "" if ':' in word: # parse colons # "3:00 in the morning" stage = 0 length = len(word) for i in range(length): if stage == 0: if word[i].isdigit(): strHH += word[i] elif word[i] == ":": stage = 1 else: stage = 2 i -= 1 elif stage == 1: if word[i].isdigit(): strMM += word[i] else: stage = 2 i -= 1 elif stage == 2: remainder = word[i:].replace(".", "") break if remainder == "": nextWord = wordNext.replace(".", "") if nextWord == "am" or nextWord == "pm": remainder = nextWord used += 1 elif nextWord == "abends": remainder = "pm" used += 1 elif wordNext == "am" and wordNextNext == "morgen": remainder = "am" used += 2 elif wordNext == "am" and wordNextNext == "nachmittag": remainder = "pm" used += 2 elif wordNext == "am" and wordNextNext == "abend": remainder = "pm" used += 2 elif wordNext == "morgens": remainder = "am" used += 1 elif wordNext == "nachmittags": remainder = "pm" used += 1 elif wordNext == "abends": remainder = "pm" used += 1 elif wordNext == "heute" and wordNextNext == "morgen": remainder = "am" used = 2 elif wordNext == "heute" and wordNextNext == "nachmittag": remainder = "pm" used = 2 elif wordNext == "heute" and wordNextNext == "abend": remainder = "pm" used = 2 elif wordNext == "nachts": if strHH > 4: remainder = "pm" else: remainder = "am" used += 1 else: if timeQualifier != "": if strHH <= 12 and \ (timeQualifier == "abends" or timeQualifier == "nachmittags"): strHH += 12 # what happens when strHH is 24? else: # try to parse # s without colons # 5 hours, 10 minutes etc. length = len(word) strNum = "" remainder = "" for i in range(length): if word[i].isdigit(): strNum += word[i] else: remainder += word[i] if remainder == "": remainder = wordNext.replace(".", "").lstrip().rstrip() if (remainder == "pm" or wordNext == "pm" or remainder == "p.m." or wordNext == "p.m."): strHH = strNum remainder = "pm" used = 1 elif (remainder == "am" or wordNext == "am" or remainder == "a.m." or wordNext == "a.m."): strHH = strNum remainder = "am" used = 1 else: if wordNext == "stund" and int(word) < 100: # "in 3 hours" hrOffset = int(word) used = 2 isTime = False hrAbs = -1 minAbs = -1 elif wordNext == "minut": # "in 10 minutes" minOffset = int(word) used = 2 isTime = False hrAbs = -1 minAbs = -1 elif wordNext == "sekund": # in 5 seconds secOffset = int(word) used = 2 isTime = False hrAbs = -1 minAbs = -1 elif wordNext == "uhr": strHH = word used += 1 isTime = True if wordNextNext == timeQualifier: strMM = "" if wordNextNext[:10] == "nachmittag": used += 1 remainder = "pm" elif wordNextNext == "am" and wordNextNextNext == \ "nachmittag": used += 2 remainder = "pm" elif wordNextNext[:5] == "abend": used += 1 remainder = "pm" elif wordNextNext == "am" and wordNextNextNext == \ "abend": used += 2 remainder = "pm" elif wordNextNext[:7] == "morgens": used += 1 remainder = "am" elif wordNextNext == "am" and wordNextNextNext == \ "morgen": used += 2 remainder = "am" elif wordNextNext == "nachts": used += 1 if 8 <= int(word) <= 12: remainder = "pm" else: remainder = "am" elif is_numeric(wordNextNext): strMM = wordNextNext used += 1 if wordNextNextNext == timeQualifier: if wordNextNextNext[:10] == "nachmittag": used += 1 remainder = "pm" elif wordNextNextNext == "am" and \ wordNextNextNextNext == "nachmittag": used += 2 remainder = "pm" elif wordNextNextNext[:5] == "abend": used += 1 remainder = "pm" elif wordNextNextNext == "am" and \ wordNextNextNextNext == "abend": used += 2 remainder = "pm" elif wordNextNextNext[:7] == "morgens": used += 1 remainder = "am" elif wordNextNextNext == "am" and \ wordNextNextNextNext == "morgen": used += 2 remainder = "am" elif wordNextNextNext == "nachts": used += 1 if 8 <= int(word) <= 12: remainder = "pm" else: remainder = "am" elif wordNext == timeQualifier: strHH = word strMM = 00 isTime = True if wordNext[:10] == "nachmittag": used += 1 remainder = "pm" elif wordNext == "am" and wordNextNext == "nachmittag": used += 2 remainder = "pm" elif wordNext[:5] == "abend": used += 1 remainder = "pm" elif wordNext == "am" and wordNextNext == "abend": used += 2 remainder = "pm" elif wordNext[:7] == "morgens": used += 1 remainder = "am" elif wordNext == "am" and wordNextNext == "morgen": used += 2 remainder = "am" elif wordNext == "nachts": used += 1 if 8 <= int(word) <= 12: remainder = "pm" else: remainder = "am" # if timeQualifier != "": # military = True # else: # isTime = False strHH = int(strHH) if strHH else 0 strMM = int(strMM) if strMM else 0 strHH = strHH + 12 if remainder == "pm" and strHH < 12 else strHH strHH = strHH - 12 if remainder == "am" and strHH >= 12 else strHH if strHH > 24 or strMM > 59: isTime = False used = 0 if isTime: hrAbs = strHH * 1 minAbs = strMM * 1 used += 1 if used > 0: # removed parsed words from the sentence for i in range(used): words[idx + i] = "" if wordPrev == "Uhr": words[words.index(wordPrev)] = "" if wordPrev == u"früh": hrOffset = -1 words[idx - 1] = "" idx -= 1 elif wordPrev == u"spät": hrOffset = 1 words[idx - 1] = "" idx -= 1 if idx > 0 and wordPrev in markers: words[idx - 1] = "" if idx > 1 and wordPrevPrev in markers: words[idx - 2] = "" idx += used - 1 found = True # check that we found a date if not date_found: return None if dayOffset is False: dayOffset = 0 # perform date manipulation extractedDate = dateNow extractedDate = extractedDate.replace(microsecond=0, second=0, minute=0, hour=0) if datestr != "": en_months = [ 'january', 'february', 'march', 'april', 'may', 'june', 'july', 'august', 'september', 'october', 'november', 'december' ] en_monthsShort = [ 'jan', 'feb', 'mar', 'apr', 'may', 'june', 'july', 'aug', 'sept', 'oct', 'nov', 'dec' ] for idx, en_month in enumerate(en_months): datestr = datestr.replace(months[idx], en_month) for idx, en_month in enumerate(en_monthsShort): datestr = datestr.replace(monthsShort[idx], en_month) temp = datetime.strptime(datestr, "%B %d") if not hasYear: temp = temp.replace(year=extractedDate.year) if extractedDate < temp: extractedDate = extractedDate.replace( year=int(currentYear), month=int(temp.strftime("%m")), day=int(temp.strftime("%d"))) else: extractedDate = extractedDate.replace( year=int(currentYear) + 1, month=int(temp.strftime("%m")), day=int(temp.strftime("%d"))) else: extractedDate = extractedDate.replace( year=int(temp.strftime("%Y")), month=int(temp.strftime("%m")), day=int(temp.strftime("%d"))) if timeStr != "": temp = datetime(timeStr) extractedDate = extractedDate.replace(hour=temp.strftime("%H"), minute=temp.strftime("%M"), second=temp.strftime("%S")) if yearOffset != 0: extractedDate = extractedDate + relativedelta(years=yearOffset) if monthOffset != 0: extractedDate = extractedDate + relativedelta(months=monthOffset) if dayOffset != 0: extractedDate = extractedDate + relativedelta(days=dayOffset) if hrAbs is None and minAbs is None and default_time: hrAbs = default_time.hour minAbs = default_time.minute if hrAbs != -1 and minAbs != -1: extractedDate = extractedDate + relativedelta(hours=hrAbs or 0, minutes=minAbs or 0) if (hrAbs or minAbs) and datestr == "": if not daySpecified and dateNow > extractedDate: extractedDate = extractedDate + relativedelta(days=1) if hrOffset != 0: extractedDate = extractedDate + relativedelta(hours=hrOffset) if minOffset != 0: extractedDate = extractedDate + relativedelta(minutes=minOffset) if secOffset != 0: extractedDate = extractedDate + relativedelta(seconds=secOffset) for idx, word in enumerate(words): if words[idx] == "und" and words[idx - 1] == "" \ and words[idx + 1] == "": words[idx] = "" resultStr = " ".join(words) resultStr = ' '.join(resultStr.split()) return [extractedDate, resultStr]
def extractnumber_it(text): """ Questa funzione prepara il testo dato per l'analisi rendendo numeri testuali come interi o frazioni. In italiano non è un modo abituale ma può essere interessante per Mycroft E' la versione portoghese riadattata in italiano args: text (str): la stringa da normalizzare Ritorna: (int) o (float): il valore del numero estratto """ aWords = text.split() count = 0 result = None while count < len(aWords): val = 0 word = aWords[count] next_next_word = None if count + 1 < len(aWords): next_word = aWords[count + 1] if count + 2 < len(aWords): next_next_word = aWords[count + 2] else: next_word = None # is current word a number? if word in it_numbers: if word == "mila": val = it_numbers[word] val = result * val result = 0 else: val = it_numbers[word] elif word.isdigit(): # doesn't work with decimals val = int(word) elif is_numeric(word): val = float(word) elif isFractional_it(word): if not result: result = 1 result = result * isFractional_it(word) # "un terzo" is 1/3 but "il terzo" is 3 if aWords[count - 1] == "il": result = 1.0 // isFractional_it(word) count += 1 continue if not val: # look for fractions like "2/3" aPieces = word.split('/') # if (len(aPieces) == 2 and is_numeric(aPieces[0]) # and is_numeric(aPieces[1])): if look_for_fractions(aPieces): val = float(aPieces[0]) / float(aPieces[1]) if not val: # cerca numero composto come ventuno ventitre centoventi" val = extractnumber_long_it(word) if val: if result is None: result = 0 # handle fractions # if next_word != "avos": result += val # else: # result = float(result) / float(val) if next_word is None: break # number word and fraction ands = ["e"] if next_word in ands: zeros = 0 if result is None: count += 1 continue newWords = aWords[count + 2:] newText = "" for word in newWords: newText += word + " " afterAndVal = extractnumber_it(newText[:-1]) if afterAndVal: if result < afterAndVal or result < 20: while afterAndVal > 1: afterAndVal = afterAndVal / 10.0 for word in newWords: if word == "zero" or word == "0": zeros += 1 else: break for _ in range(0, zeros): afterAndVal = afterAndVal / 10.0 result += afterAndVal break elif next_next_word is not None: if next_next_word in ands: newWords = aWords[count + 3:] newText = "" for word in newWords: newText += word + " " afterAndVal = extractnumber_it(newText[:-1]) if afterAndVal: if result is None: result = 0 result += afterAndVal break decimals = ["punto", "virgola", ".", ","] if next_word in decimals: zeros = 0 newWords = aWords[count + 2:] newText = "" for word in newWords: newText += word + " " for word in newWords: if word == "zero" or word == "0": zeros += 1 else: break afterDotVal = str(extractnumber_it(newText[:-1])) afterDotVal = zeros * "0" + afterDotVal result = float(str(result) + "." + afterDotVal) break count += 1 if result is None: return False # Return the $str with the number related words removed # (now empty strings, so strlen == 0) # aWords = [word for word in aWords if len(word) > 0] # text = ' '.join(aWords) if "." in str(result): integer, dec = str(result).split(".") # cast float to int if dec == "0": result = int(integer) return result
def extractnumber_it(text): """ Questa funzione prepara il testo dato per l'analisi rendendo numeri testuali come interi o frazioni. In italiano non è un modo abituale ma può essere interessante per Mycroft E' la versione portoghese riadattata in italiano args: text (str): la stringa da normalizzare Ritorna: (int) o (float): il valore del numero estratto """ aWords = text.split() count = 0 result = None while count < len(aWords): val = 0 word = aWords[count] next_next_word = None if count + 1 < len(aWords): next_word = aWords[count + 1] if count + 2 < len(aWords): next_next_word = aWords[count + 2] else: next_word = None # is current word a number? if word in it_numbers: if word == "mila": val = it_numbers[word] val = result * val result = 0 else: val = it_numbers[word] elif word.isdigit(): # doesn't work with decimals val = int(word) elif is_numeric(word): val = float(word) elif isFractional_it(word): if not result: result = 1 result = result * isFractional_it(word) # "un terzo" is 1/3 but "il terzo" is 3 if aWords[count - 1] == "il": result = 1.0 // isFractional_it(word) count += 1 continue if not val: # look for fractions like "2/3" aPieces = word.split('/') # if (len(aPieces) == 2 and is_numeric(aPieces[0]) # and is_numeric(aPieces[1])): if look_for_fractions(aPieces): val = float(aPieces[0]) / float(aPieces[1]) if not val: # cerca numero composto come ventuno ventitre centoventi" val = extractnumber_long_it(word) if val: if result is None: result = 0 # handle fractions # if next_word != "avos": result += val # else: # result = float(result) / float(val) if next_word is None: break # number word and fraction ands = ["e"] if next_word in ands: zeros = 0 if result is None: count += 1 continue newWords = aWords[count + 2:] newText = "" for word in newWords: newText += word + " " afterAndVal = extractnumber_it(newText[:-1]) if afterAndVal: if result < afterAndVal or result < 20: while afterAndVal > 1: afterAndVal = afterAndVal / 10.0 for word in newWords: if word == "zero" or word == "0": zeros += 1 else: break for _ in range(0, zeros): afterAndVal = afterAndVal / 10.0 result += afterAndVal break elif next_next_word is not None: if next_next_word in ands: newWords = aWords[count + 3:] newText = "" for word in newWords: newText += word + " " afterAndVal = extractnumber_it(newText[:-1]) if afterAndVal: if result is None: result = 0 result += afterAndVal break decimals = ["punto", "virgola", ".", ","] if next_word in decimals: zeros = 0 newWords = aWords[count + 2:] newText = "" for word in newWords: newText += word + " " for word in newWords: if word == "zero" or word == "0": zeros += 1 else: break afterDotVal = str(extractnumber_it(newText[:-1])) afterDotVal = zeros * "0" + afterDotVal result = float(str(result) + "." + afterDotVal) break count += 1 if result is None: return False # Return the $str with the number related words removed # (now empty strings, so strlen == 0) # aWords = [word for word in aWords if len(word) > 0] # text = ' '.join(aWords) if "." in str(result): integer, dec = str(result).split(".") # cast float to int if dec == "0": result = int(integer) return result
