def preprocess(self, corpus):
'''
This iterates through the corpus line by line tokenizing,
spellchecking, normalizing abbreviations, etc.
It's debatable how you want to tokenize, I chose to do it by sentence
but it could be done with regex like so "[\w' ]+" which would give you
smaller phrases.
I: list of text strings
O: preprocessed list of text strings
'''
sent_detector = nltk.data.load('tokenizers/punkt/english.pickle')
#lowercase and tokenize by sentence
corpus_formatted = [tokenized_line for line in corpus \
for tokenized_line in sent_detector.tokenize(line.strip().lower())]
#consider re-designing to a helper function that does multiple manipulations to a string
#expanding the abreviations
corpus_formatted_expanded = [multiple_replace(line) for line in corpus_formatted]
#spell check *note this function takes quite a bit of time
corpus_formatted_expanded_correct = [' '.join([spell_checker.correct(word) for word in line.strip().split()])
for line in corpus_formatted_expanded]
#*FIX HACK
#further refinement
corpus_formatted_expanded_correct = [add_question_mark_or_period_to_sentence(line) for line in corpus_formatted_expanded_correct]
return [custome_refine(line) for line in corpus_formatted_expanded_correct]
def preprocess(text, twit=False):
""" Preprocess text for sentiment analysis
"""
tagger = twit_pos if twit else postag
toks, pos = tagger(text)
tokens = [spell.correct(t) for t in toks]
pairs = map(list, zip(tokens,pos))
return reduce_form(pairs)
def spellcheck(raw):
""" Performs spell correction on the input
"""
l = []
sents = nltk.tokenize.sent_tokenize(raw)
for sent in sents:
words = nltk.word_tokenize(sent)
for word in words:
l.append(spell.correct(word))
return ' '.join(l)
def spell_check(l):
"""
Return a list of spell-checked, pos-tagged tweets in the format of [tokens, pos]
"""
print 'spellchecking '
ret = []
for tweet in l:
tokens = []
for t in tweet[0]:
tokens.append(s.correct(t))
ret.append([tokens, tweet[1]])
print 'done spellchecking'
return ret
def spell_check(l): """ Return a list of spell-checked, pos-tagged tweets in the format of [tokens, pos] """ print 'spellchecking ' ret = [] for tweet in l: tokens = [] for t in tweet[0]: tokens.append(s.correct(t)) ret.append([tokens, tweet[1]]) print 'done spellchecking' return ret
def preprocess(self, corpus):
'''
This iterates through the corpus line by line tokenizing,
spellchecking, normalizing abbreviations, etc.
It's debatable how you want to tokenize, I chose to do it by sentence
but it could be done with regex like so "[\w' ]+" which would give you
smaller phrases.
I: list of text strings
O: preprocessed list of text strings
'''
sent_detector = nltk.data.load('tokenizers/punkt/english.pickle')
#lowercase and tokenize by sentence
corpus_formatted = [tokenized_line for line in corpus \
for tokenized_line in sent_detector.tokenize(line.strip().lower())]
#consider re-designing to a helper function that does multiple manipulations to a string
#expanding the abreviations
corpus_formatted_expanded = [
multiple_replace(line) for line in corpus_formatted
]
#spell check *note this function takes quite a bit of time
corpus_formatted_expanded_correct = [
' '.join(
[spell_checker.correct(word) for word in line.strip().split()])
for line in corpus_formatted_expanded
]
#*FIX HACK
#further refinement
corpus_formatted_expanded_correct = [
add_question_mark_or_period_to_sentence(line)
for line in corpus_formatted_expanded_correct
]
return [
custome_refine(line) for line in corpus_formatted_expanded_correct
]
def predict(self, e):
s = spell_checker.correct(self.editname.GetValue())
self.result.SetLabel(s)
def predict(self, e):
s=spell_checker.correct(self.editname.GetValue())
self.result.SetLabel(s)
#!/usr/bin/python
import spell_checker
file = 'commonmisspelled.txt'
f = open(file)
error = []
correct = []
for line in f:
error.append(line.split()[0])
correct.append(line.split()[1])
meanReciprocalRank = 0
avgTime = 0.0
for i in range(len(error)):
suggestions = spell_checker.correct(error[i])
#avgTime += time
reciprocalRank = 0.0
for j in range(5):
if suggestions[j][0] == correct[i]:
reciprocalRank = 1.0/(j+1)
break
meanReciprocalRank += reciprocalRank
print i
#print float(avgTime)/len(error)
print float(meanReciprocalRank)/len(error)
def GET(self,name):
user_data = web.input()
return correct(user_data.text.replace('sapiAGRI ',''))
def test(input_file_name,output_file_name):
file = open(input_file_name)
data = file.read()
element = ET.XML(data)
outfile = open(output_file_name,"w")