def extract_datetime_en(string, dateNow, default_time): """ Convert a human date reference into an exact datetime Convert things like "today" "tomorrow afternoon" "next Tuesday at 4pm" "August 3rd" into a datetime. If a reference date is not provided, the current local time is used. Also consumes the words used to define the date returning the remaining string. For example, the string "what is Tuesday's weather forecast" returns the date for the forthcoming Tuesday relative to the reference date and the remainder string "what is weather forecast". Args: string (str): string containing date words dateNow (datetime): A reference date/time for "tommorrow", etc default_time (time): Time to set if no time was found in the string Returns: [datetime, str]: An array containing the datetime and the remaining text not consumed in the parsing, or None if no date or time related text was found. """ def clean_string(s): # clean unneeded punctuation and capitalization among other things. s = s.lower().replace('?', '').replace('.', '').replace(',', '') \ .replace(' the ', ' ').replace(' a ', ' ').replace(' an ', ' ') \ .replace("o' clock", "o'clock").replace("o clock", "o'clock") \ .replace("o ' clock", "o'clock").replace("o 'clock", "o'clock") \ .replace("oclock", "o'clock").replace("couple", "2") \ .replace("centuries", "century").replace("decades", "decade") \ .replace("millenniums", "millennium") wordList = s.split() for idx, word in enumerate(wordList): word = word.replace("'s", "") ordinals = ["rd", "st", "nd", "th"] if word[0].isdigit(): for ordinal in ordinals: # "second" is the only case we should not do this if ordinal in word and "second" not in word: word = word.replace(ordinal, "") wordList[idx] = word return wordList def date_found(): return found or \ ( datestr != "" or yearOffset != 0 or monthOffset != 0 or dayOffset is True or hrOffset != 0 or hrAbs or minOffset != 0 or minAbs or secOffset != 0 ) if string == "" or not dateNow: return None found = False daySpecified = False dayOffset = False monthOffset = 0 yearOffset = 0 today = dateNow.strftime("%w") currentYear = dateNow.strftime("%Y") fromFlag = False datestr = "" hasYear = False timeQualifier = "" timeQualifiersAM = ['morning'] timeQualifiersPM = ['afternoon', 'evening', 'tonight', 'night'] timeQualifiersList = set(timeQualifiersAM + timeQualifiersPM) markers = ['at', 'in', 'on', 'by', 'this', 'around', 'for', 'of', "within"] days = ['monday', 'tuesday', 'wednesday', 'thursday', 'friday', 'saturday', 'sunday'] months = ['january', 'february', 'march', 'april', 'may', 'june', 'july', 'august', 'september', 'october', 'november', 'december'] monthsShort = ['jan', 'feb', 'mar', 'apr', 'may', 'june', 'july', 'aug', 'sept', 'oct', 'nov', 'dec'] year_multiples = ["decade", "century", "millennium"] day_multiples = ["weeks", "months", "years"] words = clean_string(string) for idx, word in enumerate(words): if word == "": continue wordPrevPrev = words[idx - 2] if idx > 1 else "" wordPrev = words[idx - 1] if idx > 0 else "" wordNext = words[idx + 1] if idx + 1 < len(words) else "" wordNextNext = words[idx + 2] if idx + 2 < len(words) else "" # this isn't in clean string because I don't want to save back to words word = word.rstrip('s') start = idx used = 0 # save timequalifier for later if word == "now" and not datestr: resultStr = " ".join(words[idx + 1:]) resultStr = ' '.join(resultStr.split()) extractedDate = dateNow.replace(microsecond=0) return [extractedDate, resultStr] elif wordNext in year_multiples: multiplier = None if is_numeric(word): multiplier = extractnumber_en(word) multiplier = multiplier or 1 multiplier = int(multiplier) used += 2 if wordNext == "decade": yearOffset = multiplier * 10 elif wordNext == "century": yearOffset = multiplier * 100 elif wordNext == "millennium": yearOffset = multiplier * 1000 # couple of elif word == "2" and wordNext == "of" and \ wordNextNext in year_multiples: multiplier = 2 used += 3 if wordNextNext == "decade": yearOffset = multiplier * 10 elif wordNextNext == "century": yearOffset = multiplier * 100 elif wordNextNext == "millennium": yearOffset = multiplier * 1000 elif word == "2" and wordNext == "of" and \ wordNextNext in day_multiples: multiplier = 2 used += 3 if wordNextNext == "years": yearOffset = multiplier elif wordNextNext == "months": monthOffset = multiplier elif wordNextNext == "weeks": dayOffset = multiplier * 7 elif word in timeQualifiersList: timeQualifier = word # parse today, tomorrow, day after tomorrow elif word == "today" and not fromFlag: dayOffset = 0 used += 1 elif word == "tomorrow" and not fromFlag: dayOffset = 1 used += 1 elif (word == "day" and wordNext == "after" and wordNextNext == "tomorrow" and not fromFlag and not wordPrev[0].isdigit()): dayOffset = 2 used = 3 if wordPrev == "the": start -= 1 used += 1 # parse 5 days, 10 weeks, last week, next week elif word == "day": if wordPrev[0].isdigit(): dayOffset += int(wordPrev) start -= 1 used = 2 elif word == "week" and not fromFlag: if wordPrev[0].isdigit(): dayOffset += int(wordPrev) * 7 start -= 1 used = 2 elif wordPrev == "next": dayOffset = 7 start -= 1 used = 2 elif wordPrev == "last": dayOffset = -7 start -= 1 used = 2 # parse 10 months, next month, last month elif word == "month" and not fromFlag: if wordPrev[0].isdigit(): monthOffset = int(wordPrev) start -= 1 used = 2 elif wordPrev == "next": monthOffset = 1 start -= 1 used = 2 elif wordPrev == "last": monthOffset = -1 start -= 1 used = 2 # parse 5 years, next year, last year elif word == "year" and not fromFlag: if wordPrev[0].isdigit(): yearOffset = int(wordPrev) start -= 1 used = 2 elif wordPrev == "next": yearOffset = 1 start -= 1 used = 2 elif wordPrev == "last": yearOffset = -1 start -= 1 used = 2 # parse Monday, Tuesday, etc., and next Monday, # last Tuesday, etc. elif word in days and not fromFlag: d = days.index(word) dayOffset = (d + 1) - int(today) used = 1 if dayOffset < 0: dayOffset += 7 if wordPrev == "next": dayOffset += 7 used += 1 start -= 1 elif wordPrev == "last": dayOffset -= 7 used += 1 start -= 1 # parse 15 of July, June 20th, Feb 18, 19 of February elif word in months or word in monthsShort and not fromFlag: try: m = months.index(word) except ValueError: m = monthsShort.index(word) used += 1 datestr = months[m] if wordPrev and (wordPrev[0].isdigit() or (wordPrev == "of" and wordPrevPrev[0].isdigit())): if wordPrev == "of" and wordPrevPrev[0].isdigit(): datestr += " " + words[idx - 2] used += 1 start -= 1 else: datestr += " " + wordPrev start -= 1 used += 1 if wordNext and wordNext[0].isdigit(): datestr += " " + wordNext used += 1 hasYear = True else: hasYear = False elif wordNext and wordNext[0].isdigit(): datestr += " " + wordNext used += 1 if wordNextNext and wordNextNext[0].isdigit(): datestr += " " + wordNextNext used += 1 hasYear = True else: hasYear = False # parse 5 days from tomorrow, 10 weeks from next thursday, # 2 months from July validFollowups = days + months + monthsShort validFollowups.append("today") validFollowups.append("tomorrow") validFollowups.append("next") validFollowups.append("last") validFollowups.append("now") if (word == "from" or word == "after") and wordNext in validFollowups: used = 2 fromFlag = True if wordNext == "tomorrow": dayOffset += 1 elif wordNext in days: d = days.index(wordNext) tmpOffset = (d + 1) - int(today) used = 2 if tmpOffset < 0: tmpOffset += 7 dayOffset += tmpOffset elif wordNextNext and wordNextNext in days: d = days.index(wordNextNext) tmpOffset = (d + 1) - int(today) used = 3 if wordNext == "next": tmpOffset += 7 used += 1 start -= 1 elif wordNext == "last": tmpOffset -= 7 used += 1 start -= 1 dayOffset += tmpOffset if used > 0: if start - 1 > 0 and words[start - 1] == "this": start -= 1 used += 1 for i in range(0, used): words[i + start] = "" if start - 1 >= 0 and words[start - 1] in markers: words[start - 1] = "" found = True daySpecified = True # parse time hrOffset = 0 minOffset = 0 secOffset = 0 hrAbs = None minAbs = None military = False for idx, word in enumerate(words): if word == "": continue wordPrevPrev = words[idx - 2] if idx > 1 else "" wordPrev = words[idx - 1] if idx > 0 else "" wordNext = words[idx + 1] if idx + 1 < len(words) else "" wordNextNext = words[idx + 2] if idx + 2 < len(words) else "" # parse noon, midnight, morning, afternoon, evening used = 0 if word == "noon": hrAbs = 12 used += 1 elif word == "midnight": hrAbs = 0 used += 1 elif word == "morning": if hrAbs is None: hrAbs = 8 used += 1 elif word == "afternoon": if hrAbs is None: hrAbs = 15 used += 1 elif word == "evening": if hrAbs is None: hrAbs = 19 used += 1 # couple of time_unit elif word == "2" and wordNext == "of" and \ wordNextNext in ["hours", "minutes", "seconds"]: used += 3 if wordNextNext == "hours": hrOffset = 2 elif wordNextNext == "minutes": minOffset = 2 elif wordNextNext == "seconds": secOffset = 2 # parse half an hour, quarter hour elif word == "hour" and \ (wordPrev in markers or wordPrevPrev in markers): if wordPrev == "half": minOffset = 30 elif wordPrev == "quarter": minOffset = 15 elif wordPrevPrev == "quarter": minOffset = 15 if idx > 2 and words[idx - 3] in markers: words[idx - 3] = "" if words[idx - 3] == "this": daySpecified = True words[idx - 2] = "" elif wordPrev == "within": hrOffset = 1 else: hrOffset = 1 if wordPrevPrev in markers: words[idx - 2] = "" if wordPrevPrev == "this": daySpecified = True words[idx - 1] = "" used += 1 hrAbs = -1 minAbs = -1 # parse 5:00 am, 12:00 p.m., etc # parse in a minute elif word == "minute" and wordPrev == "in": minOffset = 1 words[idx - 1] = "" used += 1 # parse in a second elif word == "second" and wordPrev == "in": secOffset = 1 words[idx - 1] = "" used += 1 elif word[0].isdigit(): isTime = True strHH = "" strMM = "" remainder = "" if ':' in word: # parse colons # "3:00 in the morning" stage = 0 length = len(word) for i in range(length): if stage == 0: if word[i].isdigit(): strHH += word[i] elif word[i] == ":": stage = 1 else: stage = 2 i -= 1 elif stage == 1: if word[i].isdigit(): strMM += word[i] else: stage = 2 i -= 1 elif stage == 2: remainder = word[i:].replace(".", "") break if remainder == "": nextWord = wordNext.replace(".", "") if nextWord == "am" or nextWord == "pm": remainder = nextWord used += 1 elif nextWord == "tonight": remainder = "pm" used += 1 elif wordNext == "in" and wordNextNext == "the" and \ words[idx + 3] == "morning": remainder = "am" used += 3 elif wordNext == "in" and wordNextNext == "the" and \ words[idx + 3] == "afternoon": remainder = "pm" used += 3 elif wordNext == "in" and wordNextNext == "the" and \ words[idx + 3] == "evening": remainder = "pm" used += 3 elif wordNext == "in" and wordNextNext == "morning": remainder = "am" used += 2 elif wordNext == "in" and wordNextNext == "afternoon": remainder = "pm" used += 2 elif wordNext == "in" and wordNextNext == "evening": remainder = "pm" used += 2 elif wordNext == "this" and wordNextNext == "morning": remainder = "am" used = 2 daySpecified = True elif wordNext == "this" and wordNextNext == "afternoon": remainder = "pm" used = 2 daySpecified = True elif wordNext == "this" and wordNextNext == "evening": remainder = "pm" used = 2 daySpecified = True elif wordNext == "at" and wordNextNext == "night": if strHH and int(strHH) > 5: remainder = "pm" else: remainder = "am" used += 2 else: if timeQualifier != "": military = True if strHH and int(strHH) <= 12 and \ (timeQualifier in timeQualifiersPM): strHH += str(int(strHH) + 12) else: # try to parse numbers without colons # 5 hours, 10 minutes etc. length = len(word) strNum = "" remainder = "" for i in range(length): if word[i].isdigit(): strNum += word[i] else: remainder += word[i] if remainder == "": remainder = wordNext.replace(".", "").lstrip().rstrip() if ( remainder == "pm" or wordNext == "pm" or remainder == "p.m." or wordNext == "p.m."): strHH = strNum remainder = "pm" used = 1 elif ( remainder == "am" or wordNext == "am" or remainder == "a.m." or wordNext == "a.m."): strHH = strNum remainder = "am" used = 1 else: if ( int(strNum) > 100 and ( wordPrev == "o" or wordPrev == "oh" )): # 0800 hours (pronounced oh-eight-hundred) strHH = str(int(strNum) // 100) strMM = str(int(strNum) % 100) military = True if wordNext == "hours": used += 1 elif ( (wordNext == "hours" or wordNext == "hour" or remainder == "hours" or remainder == "hour") and word[0] != '0' and ( int(strNum) < 100 or int(strNum) > 2400 )): # ignores military time # "in 3 hours" hrOffset = int(strNum) used = 2 isTime = False hrAbs = -1 minAbs = -1 elif wordNext == "minutes" or wordNext == "minute" or \ remainder == "minutes" or remainder == "minute": # "in 10 minutes" minOffset = int(strNum) used = 2 isTime = False hrAbs = -1 minAbs = -1 elif wordNext == "seconds" or wordNext == "second" \ or remainder == "seconds" or remainder == "second": # in 5 seconds secOffset = int(strNum) used = 2 isTime = False hrAbs = -1 minAbs = -1 elif int(strNum) > 100: # military time, eg. "3300 hours" strHH = str(int(strNum) // 100) strMM = str(int(strNum) % 100) military = True if wordNext == "hours" or wordNext == "hour" or \ remainder == "hours" or remainder == "hour": used += 1 elif wordNext and wordNext[0].isdigit(): # military time, e.g. "04 38 hours" strHH = strNum strMM = wordNext military = True used += 1 if (wordNextNext == "hours" or wordNextNext == "hour" or remainder == "hours" or remainder == "hour"): used += 1 elif ( wordNext == "" or wordNext == "o'clock" or ( wordNext == "in" and ( wordNextNext == "the" or wordNextNext == timeQualifier ) )): strHH = strNum strMM = "00" if wordNext == "o'clock": used += 1 if wordNext == "in" or wordNextNext == "in": used += (1 if wordNext == "in" else 2) wordNextNextNext = words[idx + 3] \ if idx + 3 < len(words) else "" if (wordNextNext and (wordNextNext in timeQualifier or wordNextNextNext in timeQualifier)): if (wordNextNext in timeQualifiersPM or wordNextNextNext in timeQualifiersPM): remainder = "pm" used += 1 if (wordNextNext in timeQualifiersAM or wordNextNextNext in timeQualifiersAM): remainder = "am" used += 1 if timeQualifier != "": used += 1 # TODO: Unsure if this is 100% accurate military = True else: isTime = False HH = int(strHH) if strHH else 0 MM = int(strMM) if strMM else 0 HH = HH + 12 if remainder == "pm" and HH < 12 else HH HH = HH - 12 if remainder == "am" and HH >= 12 else HH if (not military and remainder not in ['am', 'pm', 'hours', 'minutes', "second", "seconds", "hour", "minute"] and ((not daySpecified) or dayOffset < 1)): # ambiguous time, detect whether they mean this evening or # the next morning based on whether it has already passed if dateNow.hour < HH: pass # No modification needed elif dateNow.hour < HH + 12: HH += 12 else: # has passed, assume the next morning dayOffset += 1 if timeQualifier in timeQualifiersPM and HH < 12: HH += 12 if HH > 24 or MM > 59: isTime = False used = 0 if isTime: hrAbs = HH minAbs = MM used += 1 if used > 0: # removed parsed words from the sentence for i in range(used): if idx + i >= len(words): break words[idx + i] = "" if wordPrev == "o" or wordPrev == "oh": words[words.index(wordPrev)] = "" if wordPrev == "early": hrOffset = -1 words[idx - 1] = "" idx -= 1 elif wordPrev == "late": hrOffset = 1 words[idx - 1] = "" idx -= 1 if idx > 0 and wordPrev in markers: words[idx - 1] = "" if wordPrev == "this": daySpecified = True if idx > 1 and wordPrevPrev in markers: words[idx - 2] = "" if wordPrevPrev == "this": daySpecified = True idx += used - 1 found = True # check that we found a date if not date_found: return None if dayOffset is False: dayOffset = 0 # perform date manipulation extractedDate = dateNow.replace(microsecond=0) if datestr != "": # date included an explicit date, e.g. "june 5" or "june 2, 2017" try: temp = datetime.strptime(datestr, "%B %d") except ValueError: # Try again, allowing the year temp = datetime.strptime(datestr, "%B %d %Y") extractedDate = extractedDate.replace(hour=0, minute=0, second=0) if not hasYear: temp = temp.replace(year=extractedDate.year, tzinfo=extractedDate.tzinfo) if extractedDate < temp: extractedDate = extractedDate.replace( year=int(currentYear), month=int(temp.strftime("%m")), day=int(temp.strftime("%d")), tzinfo=extractedDate.tzinfo) else: extractedDate = extractedDate.replace( year=int(currentYear) + 1, month=int(temp.strftime("%m")), day=int(temp.strftime("%d")), tzinfo=extractedDate.tzinfo) else: extractedDate = extractedDate.replace( year=int(temp.strftime("%Y")), month=int(temp.strftime("%m")), day=int(temp.strftime("%d")), tzinfo=extractedDate.tzinfo) else: # ignore the current HH:MM:SS if relative using days or greater if hrOffset == 0 and minOffset == 0 and secOffset == 0: extractedDate = extractedDate.replace(hour=0, minute=0, second=0) if yearOffset != 0: extractedDate = extractedDate + relativedelta(years=yearOffset) if monthOffset != 0: extractedDate = extractedDate + relativedelta(months=monthOffset) if dayOffset != 0: extractedDate = extractedDate + relativedelta(days=dayOffset) if hrAbs != -1 and minAbs != -1: # If no time was supplied in the string set the time to default # time if it's available if hrAbs is None and minAbs is None and default_time is not None: hrAbs, minAbs = default_time.hour, default_time.minute else: hrAbs = hrAbs or 0 minAbs = minAbs or 0 extractedDate = extractedDate + relativedelta(hours=hrAbs, minutes=minAbs) if (hrAbs != 0 or minAbs != 0) and datestr == "": if not daySpecified and dateNow > extractedDate: extractedDate = extractedDate + relativedelta(days=1) if hrOffset != 0: extractedDate = extractedDate + relativedelta(hours=hrOffset) if minOffset != 0: extractedDate = extractedDate + relativedelta(minutes=minOffset) if secOffset != 0: extractedDate = extractedDate + relativedelta(seconds=secOffset) for idx, word in enumerate(words): if words[idx] == "and" and \ words[idx - 1] == "" and words[idx + 1] == "": words[idx] = "" resultStr = " ".join(words) resultStr = ' '.join(resultStr.split()) return [extractedDate, resultStr]
def extractnumber_en(text): """ This function prepares the given text for parsing by making numbers consistent, getting rid of contractions, etc. Args: text (str): the string to normalize Returns: (int) or (float): The value of extracted number """ aWords = text.split() aWords = [word for word in aWords if word not in ["the", "a", "an"]] and_pass = False valPreAnd = False val = False count = 0 while count < len(aWords): word = aWords[count] if is_numeric(word): # if word.isdigit(): # doesn't work with decimals val = float(word) elif word == "first": val = 1 elif word == "second": val = 2 elif isFractional_en(word): val = isFractional_en(word) else: if word == "one": val = 1 elif word == "two": val = 2 elif word == "three": val = 3 elif word == "four": val = 4 elif word == "five": val = 5 elif word == "six": val = 6 elif word == "seven": val = 7 elif word == "eight": val = 8 elif word == "nine": val = 9 elif word == "ten": val = 10 if val: if count < (len(aWords) - 1): wordNext = aWords[count + 1] else: wordNext = "" valNext = isFractional_en(wordNext) if valNext: val = val * valNext aWords[count + 1] = "" # if val == False: if not val: # look for fractions like "2/3" aPieces = word.split('/') # if (len(aPieces) == 2 and is_numeric(aPieces[0]) # and is_numeric(aPieces[1])): if look_for_fractions(aPieces): val = float(aPieces[0]) / float(aPieces[1]) elif and_pass: # added to value, quit here val = valPreAnd break else: count += 1 continue aWords[count] = "" if and_pass: aWords[count - 1] = '' # remove "and" val += valPreAnd elif count + 1 < len(aWords) and aWords[count + 1] == 'and': and_pass = True valPreAnd = val val = False count += 2 continue elif count + 2 < len(aWords) and aWords[count + 2] == 'and': and_pass = True valPreAnd = val val = False count += 3 continue break # if val == False: if not val: return False # Return the string with the number related words removed # (now empty strings, so strlen == 0) aWords = [word for word in aWords if len(word) > 0] text = ' '.join(aWords) return val
def extract_datetime_da(string, currentDate, default_time): def clean_string(s): """ cleans the input string of unneeded punctuation and capitalization among other things. 'am' is a preposition, so cannot currently be used for 12 hour date format """ s = s.lower().replace('?', '').replace('.', '').replace(',', '') \ .replace(' den ', ' ').replace(' det ', ' ').replace(' om ', ' ').replace( ' om ', ' ') \ .replace(' på ', ' ').replace(' om ', ' ') wordList = s.split() for idx, word in enumerate(wordList): if isOrdinal_da(word) is not False: word = str(isOrdinal_da(word)) wordList[idx] = word return wordList def date_found(): return found or \ ( datestr != "" or timeStr != "" or yearOffset != 0 or monthOffset != 0 or dayOffset is True or hrOffset != 0 or hrAbs or minOffset != 0 or minAbs or secOffset != 0 ) if string == "" or not currentDate: return None found = False daySpecified = False dayOffset = False monthOffset = 0 yearOffset = 0 dateNow = currentDate today = dateNow.strftime("%w") currentYear = dateNow.strftime("%Y") fromFlag = False datestr = "" hasYear = False timeQualifier = "" timeQualifiersList = ['tidlig', 'morgen', 'morgenen', 'formidag', 'formiddagen', 'eftermiddag', 'eftermiddagen', 'aften', 'aftenen', 'nat', 'natten'] markers = ['i', 'om', 'på', 'klokken', 'ved'] days = ['mandag', 'tirsdag', 'onsdag', 'torsdag', 'fredag', 'lørdag', 'søndag'] months = ['januar', 'februar', 'marts', 'april', 'maj', 'juni', 'juli', 'august', 'september', 'oktober', 'november', 'desember'] monthsShort = ['jan', 'feb', 'mar', 'apr', 'maj', 'juni', 'juli', 'aug', 'sep', 'okt', 'nov', 'des'] validFollowups = days + months + monthsShort validFollowups.append("i dag") validFollowups.append("morgen") validFollowups.append("næste") validFollowups.append("forige") validFollowups.append("nu") words = clean_string(string) for idx, word in enumerate(words): if word == "": continue wordPrevPrev = words[idx - 2] if idx > 1 else "" wordPrev = words[idx - 1] if idx > 0 else "" wordNext = words[idx + 1] if idx + 1 < len(words) else "" wordNextNext = words[idx + 2] if idx + 2 < len(words) else "" start = idx used = 0 # save timequalifier for later if word in timeQualifiersList: timeQualifier = word # parse today, tomorrow, day after tomorrow elif word == "dag" and not fromFlag: dayOffset = 0 used += 1 elif word == "morgen" and not fromFlag and wordPrev != "om" and \ wordPrev not in days: # morgen means tomorrow if not "am # Morgen" and not [day of the week] morgen dayOffset = 1 used += 1 elif word == "overmorgen" and not fromFlag: dayOffset = 2 used += 1 # parse 5 days, 10 weeks, last week, next week elif word == "dag" or word == "dage": if wordPrev[0].isdigit(): dayOffset += int(wordPrev) start -= 1 used = 2 elif word == "uge" or word == "uger" and not fromFlag: if wordPrev[0].isdigit(): dayOffset += int(wordPrev) * 7 start -= 1 used = 2 elif wordPrev[:6] == "næste": dayOffset = 7 start -= 1 used = 2 elif wordPrev[:5] == "forige": dayOffset = -7 start -= 1 used = 2 # parse 10 months, next month, last month elif word == "måned" and not fromFlag: if wordPrev[0].isdigit(): monthOffset = int(wordPrev) start -= 1 used = 2 elif wordPrev[:6] == "næste": monthOffset = 1 start -= 1 used = 2 elif wordPrev[:5] == "forige": monthOffset = -1 start -= 1 used = 2 # parse 5 years, next year, last year elif word == "år" and not fromFlag: if wordPrev[0].isdigit(): yearOffset = int(wordPrev) start -= 1 used = 2 elif wordPrev[:6] == " næste": yearOffset = 1 start -= 1 used = 2 elif wordPrev[:6] == "næste": yearOffset = -1 start -= 1 used = 2 # parse Monday, Tuesday, etc., and next Monday, # last Tuesday, etc. elif word in days and not fromFlag: d = days.index(word) dayOffset = (d + 1) - int(today) used = 1 if dayOffset < 0: dayOffset += 7 if wordNext == "morgen": # morgen means morning if preceded by # the day of the week words[idx + 1] = "tidlig" if wordPrev[:6] == "næste": dayOffset += 7 used += 1 start -= 1 elif wordPrev[:5] == "forige": dayOffset -= 7 used += 1 start -= 1 # parse 15 of July, June 20th, Feb 18, 19 of February elif word