# regex for single word.
regex_single = re.compile('\w+')
# regex for phrase.
regex_phrase = re.compile('\w+|[,|:|(|)|\'|\"]')
# creating the regex obj.
regex = re.compile('\w+|[.|,|:|(|)|\'|\"]')
for subelement in element:
if(subelement.attrib['id'][0] == '1'): # testcase1.xml
outfile.write(subelement.attrib['id'])
single_word = regex_single.findall(subelement.text)
suggestions = spell_checker.correct(single_word[0])
# printing the misspelled word.
outfile.write(", "+single_word[0])
i = 1
# printing the suggestions.
for each_suggestion in suggestions:
outfile.write(", ")
outfile.write(each_suggestion[0])
i += 1
if( i > MAX_SUGGESTIONS):
outfile.write("\n")
break
elif( subelement.attrib['id'][0] == '2'): # testcase2.xml
words_in_phrase = regex_phrase.findall( subelement.text)
# identifying the incorrect words in the phrase.
i = 0;
for i in xrange(0,len(words_in_phrase)):
# if it is a punctuations then we should not query the dictionary.
if( isPunctuation(words_in_phrase[i])):
continue
suggestions = spell_checker.correct(words_in_phrase[i])
if( len(suggestions) == 1): # correct word.
continue
outfile.write(subelement.attrib['id'])
incorrect_word = words_in_phrase[i]
outfile.write(", "+incorrect_word)
temp_list = []
for each_suggestion in suggestions[:MAX_UNIGRAM_SUGGESTIONS] :
words_in_phrase[i] = each_suggestion[0]
temp_list.append( ( each_suggestion[0], ngrams.getProbabilityOfPhrase(words_in_phrase,100,False)))
temp_list.sort( key = getSecond, reverse=True)
j = 1
# printing the suggestions.
for each_temp in temp_list:
outfile.write(", ")
outfile.write(each_temp[0])
j += 1
if( j > MAX_SUGGESTIONS):
outfile.write("\n")
break
# handling the testcases3
elif(subelement.attrib['id'][0] == '3'): # testcases3.xml
sentence = subelement.text
words_in_sentence = regex.findall(sentence) # TODO: dot is added here.
# identifying the incorrect words in the sentence.
i = 0;
for i in xrange(0,len(words_in_sentence)):
# if it is a punctuations then we should not query the dictionary.
if( isPunctuation(words_in_sentence[i])):
continue
suggestions = spell_checker.correct(words_in_sentence[i])
if( len(suggestions) == 1): # correct word.
continue
# patch.
for k in xrange(0,len(suggestions)):
suggestions[k] = suggestions[k][0]
outfile.write(subelement.attrib['id'])
incorrect_word = words_in_sentence[i]
outfile.write(", "+incorrect_word)
# creating a new list to call the correctSentences function.
aux_list = []
for t in xrange(0,(len(words_in_sentence)-1)):
aux_list.append([words_in_sentence[t]])
aux_list[i] = suggestions[:MAX_SUGGESTIONS]
print aux_list
sorted_suggestions = pos_ngrams.correctSentences(aux_list)
j = 1
# printing the suggestions.
for each_sorted_suggestion in sorted_suggestions:
outfile.write(", ")
outfile.write(each_sorted_suggestion)
j += 1
if( j > MAX_SUGGESTIONS):
outfile.write("\n")
break
# handling the testcases4
elif(subelement.attrib['id'][0] == '4'): # testcases4.xml
sentences = subelement.text.split('.') # dot is the sentence separator.
for each_sentence in sentences:
words_in_sentence = regex.findall(each_sentence) # a list is returned.