in months or word in monthsShort and not fromFlag: try: m = months.index(word) except ValueError: m = monthsShort.index(word) used += 1 datestr = months[m] if wordPrev and (wordPrev[0].isdigit() or (wordPrev == "of" and wordPrevPrev[0].isdigit())): if wordPrev == "of" and wordPrevPrev[0].isdigit(): datestr += " " + words[idx - 2] used += 1 start -= 1 else: datestr += " " + wordPrev start -= 1 used += 1 if wordNext and wordNext[0].isdigit(): datestr += " " + wordNext used += 1 hasYear = True else: hasYear = False elif wordNext and wordNext[0].isdigit(): datestr += " " + wordNext used += 1 if wordNextNext and wordNextNext[0].isdigit(): datestr += " " + wordNextNext used += 1 hasYear = True else: hasYear = False # parse 5 days from tomorrow, 10 weeks from next thursday, # 2 months from July if ( word == "fra" or word == "til" or word == "om") and wordNext \ in validFollowups: used = 2 fromFlag = True if wordNext == "morgenen" and \ wordPrev != "om" and \ wordPrev not in days: # morgen means tomorrow if not "am Morgen" and not # [day of the week] morgen: dayOffset += 1 elif wordNext in days: d = days.index(wordNext) tmpOffset = (d + 1) - int(today) used = 2 if tmpOffset < 0: tmpOffset += 7 dayOffset += tmpOffset elif wordNextNext and wordNextNext in days: d = days.index(wordNextNext) tmpOffset = (d + 1) - int(today) used = 3 if wordNext[:6] == "næste": tmpOffset += 7 used += 1 start -= 1 elif wordNext[:5] == "forige": tmpOffset -= 7 used += 1 start -= 1 dayOffset += tmpOffset if used > 0: if start - 1 > 0 and words[start - 1].startswith("denne"): start -= 1 used += 1 for i in range(0, used): words[i + start] = "" if start - 1 >= 0 and words[start - 1] in markers: words[start - 1] = "" found = True daySpecified = True # parse time timeStr = "" hrOffset = 0 minOffset = 0 secOffset = 0 hrAbs = None minAbs = None for idx, word in enumerate(words): if word == "": continue wordPrevPrev = words[idx - 2] if idx > 1 else "" wordPrev = words[idx - 1] if idx > 0 else "" wordNext = words[idx + 1] if idx + 1 < len(words) else "" wordNextNext = words[idx + 2] if idx + 2 < len(words) else "" wordNextNextNext = words[idx + 3] if idx + 3 < len(words) else "" wordNextNextNextNext = words[idx + 4] if idx + 4 < len(words) else "" # parse noon, midnight, morning, afternoon, evening used = 0 if word[:6] == "middag": hrAbs = 12 used += 1 elif word[:11] == "midnat": hrAbs = 0 used += 1 elif word == "morgenen" or ( wordPrev == "om" and word == "morgenen") or word == "tidlig": if not hrAbs: hrAbs = 8 used += 1 elif word[:11] == "eftermiddag": if not hrAbs: hrAbs = 15 used += 1 elif word[:5] == "aften": if not hrAbs: hrAbs = 19 used += 1 # parse half an hour, quarter hour elif word == "time" and \ (wordPrev in markers or wordPrevPrev in markers): if wordPrev[:4] == "halv": minOffset = 30 elif wordPrev == "kvarter": minOffset = 15 elif wordPrev == "trekvarter": minOffset = 45 else: hrOffset = 1 if wordPrevPrev in markers: words[idx - 2] = "" words[idx - 1] = "" used += 1 hrAbs = -1 minAbs = -1 # parse 5:00 am, 12:00 p.m., etc elif word[0].isdigit(): isTime = True strHH = "" strMM = "" remainder = "" if ':' in word: # parse colons # "3:00 in the morning" stage = 0 length = len(word) for i in range(length): if stage == 0: if word[i].isdigit(): strHH += word[i] elif word[i] == ":": stage = 1 else: stage = 2 i -= 1 elif stage == 1: if word[i].isdigit(): strMM += word[i] else: stage = 2 i -= 1 elif stage == 2: remainder = word[i:].replace(".", "") break if remainder == "": nextWord = wordNext.replace(".", "") if nextWord == "am" or nextWord == "pm": remainder = nextWord used += 1 elif nextWord == "aften": remainder = "pm" used += 1 elif wordNext == "om" and wordNextNext == "morgenen": remainder = "am" used += 2 elif wordNext == "om" and wordNextNext == "eftermiddagen": remainder = "pm" used += 2 elif wordNext == "om" and wordNextNext == "aftenen": remainder = "pm" used += 2 elif wordNext == "morgen": remainder = "am" used += 1 elif wordNext == "eftermiddag": remainder = "pm" used += 1 elif wordNext == "aften": remainder = "pm" used += 1 elif wordNext == "i" and wordNextNext == "morgen": remainder = "am" used = 2 elif wordNext == "i" and wordNextNext == "eftermiddag": remainder = "pm" used = 2 elif wordNext == "i" and wordNextNext == "aften": remainder = "pm" used = 2 elif wordNext == "natten": if strHH > 4: remainder = "pm" else: remainder = "am" used += 1 else: if timeQualifier != "": if strHH <= 12 and \ (timeQualifier == "aftenen" or timeQualifier == "eftermiddagen"): strHH += 12 # what happens when strHH is 24? else: # try to parse # s without colons # 5 hours, 10 minutes etc. length = len(word) strNum = "" remainder = "" for i in range(length): if word[i].isdigit(): strNum += word[i] else: remainder += word[i] if remainder == "": remainder = wordNext.replace(".", "").lstrip().rstrip() if ( remainder == "pm" or wordNext == "pm" or remainder == "p.m." or wordNext == "p.m."): strHH = strNum remainder = "pm" used = 1 elif ( remainder == "am" or wordNext == "am" or remainder == "a.m." or wordNext == "a.m."): strHH = strNum remainder = "am" used = 1 else: if wordNext == "time" and int(word) < 100: # "in 3 hours" hrOffset = int(word) used = 2 isTime = False hrAbs = -1 minAbs = -1 elif wordNext == "minut": # "in 10 minutes" minOffset = int(word) used = 2 isTime = False hrAbs = -1 minAbs = -1 elif wordNext == "sekund": # in 5 seconds secOffset = int(word) used = 2 isTime = False hrAbs = -1 minAbs = -1 elif wordNext == "time": strHH = word used += 1 isTime = True if wordNextNext == timeQualifier: strMM = "" if wordNextNext[:11] == "eftermiddag": used += 1 remainder = "pm" elif wordNextNext == "om" and wordNextNextNext == \ "eftermiddagen": used += 2 remainder = "pm" elif wordNextNext[:5] == "aften": used += 1 remainder = "pm" elif wordNextNext == "om" and wordNextNextNext == \ "aftenen": used += 2 remainder = "pm" elif wordNextNext[:6] == "morgen": used += 1 remainder = "am" elif wordNextNext == "om" and wordNextNextNext == \ "morgenen": used += 2 remainder = "am" elif wordNextNext == "natten": used += 1 if 8 <= int(word) <= 12: remainder = "pm" else: remainder = "am" elif is_numeric(wordNextNext): strMM = wordNextNext used += 1 if wordNextNextNext == timeQualifier: if wordNextNextNext[:11] == "eftermiddag": used += 1 remainder = "pm" elif wordNextNextNext == "om" and \ wordNextNextNextNext == \ "eftermiddagen": used += 2 remainder = "pm" elif wordNextNextNext[:6] == "natten": used += 1 remainder = "pm" elif wordNextNextNext == "am" and \ wordNextNextNextNext == "natten": used += 2 remainder = "pm" elif wordNextNextNext[:7] == "morgenen": used += 1 remainder = "am" elif wordNextNextNext == "om" and \ wordNextNextNextNext == "morgenen": used += 2 remainder = "am" elif wordNextNextNext == "natten": used += 1 if 8 <= int(word) <= 12: remainder = "pm" else: remainder = "am" elif wordNext == timeQualifier: strHH = word strMM = 00 isTime = True if wordNext[:10] == "eftermidag": used += 1 remainder = "pm" elif wordNext == "om" and \ wordNextNext == "eftermiddanen": used += 2 remainder = "pm" elif wordNext[:7] == "aftenen": used += 1 remainder = "pm" elif wordNext == "om" and wordNextNext == "aftenen": used += 2 remainder = "pm" elif wordNext[:7] == "morgenen": used += 1 remainder = "am" elif wordNext == "ao" and wordNextNext == "morgenen": used += 2 remainder = "am" elif wordNext == "natten": used += 1 if 8 <= int(word) <= 12: remainder = "pm" else: remainder = "am" # if timeQualifier != "": # military = True # else: # isTime = False strHH = int(strHH) if strHH else 0 strMM = int(strMM) if strMM else 0 strHH = strHH + 12 if remainder == "pm" and strHH < 12 else strHH strHH = strHH - 12 if remainder == "am" and strHH >= 12 else strHH if strHH > 24 or strMM > 59: isTime = False used = 0 if isTime: hrAbs = strHH * 1 minAbs = strMM * 1 used += 1 if used > 0: # removed parsed words from the sentence for i in range(used): words[idx + i] = "" if wordPrev == "tidlig": hrOffset = -1 words[idx - 1] = "" idx -= 1 elif wordPrev == "sen": hrOffset = 1 words[idx - 1] = "" idx -= 1 if idx > 0 and wordPrev in markers: words[idx - 1] = "" if idx > 1 and wordPrevPrev in markers: words[idx - 2] = "" idx += used - 1 found = True # check that we found a date if not date_found: return None if dayOffset is False: dayOffset = 0 # perform date manipulation extractedDate = dateNow extractedDate = extractedDate.replace(microsecond=0, second=0, minute=0, hour=0) if datestr != "": en_months = ['january', 'february', 'march', 'april', 'may', 'june', 'july', 'august', 'september', 'october', 'november', 'december'] en_monthsShort = ['jan', 'feb', 'mar', 'apr', 'may', 'june', 'july', 'aug', 'sept', 'oct', 'nov', 'dec'] for idx, en_month in enumerate(en_months): datestr = datestr.replace(months[idx], en_month) for idx, en_month in enumerate(en_monthsShort): datestr = datestr.replace(monthsShort[idx], en_month) temp = datetime.strptime(datestr, "%B %d") if not hasYear: temp = temp.replace(year=extractedDate.year) if extractedDate < temp: extractedDate = extractedDate.replace(year=int(currentYear), month=int( temp.strftime( "%m")), day=int(temp.strftime( "%d"))) else: extractedDate = extractedDate.replace( year=int(currentYear) + 1, month=int(temp.strftime("%m")), day=int(temp.strftime("%d"))) else: extractedDate = extractedDate.replace( year=int(temp.strftime("%Y")), month=int(temp.strftime("%m")), day=int(temp.strftime("%d"))) if timeStr != "": temp = datetime(timeStr) extractedDate = extractedDate.replace(hour=temp.strftime("%H"), minute=temp.strftime("%M"), second=temp.strftime("%S")) if yearOffset != 0: extractedDate = extractedDate + relativedelta(years=yearOffset) if monthOffset != 0: extractedDate = extractedDate + relativedelta(months=monthOffset) if dayOffset != 0: extractedDate = extractedDate + relativedelta(days=dayOffset) if hrAbs is None and minAbs is None and default_time: hrAbs = default_time.hour minAbs = default_time.minute if hrAbs != -1 and minAbs != -1: extractedDate = extractedDate + relativedelta(hours=hrAbs or 0, minutes=minAbs or 0) if (hrAbs or minAbs) and datestr == "": if not daySpecified and dateNow > extractedDate: extractedDate = extractedDate + relativedelta(days=1) if hrOffset != 0: extractedDate = extractedDate + relativedelta(hours=hrOffset) if minOffset != 0: extractedDate = extractedDate + relativedelta(minutes=minOffset) if secOffset != 0: extractedDate = extractedDate + relativedelta(seconds=secOffset) for idx, word in enumerate(words): if words[idx] == "og" and words[idx - 1] == "" \ and words[idx + 1] == "": words[idx] = "" resultStr = " ".join(words) resultStr = ' '.join(resultStr.split()) return [extractedDate, resultStr]
def extractnumber_da(text): """ This function prepares the given text for parsing by making numbers consistent, getting rid of contractions, etc. Args: text (str): the string to normalize Returns: (int) or (float): The value of extracted number undefined articles cannot be suppressed in German: 'ein Pferd' means 'one horse' and 'a horse' """ aWords = text.split() aWords = [word for word in aWords if word not in ["den", "det"]] and_pass = False valPreAnd = False val = False count = 0 while count < len(aWords): word = aWords[count] if is_numeric(word): if word.isdigit(): # doesn't work with decimals val = float(word) elif isFractional_da(word): val = isFractional_da(word) elif isOrdinal_da(word): val = isOrdinal_da(word) else: if word in da_numbers: val = da_numbers[word] if count < (len(aWords) - 1): wordNext = aWords[count + 1] else: wordNext = "" valNext = isFractional_da(wordNext) if valNext: val = val * valNext aWords[count + 1] = "" if not val: # look for fractions like "2/3" aPieces = word.split('/') # if (len(aPieces) == 2 and is_numeric(aPieces[0]) # and is_numeric(aPieces[1])): if look_for_fractions(aPieces): val = float(aPieces[0]) / float(aPieces[1]) elif and_pass: # added to value, quit here val = valPreAnd break else: count += 1 continue aWords[count] = "" if and_pass: aWords[count - 1] = '' # remove "og" val += valPreAnd elif count + 1 < len(aWords) and aWords[count + 1] == 'og': and_pass = True valPreAnd = val val = False count += 2 continue elif count + 2 < len(aWords) and aWords[count + 2] == 'og': and_pass = True valPreAnd = val val = False count += 3 continue break if not val: return False return val
def extract_datetime_en(string, dateNow, default_time): """ Convert a human date reference into an exact datetime Convert things like "today" "tomorrow afternoon" "next Tuesday at 4pm" "August 3rd" into a datetime. If a reference date is not provided, the current local time is used. Also consumes the words used to define the date returning the remaining string. For example, the string "what is Tuesday's weather forecast" returns the date for the forthcoming Tuesday relative to the reference date and the remainder string "what is weather forecast". Args: string (str): string containing date words dateNow (datetime): A reference date/time for "tommorrow", etc default_time (time): Time to set if no time was found in the string Returns: [datetime, str]: An array containing the datetime and the remaining text not consumed in the parsing, or None if no date or time related text was found. """ def clean_string(s): # clean unneeded punctuation and capitalization among other things. s = s.lower().replace('?', '').replace('.', '').replace(',', '') \ .replace(' the ', ' ').replace(' a ', ' ').replace(' an ', ' ') \ .replace("o' clock", "o'clock").replace("o clock", "o'clock") \ .replace("o ' clock", "o'clock").replace("o 'clock", "o'clock") \ .replace("oclock", "o'clock").replace("couple", "2") \ .replace("centuries", "century").replace("decades", "decade") \ .replace("millenniums", "millennium") wordList = s.split() for idx, word in enumerate(wordList): word = word.replace("'s", "") ordinals = ["rd", "st", "nd", "th"] if word[0].isdigit(): for ordinal in ordinals: # "second" is the only case we should not do this if ordinal in word and "second" not in word: word = word.replace(ordinal, "") wordList[idx] = word return wordList def date_found(): return found or \ ( datestr != "" or yearOffset != 0 or monthOffset != 0 or dayOffset is True or hrOffset != 0 or hrAbs or minOffset != 0 or minAbs or secOffset != 0 ) if string == "" or not dateNow: return None found = False daySpecified = False dayOffset = False monthOffset = 0 yearOffset = 0 today = dateNow.strftime("%w") currentYear = dateNow.strftime("%Y") fromFlag = False datestr = "" hasYear = False timeQualifier = "" timeQualifiersAM = ['morning'] timeQualifiersPM = ['afternoon', 'evening', 'tonight', 'night'] timeQualifiersList = set(timeQualifiersAM + timeQualifiersPM) markers = ['at', 'in', 'on', 'by', 'this', 'around', 'for', 'of', "within"] days = [ 'monday', 'tuesday', 'wednesday', 'thursday', 'friday', 'saturday', 'sunday' ] months = [ 'january', 'february', 'march', 'april', 'may', 'june', 'july', 'august', 'september', 'october', 'november', 'december' ] monthsShort = [ 'jan', 'feb', 'mar', 'apr', 'may', 'june', 'july', 'aug', 'sept', 'oct', 'nov', 'dec' ] year_multiples = ["decade", "century", "millennium"] day_multiples = ["weeks", "months", "years"] words = clean_string(string) for idx, word in enumerate(words): if word == "": continue wordPrevPrev = words[idx - 2] if idx > 1 else "" wordPrev = words[idx - 1] if idx > 0 else "" wordNext = words[idx + 1] if idx + 1 < len(words) else "" wordNextNext = words[idx + 2] if idx + 2 < len(words) else "" # this isn't in clean string because I don't want to save back to words word = word.rstrip('s') start = idx used = 0 # save timequalifier for later if word == "now" and not datestr: resultStr = " ".join(words[idx + 1:]) resultStr = ' '.join(resultStr.split()) extractedDate = dateNow.replace(microsecond=0) return [extractedDate, resultStr] elif wordNext in year_multiples: multiplier = None if is_numeric(word): multiplier = extractnumber_en(word) multiplier = multiplier or 1 multiplier = int(multiplier) used += 2 if wordNext == "decade": yearOffset = multiplier * 10 elif wordNext == "century": yearOffset = multiplier * 100 elif wordNext == "millennium": yearOffset = multiplier * 1000 # couple of elif word == "2" and wordNext == "of" and \ wordNextNext in year_multiples: multiplier = 2 used += 3 if wordNextNext == "decade": yearOffset = multiplier * 10 elif wordNextNext == "century": yearOffset = multiplier * 100 elif wordNextNext == "millennium": yearOffset = multiplier * 1000 elif word == "2" and wordNext == "of" and \ wordNextNext in day_multiples: multiplier = 2 used += 3 if wordNextNext == "years": yearOffset = multiplier elif wordNextNext == "months": monthOffset = multiplier elif wordNextNext == "weeks": dayOffset = multiplier * 7 elif word in timeQualifiersList: timeQualifier = word # parse today, tomorrow, day after tomorrow elif word == "today" and not fromFlag: dayOffset = 0 used += 1 elif word == "tomorrow" and not fromFlag: dayOffset = 1 used += 1 elif (word == "day" and wordNext == "after" and wordNextNext == "tomorrow" and not fromFlag and not wordPrev[0].isdigit()): dayOffset = 2 used = 3 if wordPrev == "the": start -= 1 used += 1 # parse 5 days, 10 weeks, last week, next week elif word == "day": if wordPrev[0].isdigit(): dayOffset += int(wordPrev) start -= 1 used = 2 elif word == "week" and not fromFlag: if wordPrev[0].isdigit(): dayOffset += int(wordPrev) * 7 start -= 1 used = 2 elif wordPrev == "next": dayOffset = 7 start -= 1 used = 2 elif wordPrev == "last": dayOffset = -7 start -= 1 used = 2 # parse 10 months, next month, last month elif word == "month" and not fromFlag: if wordPrev[0].isdigit(): monthOffset = int(wordPrev) start -= 1 used = 2 elif wordPrev == "next": monthOffset = 1 start -= 1 used = 2 elif wordPrev == "last": monthOffset = -1 start -= 1 used = 2 # parse 5 years, next year, last year elif word == "year" and not fromFlag: if wordPrev[0].isdigit(): yearOffset = int(wordPrev) start -= 1 used = 2 elif wordPrev == "next": yearOffset = 1 start -= 1 used = 2 elif wordPrev == "last": yearOffset = -1 start -= 1 used = 2 # parse Monday, Tuesday, etc., and next Monday, # last Tuesday, etc. elif word in days and not fromFlag: d = days.index(word) dayOffset = (d + 1) - int(today) used = 1 if dayOffset < 0: dayOffset += 7 if wordPrev == "next": dayOffset += 7 used += 1 start -= 1 elif wordPrev == "last": dayOffset -= 7 used += 1 start -= 1 # parse 15 of July, June 20th, Feb 18, 19 of February elif word in months or word in monthsShort and not fromFlag: try: m = months.index(word) except ValueError: m = monthsShort.index(word) used += 1 datestr = months[m] if wordPrev and (wordPrev[0].isdigit() or (wordPrev == "of" and wordPrevPrev[0].isdigit())): if wordPrev == "of" and wordPrevPrev[0].isdigit(): datestr += " " + words[idx - 2] used += 1 start -= 1 else: datestr += " " + wordPrev start -= 1 used += 1 if wordNext and wordNext[0].isdigit(): datestr += " " + wordNext used += 1 hasYear = True else: hasYear = False elif wordNext and wordNext[0].isdigit(): datestr += " " + wordNext used += 1 if wordNextNext and wordNextNext[0].isdigit(): datestr += " " + wordNextNext used += 1 hasYear = True else: hasYear = False # parse 5 days from tomorrow, 10 weeks from next thursday, # 2 months from July validFollowups = days + months + monthsShort validFollowups.append("today") validFollowups.append("tomorrow") validFollowups.append("next") validFollowups.append("last") validFollowups.append("now") if (word == "from" or word == "after") and wordNext in validFollowups: used = 2 fromFlag = True if wordNext == "tomorrow": dayOffset += 1 elif wordNext in days: d = days.index(wordNext) tmpOffset = (d + 1) - int(today) used = 2 if tmpOffset < 0: tmpOffset += 7 dayOffset += tmpOffset elif wordNextNext and wordNextNext in days: d = days.index(wordNextNext) tmpOffset = (d + 1) - int(today) used = 3 if wordNext == "next": tmpOffset += 7 used += 1 start -= 1 elif wordNext == "last": tmpOffset -= 7 used += 1 start -= 1 dayOffset += tmpOffset if used > 0: if start - 1 > 0 and words[start - 1] == "this": start -= 1 used += 1 for i in range(0, used): words[i + start] = "" if start - 1 >= 0 and words[start - 1] in markers: words[start - 1] = "" found = True daySpecified = True # parse time hrOffset = 0 minOffset = 0 secOffset = 0 hrAbs = None minAbs = None military = False for idx, word in enumerate(words): if word == "": continue wordPrevPrev = words[idx - 2] if idx > 1 else "" wordPrev = words[idx - 1] if idx > 0 else "" wordNext = words[idx + 1] if idx + 1 < len(words) else "" wordNextNext = words[idx + 2] if idx + 2 < len(words) else "" # parse noon, midnight, morning, afternoon, evening used = 0 if word == "noon": hrAbs = 12 used += 1 elif word == "midnight": hrAbs = 0 used += 1 elif word == "morning": if hrAbs is None: hrAbs = 8 used += 1 elif word == "afternoon": if hrAbs is None: hrAbs = 15 used += 1 elif word == "evening": if hrAbs is None: hrAbs = 19 used += 1 # couple of time_unit elif word == "2" and wordNext == "of" and \ wordNextNext in ["hours", "minutes", "seconds"]: used += 3 if wordNextNext == "hours": hrOffset = 2 elif wordNextNext == "minutes": minOffset = 2 elif wordNextNext == "seconds": secOffset = 2 # parse half an hour, quarter hour elif word == "hour" and \ (wordPrev in markers or wordPrevPrev in markers): if wordPrev == "half": minOffset = 30 elif wordPrev == "quarter": minOffset = 15 elif wordPrevPrev == "quarter": minOffset = 15 if idx > 2 and words[idx - 3] in markers: words[idx - 3] = "" if words[idx - 3] == "this": daySpecified = True words[idx - 2] = "" elif wordPrev == "within": hrOffset = 1 else: hrOffset = 1 if wordPrevPrev in markers: words[idx - 2] = "" if wordPrevPrev == "this": daySpecified = True words[idx - 1] = "" used += 1 hrAbs = -1 minAbs = -1 # parse 5:00 am, 12:00 p.m., etc # parse in a minute elif word == "minute" and wordPrev == "in": minOffset = 1 words[idx - 1] = "" used += 1 # parse in a second elif word == "second" and wordPrev == "in": secOffset = 1 words[idx - 1] = "" used += 1 elif word[0].isdigit(): isTime = True strHH = "" strMM = "" remainder = "" if ':' in word: # parse colons # "3:00 in the morning" stage = 0 length = len(word) for i in range(length): if stage == 0: if word[i].isdigit(): strHH += word[i] elif word[i] == ":": stage = 1 else: stage = 2 i -= 1 elif stage == 1: if word[i].isdigit(): strMM += word[i] else: stage = 2 i -= 1 elif stage == 2: remainder = word[i:].replace(".", "") break if remainder == "": nextWord = wordNext.replace(".", "") if nextWord == "am" or nextWord == "pm": remainder = nextWord used += 1 elif nextWord == "tonight": remainder = "pm" used += 1 elif wordNext == "in" and wordNextNext == "the" and \ words[idx + 3] == "morning": remainder = "am" used += 3 elif wordNext == "in" and wordNextNext == "the" and \ words[idx + 3] == "afternoon": remainder = "pm" used += 3 elif wordNext == "in" and wordNextNext == "the" and \ words[idx + 3] == "evening": remainder = "pm" used += 3 elif wordNext == "in" and wordNextNext == "morning": remainder = "am" used += 2 elif wordNext == "in" and wordNextNext == "afternoon": remainder = "pm" used += 2 elif wordNext == "in" and wordNextNext == "evening": remainder = "pm" used += 2 elif wordNext == "this" and wordNextNext == "morning": remainder = "am" used = 2 daySpecified = True elif wordNext == "this" and wordNextNext == "afternoon": remainder = "pm" used = 2 daySpecified = True elif wordNext == "this" and wordNextNext == "evening": remainder = "pm" used = 2 daySpecified = True elif wordNext == "at" and wordNextNext == "night": if strHH and int(strHH) > 5: remainder = "pm" else: remainder = "am" used += 2 else: if timeQualifier != "": military = True if strHH and int(strHH) <= 12 and \ (timeQualifier in timeQualifiersPM): strHH += str(int(strHH) + 12) else: # try to parse numbers without colons # 5 hours, 10 minutes etc. length = len(word) strNum = "" remainder = "" for i in range(length): if word[i].isdigit(): strNum += word[i] else: remainder += word[i] if remainder == "": remainder = wordNext.replace(".", "").lstrip().rstrip() if (remainder == "pm" or wordNext == "pm" or remainder == "p.m." or wordNext == "p.m."): strHH = strNum remainder = "pm" used = 1 elif (remainder == "am" or wordNext == "am" or remainder == "a.m." or wordNext == "a.m."): strHH = strNum remainder = "am" used = 1 else: if (int(strNum) > 100 and (wordPrev == "o" or wordPrev == "oh")): # 0800 hours (pronounced oh-eight-hundred) strHH = str(int(strNum) // 100) strMM = str(int(strNum) % 100) military = True if wordNext == "hours": used += 1 elif ((wordNext == "hours" or wordNext == "hour" or remainder == "hours" or remainder == "hour") and word[0] != '0' and (int(strNum) < 100 or int(strNum) > 2400)): # ignores military time # "in 3 hours" hrOffset = int(strNum) used = 2 isTime = False hrAbs = -1 minAbs = -1 elif wordNext == "minutes" or wordNext == "minute" or \ remainder == "minutes" or remainder == "minute": # "in 10 minutes" minOffset = int(strNum) used = 2 isTime = False hrAbs = -1 minAbs = -1 elif wordNext == "seconds" or wordNext == "second" \ or remainder == "seconds" or remainder == "second": # in 5 seconds secOffset = int(strNum) used = 2 isTime = False hrAbs = -1 minAbs = -1 elif int(strNum) > 100: # military time, eg. "3300 hours" strHH = str(int(strNum) // 100) strMM = str(int(strNum) % 100) military = True if wordNext == "hours" or wordNext == "hour" or \ remainder == "hours" or remainder == "hour": used += 1 elif wordNext and wordNext[0].isdigit(): # military time, e.g. "04 38 hours" strHH = strNum strMM = wordNext military = True used += 1 if (wordNextNext == "hours" or wordNextNext == "hour" or remainder == "hours" or remainder == "hour"): used += 1 elif (wordNext == "" or wordNext == "o'clock" or (wordNext == "in" and (wordNextNext == "the" or wordNextNext == timeQualifier))): strHH = strNum strMM = "00" if wordNext == "o'clock": used += 1 if wordNext == "in" or wordNextNext == "in": used += (1 if wordNext == "in" else 2) wordNextNextNext = words[idx + 3] \ if idx + 3 < len(words) else "" if (wordNextNext and (wordNextNext in timeQualifier or wordNextNextNext in timeQualifier)): if (wordNextNext in timeQualifiersPM or wordNextNextNext in timeQualifiersPM): remainder = "pm" used += 1 if (wordNextNext in timeQualifiersAM or wordNextNextNext in timeQualifiersAM): remainder = "am" used += 1 if timeQualifier != "": used += 1 # TODO: Unsure if this is 100% accurate military = True else: isTime = False HH = int(strHH) if strHH else 0 MM = int(strMM) if strMM else 0 HH = HH + 12 if remainder == "pm" and HH < 12 else HH HH = HH - 12 if remainder == "am" and HH >= 12 else HH if (not military and remainder not in [ 'am', 'pm', 'hours', 'minutes', "second", "seconds", "hour", "minute" ] and ((not daySpecified) or dayOffset < 1)): # ambiguous time, detect whether they mean this evening or # the next morning based on whether it has already passed if dateNow.hour < HH: pass # No modification needed elif dateNow.hour < HH + 12: HH += 12 else: # has passed, assume the next morning dayOffset += 1 if timeQualifier in timeQualifiersPM and HH < 12: HH += 12 if HH > 24 or MM > 59: isTime = False used = 0 if isTime: hrAbs = HH minAbs = MM used += 1 if used > 0: # removed parsed words from the sentence for i in range(used): if idx + i >= len(words): break words[idx + i] = "" if wordPrev == "o" or wordPrev == "oh": words[words.index(wordPrev)] = "" if wordPrev == "early": hrOffset = -1 words[idx - 1] = "" idx -= 1 elif wordPrev == "late": hrOffset = 1 words[idx - 1] = "" idx -= 1 if idx > 0 and wordPrev in markers: words[idx - 1] = "" if wordPrev == "this": daySpecified = True if idx > 1 and wordPrevPrev in markers: words[idx - 2] = "" if wordPrevPrev == "this": daySpecified = True idx += used - 1 found = True # check that we found a date if not date_found: return None if dayOffset is False: dayOffset = 0 # perform date manipulation extractedDate = dateNow.replace(microsecond=0) if datestr != "": # date included an explicit date, e.g. "june 5" or "june 2, 2017" try: temp = datetime.strptime(datestr, "%B %d") except ValueError: # Try again, allowing the year temp = datetime.strptime(datestr, "%B %d %Y") extractedDate = extractedDate.replace(hour=0, minute=0, second=0) if not hasYear: temp = temp.replace(year=extractedDate.year, tzinfo=extractedDate.tzinfo) if extractedDate < temp: extractedDate = extractedDate.replace( year=int(currentYear), month=int(temp.strftime("%m")), day=int(temp.strftime("%d")), tzinfo=extractedDate.tzinfo) else: extractedDate = extractedDate.replace( year=int(currentYear) + 1, month=int(temp.strftime("%m")), day=int(temp.strftime("%d")), tzinfo=extractedDate.tzinfo) else: extractedDate = extractedDate.replace( year=int(temp.strftime("%Y")), month=int(temp.strftime("%m")), day=int(temp.strftime("%d")), tzinfo=extractedDate.tzinfo) else: # ignore the current HH:MM:SS if relative using days or greater if hrOffset == 0 and minOffset == 0 and secOffset == 0: extractedDate = extractedDate.replace(hour=0, minute=0, second=0) if yearOffset != 0: extractedDate = extractedDate + relativedelta(years=yearOffset) if monthOffset != 0: extractedDate = extractedDate + relativedelta(months=monthOffset) if dayOffset != 0: extractedDate = extractedDate + relativedelta(days=dayOffset) if hrAbs != -1 and minAbs != -1: # If no time was supplied in the string set the time to default # time if it's available if hrAbs is None and minAbs is None and default_time is not None: hrAbs, minAbs = default_time.hour, default_time.minute else: hrAbs = hrAbs or 0 minAbs = minAbs or 0 extractedDate = extractedDate + relativedelta(hours=hrAbs, minutes=minAbs) if (hrAbs != 0 or minAbs != 0) and datestr == "": if not daySpecified and dateNow > extractedDate: extractedDate = extractedDate + relativedelta(days=1) if hrOffset != 0: extractedDate = extractedDate + relativedelta(hours=hrOffset) if minOffset != 0: extractedDate = extractedDate + relativedelta(minutes=minOffset) if secOffset != 0: extractedDate = extractedDate + relativedelta(seconds=secOffset) for idx, word in enumerate(words): if words[idx] == "and" and \ words[idx - 1] == "" and words[idx + 1] == "": words[idx] = "" resultStr = " ".join(words) resultStr = ' '.join(resultStr.split()) return [extractedDate, resultStr]
def extractnumber_pt(text): """ This function prepares the given text for parsing by making numbers consistent, getting rid of contractions, etc. Args: text (str): the string to normalize Returns: (int) or (float): The value of extracted number """ aWords = text.split() count = 0 result = None while count < len(aWords): val = 0 word = aWords[count] next_next_word = None if count + 1 < len(aWords): next_word = aWords[count + 1] if count + 2 < len(aWords): next_next_word = aWords[count + 2] else: next_word = None # is current word a number? if word in pt_numbers: val = pt_numbers[word] elif word.isdigit(): # doesn't work with decimals val = int(word) elif is_numeric(word): val = float(word) elif isFractional_pt(word): if not result: result = 1 result = result * isFractional_pt(word) count += 1 continue if not val: # look for fractions like "2/3" aPieces = word.split('/') # if (len(aPieces) == 2 and is_numeric(aPieces[0]) # and is_numeric(aPieces[1])): if look_for_fractions(aPieces): val = float(aPieces[0]) / float(aPieces[1]) if val: if result is None: result = 0 # handle fractions if next_word != "avos": result += val else: result = float(result) / float(val) if next_word is None: break # number word and fraction ands = ["e"] if next_word in ands: zeros = 0 if result is None: count += 1 continue newWords = aWords[count + 2:] newText = "" for word in newWords: newText += word + " " afterAndVal = extractnumber_pt(newText[:-1]) if afterAndVal: if result < afterAndVal or result < 20: while afterAndVal > 1: afterAndVal = afterAndVal / 10.0 for word in newWords: if word == "zero" or word == "0": zeros += 1 else: break for _ in range(0, zeros): afterAndVal = afterAndVal / 10.0 result += afterAndVal break elif next_next_word is not None: if next_next_word in ands: newWords = aWords[count + 3:] newText = "" for word in newWords: newText += word + " " afterAndVal = extractnumber_pt(newText[:-1]) if afterAndVal: if result is None: result = 0 result += afterAndVal break decimals = ["ponto", "virgula", u"v�rgula", ".", ","] if next_word in decimals: zeros = 0 newWords = aWords[count + 2:] newText = "" for word in newWords: newText += word + " " for word in newWords: if word == "zero" or word == "0": zeros += 1 else: break afterDotVal = str(extractnumber_pt(newText[:-1])) afterDotVal = zeros * "0" + afterDotVal result = float(str(result) + "." + afterDotVal) break count += 1 if result is None: return False # Return the $str with the number related words removed # (now empty strings, so strlen == 0) # aWords = [word for word in aWords if len(word) > 0] # text = ' '.join(aWords) if "." in str(result): integer, dec = str(result).split(".") # cast float to int if dec == "0": result = int(integer) return result
def extractnumber_en(text): """ This function prepares the given text for parsing by making numbers consistent, getting rid of contractions, etc. Args: text (str): the string to normalize Returns: (int) or (float): The value of extracted number """ aWords = text.split() aWords = [word for word in aWords if word not in ["the", "a", "an"]] and_pass = False valPreAnd = False val = False count = 0 while count < len(aWords): word = aWords[count] if is_numeric(word): # if word.isdigit(): # doesn't work with decimals val = float(word) elif word == "first": val = 1 elif word == "second": val = 2 elif isFractional_en(word): val = isFractional_en(word) else: if word == "one": val = 1 elif word == "two": val = 2 elif word == "three": val = 3 elif word == "four": val = 4 elif word == "five": val = 5 elif word == "six": val = 6 elif word == "seven": val = 7 elif word == "eight": val = 8 elif word == "nine": val = 9 elif word == "ten": val = 10 if val: if count < (len(aWords) - 1): wordNext = aWords[count + 1] else: wordNext = "" valNext = isFractional_en(wordNext) if valNext: val = val * valNext aWords[count + 1] = "" # if val == False: if not val: # look for fractions like "2/3" aPieces = word.split('/') # if (len(aPieces) == 2 and is_numeric(aPieces[0]) # and is_numeric(aPieces[1])): if look_for_fractions(aPieces): val = float(aPieces[0]) / float(aPieces[1]) elif and_pass: # added to value, quit here val = valPreAnd break else: count += 1 continue aWords[count] = "" if and_pass: aWords[count - 1] = '' # remove "and" val += valPreAnd elif count + 1 < len(aWords) and aWords[count + 1] == 'and': and_pass = True valPreAnd = val val = False count += 2 continue elif count + 2 < len(aWords) and aWords[count + 2] == 'and': and_pass = True valPreAnd = val val = False count += 3 continue break # if val == False: if not val: return False # Return the string with the number related words removed # (now empty strings, so strlen == 0) aWords = [word for word in aWords if len(word) > 0] text = ' '.join(aWords) return val
def extractnumber_it(text, short_scale=False, ordinals=False): """ This function extracts a number from a text string, handles pronunciations in long scale and short scale https://en.wikipedia.org/wiki/Names_of_large_numbers Args: text (str): the string to normalize short_scale (bool): use short scale if True, long scale if False ordinals (bool): consider ordinal numbers, third=3 instead of 1/3 Returns: (int) or (float) or False: The extracted number or False if no number was found """ string_num_ordinal_it = {} # first, second... if ordinals: if short_scale: for num in SHORT_ORDINAL_STRING_IT: num_string = SHORT_ORDINAL_STRING_IT[num] string_num_ordinal_it[num_string] = num STRING_NUM_ITA[num_string] = num else: for num in LONG_ORDINAL_STRING_IT: num_string = LONG_ORDINAL_STRING_IT[num] string_num_ordinal_it[num_string] = num STRING_NUM_ITA[num_string] = num # negate next number (-2 = 0 - 2) negatives = ['meno'] # 'negativo' non è usuale in italiano # multiply the previous number (one hundred = 1 * 100) multiplies = [ 'decina', 'decine', 'dozzina', 'dozzine', 'centinaia', 'centinaio', 'migliaia', 'migliaio', 'mila' ] # split sentence parse separately and sum ( 2 and a half = 2 + 0.5 ) fraction_marker = [' e '] # decimal marker ( 1 point 5 = 1 + 0.5) decimal_marker = [' punto ', ' virgola '] if short_scale: for num in SHORT_SCALE_IT: num_string = SHORT_SCALE_IT[num] STRING_NUM_ITA[num_string] = num multiplies.append(num_string) else: for num in LONG_SCALE_IT: num_string = LONG_SCALE_IT[num] STRING_NUM_ITA[num_string] = num multiplies.append(num_string) # 2 e 3/4 ed altri casi for separator in fraction_marker: components = text.split(separator) zeros = 0 if len(components) == 2: # count zeros in fraction part sub_components = components[1].split(' ') for element in sub_components: if element == 'zero' or element == '0': zeros += 1 else: break # ensure first is not a fraction and second is a fraction num1 = extractnumber_it(components[0]) num2 = extractnumber_it(components[1]) if num1 is not None and num2 is not None \ and num1 >= 1 and 0 < num2 < 1: return num1 + num2 # sette e quaranta sette e zero zero due elif num1 is not None and num2 is not None \ and num1 >= 1 and num2 > 1: return num1 + num2 / pow(10, len(str(num2)) + zeros) # 2 punto 5 for separator in decimal_marker: zeros = 0 # count zeros in fraction part components = text.split(separator) if len(components) == 2: sub_components = components[1].split(' ') for element in sub_components: if element == 'zero' or element == '0': zeros += 1 else: break number = int(extractnumber_it(components[0])) decimal = int(extractnumber_it(components[1])) if number is not None and decimal is not None: if '.' not in str(decimal): return number + decimal / pow(10, len(str(decimal)) + zeros) all_words = text.split() val = False prev_val = None to_sum = [] for idx, word in enumerate(all_words): if not word: continue prev_word = all_words[idx - 1] if idx > 0 else '' next_word = all_words[idx + 1] if idx + 1 < len(all_words) else '' # is this word already a number ? if is_numeric(word): val = float(word) # is this word the name of a number ? if word in STRING_NUM_ITA: val = STRING_NUM_ITA[word] # tre quarti un quarto trenta secondi if isFractional_it(word) and prev_val: if word[:-1] == 'second' and not ordinals: val = prev_val * 2 else: val = prev_val # is the prev word a number and should we multiply it? # twenty hundred, six hundred if word in multiplies: if not prev_val: prev_val = 1 val = prev_val * val # is this a spoken fraction? # mezza tazza if val is False: val = isFractional_it(word, short_scale=short_scale) # 2 quinti if not ordinals: next_value = isFractional_it(next_word, short_scale=short_scale) if next_value: if not val: val = 1 val = val * next_value # is this a negative number? if val and prev_word and prev_word in negatives: val = 0 - val if not val: val = extractnumber_long_it(word) # let's make sure it isn't a fraction if not val: # look for fractions like '2/3' all_pieces = word.split('/') if look_for_fractions(all_pieces): val = float(all_pieces[0]) / float(all_pieces[1]) else: prev_val = val # handle long numbers # six hundred sixty six # two million five hundred thousand if word in multiplies and next_word not in multiplies: to_sum.append(val) val = 0 prev_val = 0 elif extractnumber_long_it(word) > 100 and \ extractnumber_long_it(next_word) and \ next_word not in multiplies: to_sum.append(val) val = 0 prev_val = 0 if val is not None: for addend in to_sum: val = val + addend return val
def extractnumber_it(text, short_scale=False, ordinals=False): """ This function extracts a number from a text string, handles pronunciations in long scale and short scale https://en.wikipedia.org/wiki/Names_of_large_numbers Args: text (str): the string to normalize short_scale (bool): use short scale if True, long scale if False ordinals (bool): consider ordinal numbers, third=3 instead of 1/3 Returns: (int) or (float) or False: The extracted number or False if no number was found """ string_num_ordinal_it = {} # first, second... if ordinals: if short_scale: for num in SHORT_ORDINAL_STRING_IT: num_string = SHORT_ORDINAL_STRING_IT[num] string_num_ordinal_it[num_string] = num STRING_NUM_ITA[num_string] = num else: for num in LONG_ORDINAL_STRING_IT: num_string = LONG_ORDINAL_STRING_IT[num] string_num_ordinal_it[num_string] = num STRING_NUM_ITA[num_string] = num # negate next number (-2 = 0 - 2) negatives = ['meno'] # 'negativo' non è usuale in italiano # multiply the previous number (one hundred = 1 * 100) multiplies = ['decina', 'decine', 'dozzina', 'dozzine', 'centinaia', 'centinaio', 'migliaia', 'migliaio', 'mila'] # split sentence parse separately and sum ( 2 and a half = 2 + 0.5 ) fraction_marker = [' e '] # decimal marker ( 1 point 5 = 1 + 0.5) decimal_marker = [' punto ', ' virgola '] if short_scale: for num in SHORT_SCALE_IT: num_string = SHORT_SCALE_IT[num] STRING_NUM_ITA[num_string] = num multiplies.append(num_string) else: for num in LONG_SCALE_IT: num_string = LONG_SCALE_IT[num] STRING_NUM_ITA[num_string] = num multiplies.append(num_string) # 2 e 3/4 ed altri casi for separator in fraction_marker: components = text.split(separator) zeros = 0 if len(components) == 2: # count zeros in fraction part sub_components = components[1].split(' ') for element in sub_components: if element == 'zero' or element == '0': zeros += 1 else: break # ensure first is not a fraction and second is a fraction num1 = extractnumber_it(components[0]) num2 = extractnumber_it(components[1]) if num1 is not None and num2 is not None \ and num1 >= 1 and 0 < num2 < 1: return num1 + num2 # sette e quaranta sette e zero zero due elif num1 is not None and num2 is not None \ and num1 >= 1 and num2 > 1: return num1 + num2 / pow(10, len(str(num2)) + zeros) # 2 punto 5 for separator in decimal_marker: zeros = 0 # count zeros in fraction part components = text.split(separator) if len(components) == 2: sub_components = components[1].split(' ') for element in sub_components: if element == 'zero' or element == '0': zeros += 1 else: break number = int(extractnumber_it(components[0])) decimal = int(extractnumber_it(components[1])) if number is not None and decimal is not None: if '.' not in str(decimal): return number + decimal / pow(10, len(str(decimal)) + zeros) all_words = text.split() val = False prev_val = None to_sum = [] for idx, word in enumerate(all_words): if not word: continue prev_word = all_words[idx - 1] if idx > 0 else '' next_word = all_words[idx + 1] if idx + 1 < len(all_words) else '' # is this word already a number ? if is_numeric(word): val = float(word) # is this word the name of a number ? if word in STRING_NUM_ITA: val = STRING_NUM_ITA[word] # tre quarti un quarto trenta secondi if isFractional_it(word) and prev_val: if word[:-1] == 'second' and not ordinals: val = prev_val * 2 else: val = prev_val # is the prev word a number and should we multiply it? # twenty hundred, six hundred if word in multiplies: if not prev_val: prev_val = 1 val = prev_val * val # is this a spoken fraction? # mezza tazza if val is False: val = isFractional_it(word, short_scale=short_scale) # 2 quinti if not ordinals: next_value = isFractional_it(next_word, short_scale=short_scale) if next_value: if not val: val = 1 val = val * next_value # is this a negative number? if val and prev_word and prev_word in negatives: val = 0 - val if not val: val = extractnumber_long_it(word) # let's make sure it isn't a fraction if not val: # look for fractions like '2/3' all_pieces = word.split('/') if look_for_fractions(all_pieces): val = float(all_pieces[0]) / float(all_pieces[1]) else: prev_val = val # handle long numbers # six hundred sixty six # two million five hundred thousand if word in multiplies and next_word not in multiplies: to_sum.append(val) val = 0 prev_val = 0 elif extractnumber_long_it(word) > 100 and \ extractnumber_long_it(next_word) and \ next_word not in multiplies: to_sum.append(val) val = 0 prev_val = 0 if val is not None: for addend in to_sum: val = val + addend return val
def extractnumber_sv(text): """ This function prepares the given text for parsing by making numbers consistent, getting rid of contractions, etc. Args: text (str): the string to normalize Returns: (int) or (float): The value of extracted number """ aWords = text.split() and_pass = False valPreAnd = False val = False count = 0 while count < len(aWords): word = aWords[count] if is_numeric(word): val = float(word) elif word == "första": val = 1 elif word == "andra": val = 2 elif word == "tredje": val = 3 elif word == "fjärde": val = 4 elif word == "femte": val = 5 elif word == "sjätte": val = 6 elif is_fractional_sv(word): val = is_fractional_sv(word) else: if word == "en": val = 1 if word == "ett": val = 1 elif word == "två": val = 2 elif word == "tre": val = 3 elif word == "fyra": val = 4 elif word == "fem": val = 5 elif word == "sex": val = 6 elif word == "sju": val = 7 elif word == "åtta": val = 8 elif word == "nio": val = 9 elif word == "tio": val = 10 if val: if count < (len(aWords) - 1): wordNext = aWords[count + 1] else: wordNext = "" valNext = is_fractional_sv(wordNext) if valNext: val = val * valNext aWords[count + 1] = "" if not val: # look for fractions like "2/3" aPieces = word.split('/') if look_for_fractions(aPieces): val = float(aPieces[0]) / float(aPieces[1]) elif and_pass: # added to value, quit here val = valPreAnd break else: count += 1 continue aWords[count] = "" if and_pass: aWords[count - 1] = '' # remove "och" val += valPreAnd elif count + 1 < len(aWords) and aWords[count + 1] == 'och': and_pass = True valPreAnd = val val = False count += 2 continue elif count + 2 < len(aWords) and aWords[count + 2] == 'och': and_pass = True valPreAnd = val val = False count += 3 continue break if not val: return False return val
def extractnumber_fr(text): """Takes in a string and extracts a number. Args: text (str): the string to extract a number from Returns: (str): The number extracted or the original text. """ # normalize text, keep articles for ordinals versus fractionals text = normalize_fr(text, False) # split words by whitespace aWords = text.split() count = 0 result = None add = False while count < len(aWords): val = None word = aWords[count] wordNext = "" wordPrev = "" if count < (len(aWords) - 1): wordNext = aWords[count + 1] if count > 0: wordPrev = aWords[count - 1] if word in articles_fr: count += 1 continue if word in ["et", "plus", "+"]: count += 1 add = True continue # is current word a numeric number? if word.isdigit(): val = int(word) count += 1 elif is_numeric(word): val = float(word) count += 1 elif wordPrev in articles_fr and getOrdinal_fr(word): val = getOrdinal_fr(word) count += 1 # is current word the denominator of a fraction? elif isFractional_fr(word): val = isFractional_fr(word) count += 1 # is current word the numerator of a fraction? if val and wordNext: valNext = isFractional_fr(wordNext) if valNext: val = float(val) * valNext count += 1 if not val: count += 1 # is current word a numeric fraction like "2/3"? aPieces = word.split('/') # if (len(aPieces) == 2 and is_numeric(aPieces[0]) # and is_numeric(aPieces[1])): if look_for_fractions(aPieces): val = float(aPieces[0]) / float(aPieces[1]) # is current word followed by a decimal value? if wordNext == "virgule": zeros = 0 newWords = aWords[count + 1:] # count the number of zeros after the decimal sign for word in newWords: if word == "zéro" or word == "0": zeros += 1 else: break afterDotVal = None # extract the number after the zeros if newWords[zeros].isdigit(): afterDotVal = newWords[zeros] countDot = count + zeros + 2 # if a number was extracted (since comma is also a # punctuation sign) if afterDotVal: count = countDot if not val: val = 0 # add the zeros afterDotString = zeros * "0" + afterDotVal val = float(str(val) + "." + afterDotString) if val: if add: result += val add = False else: result = val # if result == False: if not result: return normalize_fr(text, True) return result
def extractnumber_en(text, short_scale=True, ordinals=False): """ This function extracts a number from a text string, handles pronunciations in long scale and short scale https://en.wikipedia.org/wiki/Names_of_large_numbers Args: text (str): the string to normalize short_scale (bool): use short scale if True, long scale if False ordinals (bool): consider ordinal numbers, third=3 instead of 1/3 Returns: (int) or (float) or False: The extracted number or False if no number was found """ string_num_en = { "half": 0.5, "halves": 0.5, "couple": 2, "hundred": 100, "hundreds": 100, "thousand": 1000, "thousands": 1000, "million": 1000000, 'millions': 1000000} string_num_ordinal_en = {} for num in NUM_STRING_EN: num_string = NUM_STRING_EN[num] string_num_en[num_string] = num # first, second... if ordinals: if short_scale: for num in SHORT_ORDINAL_STRING_EN: num_string = SHORT_ORDINAL_STRING_EN[num] string_num_ordinal_en[num_string] = num string_num_en[num_string] = num else: for num in LONG_ORDINAL_STRING_EN: num_string = LONG_ORDINAL_STRING_EN[num] string_num_ordinal_en[num_string] = num string_num_en[num_string] = num # negate next number (-2 = 0 - 2) negatives = ["negative", "minus"] # sum the next number (twenty two = 20 + 2) sums = ['twenty', 'thirty', 'forty', 'fifty', 'sixty', 'seventy', 'eighty', 'ninety'] # multiply the previous number (one hundred = 1 * 100) multiplies = ["hundred", "thousand", "hundreds", "thousands", "million", "millions"] # split sentence parse separately and sum ( 2 and a half = 2 + 0.5 ) fraction_marker = [" and "] # decimal marker ( 1 point 5 = 1 + 0.5) decimal_marker = [" point ", " dot "] if short_scale: for num in SHORT_SCALE_EN: num_string = SHORT_SCALE_EN[num] string_num_en[num_string] = num string_num_en[num_string + "s"] = num multiplies.append(num_string) multiplies.append(num_string + "s") else: for num in LONG_SCALE_EN: num_string = LONG_SCALE_EN[num] string_num_en[num_string] = num string_num_en[num_string + "s"] = num multiplies.append(num_string) multiplies.append(num_string + "s") # 2 and 3/4 for c in fraction_marker: components = text.split(c) if len(components) == 2: # ensure first is not a fraction and second is a fraction num1 = extractnumber_en(components[0]) num2 = extractnumber_en(components[1]) if num1 is not None and num2 is not None \ and num1 >= 1 and 0 < num2 < 1: return num1 + num2 # 2 point 5 for c in decimal_marker: components = text.split(c) if len(components) == 2: number = extractnumber_en(components[0]) decimal = extractnumber_en(components[1]) if number is not None and decimal is not None: # TODO handle number dot number number number if "." not in str(decimal): return number + float("0." + str(decimal)) aWords = text.split() aWords = [word for word in aWords if word not in ["the", "a", "an"]] val = False prev_val = None to_sum = [] for idx, word in enumerate(aWords): if not word: continue prev_word = aWords[idx - 1] if idx > 0 else "" next_word = aWords[idx + 1] if idx + 1 < len(aWords) else "" # is this word already a number ? if is_numeric(word): # if word.isdigit(): # doesn't work with decimals val = float(word) # is this word the name of a number ? if word in string_num_en: val = string_num_en[word] # is the prev word an ordinal number and current word is one? # second one, third one if ordinals and prev_word in string_num_ordinal_en and val is 1: val = prev_val # is the prev word a number and should we sum it? # twenty two, fifty six if prev_word in sums and word in string_num_en: if val and val < 10: val = prev_val + val # is the prev word a number and should we multiply it? # twenty hundred, six hundred if word in multiplies: if not prev_val: prev_val = 1 val = prev_val * val # is this a spoken fraction? # half cup if val is False: val = isFractional_en(word, short_scale=short_scale) # 2 fifths if not ordinals: next_value = isFractional_en(next_word, short_scale=short_scale) if next_value: if not val: val = 1 val = val * next_value # is this a negative number? if val and prev_word and prev_word in negatives: val = 0 - val # let's make sure it isn't a fraction if not val: # look for fractions like "2/3" aPieces = word.split('/') if look_for_fractions(aPieces): val = float(aPieces[0]) / float(aPieces[1]) else: prev_val = val # handle long numbers # six hundred sixty six # two million five hundred thousand if word in multiplies and next_word not in multiplies: to_sum.append(val) val = 0 prev_val = 0 if val is not None: for v in to_sum: val = val + v return val
def extractnumber_pt(text): """ This function prepares the given text for parsing by making numbers consistent, getting rid of contractions, etc. Args: text (str): the string to normalize Returns: (int) or (float): The value of extracted number """ aWords = text.split() count = 0 result = None while count < len(aWords): val = 0 word = aWords[count] next_next_word = None if count + 1 < len(aWords): next_word = aWords[count + 1] if count + 2 < len(aWords): next_next_word = aWords[count + 2] else: next_word = None # is current word a number? if word in pt_numbers: val = pt_numbers[word] elif word.isdigit(): # doesn't work with decimals val = int(word) elif is_numeric(word): val = float(word) elif isFractional_pt(word): if not result: result = 1 result = result * isFractional_pt(word) count += 1 continue if not val: # look for fractions like "2/3" aPieces = word.split('/') # if (len(aPieces) == 2 and is_numeric(aPieces[0]) # and is_numeric(aPieces[1])): if look_for_fractions(aPieces): val = float(aPieces[0]) / float(aPieces[1]) if val: if result is None: result = 0 # handle fractions if next_word != "avos": result += val else: result = float(result) / float(val) if next_word is None: break # number word and fraction ands = ["e"] if next_word in ands: zeros = 0 if result is None: count += 1 continue newWords = aWords[count + 2:] newText = "" for word in newWords: newText += word + " " afterAndVal = extractnumber_pt(newText[:-1]) if afterAndVal: if result < afterAndVal or result < 20: while afterAndVal > 1: afterAndVal = afterAndVal / 10.0 for word in newWords: if word == "zero" or word == "0": zeros += 1 else: break for _ in range(0, zeros): afterAndVal = afterAndVal / 10.0 result += afterAndVal break elif next_next_word is not None: if next_next_word in ands: newWords = aWords[count + 3:] newText = "" for word in newWords: newText += word + " " afterAndVal = extractnumber_pt(newText[:-1]) if afterAndVal: if result is None: result = 0 result += afterAndVal break decimals = ["ponto", "virgula", u"v�rgula", ".", ","] if next_word in decimals: zeros = 0 newWords = aWords[count + 2:] newText = "" for word in newWords: newText += word + " " for word in newWords: if word == "zero" or word == "0": zeros += 1 else: break afterDotVal = str(extractnumber_pt(newText[:-1])) afterDotVal = zeros * "0" + afterDotVal result = float(str(result) + "." + afterDotVal) break count += 1 if result is None: return False # Return the $str with the number related words removed # (now empty strings, so strlen == 0) # aWords = [word for word in aWords if len(word) > 0] # text = ' '.join(aWords) if "." in str(result): integer, dec = str(result).split(".") # cast float to int if dec == "0": result = int(integer) return result
def extractnumber_en(text, short_scale=True, ordinals=False): """ This function extracts a number from a text string, handles pronunciations in long scale and short scale https://en.wikipedia.org/wiki/Names_of_large_numbers Args: text (str): the string to normalize short_scale (bool): use short scale if True, long scale if False ordinals (bool): consider ordinal numbers, third=3 instead of 1/3 Returns: (int) or (float) or False: The extracted number or False if no number was found """ string_num_en = { "half": 0.5, "halves": 0.5, "hundred": 100, "hundreds": 100, "thousand": 1000, "thousands": 1000, "million": 1000000, 'millions': 1000000 } for num in NUM_STRING_EN: num_string = NUM_STRING_EN[num] string_num_en[num_string] = num # first, second... if ordinals: if short_scale: for num in SHORT_ORDINAL_STRING_EN: num_string = SHORT_ORDINAL_STRING_EN[num] string_num_en[num_string] = num else: for num in LONG_ORDINAL_STRING_EN: num_string = LONG_ORDINAL_STRING_EN[num] string_num_en[num_string] = num # negate next number (-2 = 0 - 2) negatives = ["negative", "minus"] # sum the next number (twenty two = 20 + 2) sums = [ 'twenty', 'thirty', 'forty', 'fifty', 'sixty', 'seventy', 'eighty', 'ninety' ] # multiply the previous number (one hundred = 1 * 100) multiplies = [ "hundred", "thousand", "hundreds", "thousands", "million", "millions" ] # split sentence parse separately and sum ( 2 and a half = 2 + 0.5 ) fraction_marker = [" and "] # decimal marker ( 1 point 5 = 1 + 0.5) decimal_marker = [" point ", " dot "] if short_scale: for num in SHORT_SCALE_EN: num_string = SHORT_SCALE_EN[num] string_num_en[num_string] = num string_num_en[num_string + "s"] = num multiplies.append(num_string) multiplies.append(num_string + "s") else: for num in LONG_SCALE_EN: num_string = LONG_SCALE_EN[num] string_num_en[num_string] = num string_num_en[num_string + "s"] = num multiplies.append(num_string) multiplies.append(num_string + "s") # 2 and 3/4 for c in fraction_marker: components = text.split(c) if len(components) == 2: # ensure first is not a fraction and second is a fraction num1 = extractnumber_en(components[0]) num2 = extractnumber_en(components[1]) if num1 is not None and num2 is not None \ and num1 >= 1 and 0 < num2 < 1: return num1 + num2 # 2 point 5 for c in decimal_marker: components = text.split(c) if len(components) == 2: number = extractnumber_en(components[0]) decimal = extractnumber_en(components[1]) if number is not None and decimal is not None: # TODO handle number dot number number number if "." not in str(decimal): return number + float("0." + str(decimal)) aWords = text.split() aWords = [word for word in aWords if word not in ["the", "a", "an"]] val = False prev_val = None to_sum = [] for idx, word in enumerate(aWords): if not word: continue prev_word = aWords[idx - 1] if idx > 0 else "" next_word = aWords[idx + 1] if idx + 1 < len(aWords) else "" # is this word already a number ? if is_numeric(word): # if word.isdigit(): # doesn't work with decimals val = float(word) # is this word the name of a number ? if word in string_num_en: val = string_num_en[word] # is the prev word a number and should we sum it? # twenty two, fifty six if prev_word in sums and word in string_num_en: if val and val < 10: val = prev_val + val # is the prev word a number and should we multiply it? # twenty hundred, six hundred if word in multiplies: if not prev_val: prev_val = 1 val = prev_val * val # is this a spoken fraction? # half cup if val is False: val = isFractional_en(word, short_scale=short_scale) # 2 fifths if not ordinals: next_value = isFractional_en(next_word, short_scale=short_scale) if next_value: if not val: val = 1 val = val * next_value # is this a negative number? if val and prev_word and prev_word in negatives: val = 0 - val # let's make sure it isn't a fraction if not val: # look for fractions like "2/3" aPieces = word.split('/') if look_for_fractions(aPieces): val = float(aPieces[0]) / float(aPieces[1]) else: prev_val = val # handle long numbers # six hundred sixty six # two million five hundred thousand if word in multiplies and next_word not in multiplies: to_sum.append(val) val = 0 prev_val = 0 if val is not None: for v in to_sum: val = val + v return val
def _extract_whole_number_with_text_en(tokens, short_scale, ordinals): """ Handle numbers not handled by the decimal or fraction functions. This is generally whole numbers. Note that phrases such as "one half" will be handled by this function, while "one and a half" are handled by the fraction function. Args: tokens [_Token]: short_scale boolean: ordinals boolean: Returns: int or float, [_Tokens] The value parsed, and tokens that it corresponds to. """ multiplies, string_num_ordinal, string_num_scale = \ _initialize_number_data(short_scale) number_words = [] # type: [_Token] val = False prev_val = None next_val = None to_sum = [] for idx, token in enumerate(tokens): current_val = None if next_val: next_val = None continue word = token.word if word in _ARTICLES or word in _NEGATIVES: number_words.append(token) continue prev_word = tokens[idx - 1].word if idx > 0 else "" next_word = tokens[idx + 1].word if idx + 1 < len(tokens) else "" if word not in string_num_scale and \ word not in _STRING_NUM_EN and \ word not in _SUMS and \ word not in multiplies and \ not (ordinals and word in string_num_ordinal) and \ not is_numeric(word) and \ not isFractional_en(word, short_scale=short_scale) and \ not look_for_fractions(word.split('/')): words_only = [token.word for token in number_words] if number_words and not all( [w in _ARTICLES | _NEGATIVES for w in words_only]): break else: number_words = [] continue elif word not in multiplies \ and prev_word not in multiplies \ and prev_word not in _SUMS \ and not (ordinals and prev_word in string_num_ordinal) \ and prev_word not in _NEGATIVES \ and prev_word not in _ARTICLES: number_words = [token] elif prev_word in _SUMS and word in _SUMS: number_words = [token] else: number_words.append(token) # is this word already a number ? if is_numeric(word): if word.isdigit(): # doesn't work with decimals val = int(word) else: val = float(word) current_val = val # is this word the name of a number ? if word in _STRING_NUM_EN: val = _STRING_NUM_EN.get(word) current_val = val elif word in string_num_scale: val = string_num_scale.get(word) current_val = val elif ordinals and word in string_num_ordinal: val = string_num_ordinal[word] current_val = val # is the prev word an ordinal number and current word is one? # second one, third one if ordinals and prev_word in string_num_ordinal and val is 1: val = prev_val # is the prev word a number and should we sum it? # twenty two, fifty six if prev_word in _SUMS and val and val < 10: val = prev_val + val # is the prev word a number and should we multiply it? # twenty hundred, six hundred if word in multiplies: if not prev_val: prev_val = 1 val = prev_val * val # is this a spoken fraction? # half cup if val is False: val = isFractional_en(word, short_scale=short_scale) current_val = val # 2 fifths if not ordinals: next_val = isFractional_en(next_word, short_scale=short_scale) if next_val: if not val: val = 1 val = val * next_val number_words.append(tokens[idx + 1]) # is this a negative number? if val and prev_word and prev_word in _NEGATIVES: val = 0 - val # let's make sure it isn't a fraction if not val: # look for fractions like "2/3" aPieces = word.split('/') if look_for_fractions(aPieces): val = float(aPieces[0]) / float(aPieces[1]) current_val = val else: if prev_word in _SUMS and word not in _SUMS and current_val >= 10: # Backtrack - we've got numbers we can't sum. number_words.pop() val = prev_val break prev_val = val # handle long numbers # six hundred sixty six # two million five hundred thousand if word in multiplies and next_word not in multiplies: to_sum.append(val) val = 0 prev_val = 0 if val is not None and to_sum: val += sum(to_sum) return val, number_words