# identifying the incorrect words in the sentence.
i = 0;
for i in xrange(0,len(words_in_sentence)):
# if it is a punctuations then we should not query the dictionary.
if( isPunctuation(words_in_sentence[i])):
continue
suggestions = spell_checker.correct(words_in_sentence[i])
if( len(suggestions) == 1): # correct word.
continue
# patch.
for k in xrange(0,len(suggestions)):
suggestions[k] = suggestions[k][0]
outfile.write(subelement.attrib['id'])
incorrect_word = words_in_sentence[i]
outfile.write(", "+incorrect_word)
# creating a new list to call the correctSentences function.
aux_list = []
for t in xrange(0,len(words_in_sentence)):
aux_list.append([words_in_sentence[t]])
aux_list[i] = suggestions[:MAX_SUGGESTIONS]
print aux_list
sorted_suggestions = pos_ngrams.correctSentences(aux_list)
j = 1
# printing the suggestions.
for each_sorted_suggestion in sorted_suggestions:
outfile.write(", ")
outfile.write(each_sorted_suggestion)
j += 1
if( j > MAX_SUGGESTIONS):
outfile.write("\n")
break
file.close()
outfile.close()
def correct_spelling(line):
return ' '.join(
[spell_checker.correct(word) for word in line.strip().split()])
def cleanText(s):
s = unicode(s).lower()
if isinstance(s, unicode):
s = correct(s)
s = s.replace(" x "," xby ")
s = s.replace("*"," xby ")
s = s.replace(" by "," xby")
s = s.replace("x0"," xby 0")
s = s.replace("x1"," xby 1")
s = s.replace("x2"," xby 2")
s = s.replace("x3"," xby 3")
s = s.replace("x4"," xby 4")
s = s.replace("x5"," xby 5")
s = s.replace("x6"," xby 6")
s = s.replace("x7"," xby 7")
s = s.replace("x8"," xby 8")
s = s.replace("x9"," xby 9")
s = s.replace("0x","0 xby ")
s = s.replace("1x","1 xby ")
s = s.replace("2x","2 xby ")
s = s.replace("3x","3 xby ")
s = s.replace("4x","4 xby ")
s = s.replace("5x","5 xby ")
s = s.replace("6x","6 xby ")
s = s.replace("7x","7 xby ")
s = s.replace("8x","8 xby ")
s = s.replace("9x","9 xby ")
s = re.sub(r"([0-9]+)( *)(square|sq) ?\.?(feet|foot|ft)\.?", r"\1sq.ft. ", s)
s = re.sub(r"([0-9]+)( *)(gallons|gallon|gal)\.?", r"\1gal. ", s)
s = re.sub(r"([0-9]+)( *)(ounces|ounce|oz)\.?", r"\1oz. ", s)
s = re.sub(r"([0-9]+)( *)(centimeters|cm)\.?", r"\1cm. ", s)
s = re.sub(r"([0-9]+)( *)(milimeters|mm)\.?", r"\1mm. ", s)
s = re.sub(r"([0-9]+)( *)(degrees|degree)\.?", r"\1deg. ", s)
s = re.sub(r"([0-9]+)( *)(volts|volt)\.?", r"\1volt. ", s)
s = re.sub(r"([0-9]+)( *)(watts|watt)\.?", r"\1watt. ", s)
s = re.sub(r"([0-9]+)( *)(amperes|ampere|amps|amp)\.?", r"\1amp. ", s)
s = s.replace("whirpool","whirlpool")
s = s.replace("whirlpoolga", "whirlpool")
s = s.replace("whirlpoolstainless","whirlpool stainless")
s = s.replace(" "," ")
s = re.sub(r"(\w)\.([A-Z])", r"\1 \2", s) #Split words with a.A
s = s.lower()
s = s.replace(" "," ")
s = s.replace(",","") #could be number / segment later
s = s.replace("$"," ")
s = s.replace("?"," ")
s = s.replace("-"," ")
s = s.replace("//","/")
s = s.replace("..",".")
s = s.replace(" / "," ")
s = s.replace(" \\ "," ")
s = s.replace("."," . ")
s = re.sub(r"(^\.|/)", r"", s)
s = re.sub(r"(\.|/)$", r"", s)
s = re.sub(r"([0-9])([a-z])", r"\1 \2", s)
s = re.sub(r"([a-z])([0-9])", r"\1 \2", s)
s = s.replace(" x "," xbi ")
s = re.sub(r"([a-z])( *)\.( *)([a-z])", r"\1 \4", s)
s = re.sub(r"([a-z])( *)/( *)([a-z])", r"\1 \4", s)
s = s.replace("*"," xbi ")
s = s.replace(" by "," xbi ")
s = re.sub(r"([0-9])( *)\.( *)([0-9])", r"\1.\4", s)
s = re.sub(r"([0-9]+)( *)(inches|inch|in|')\.?", r"\1in. ", s)
s = re.sub(r"([0-9]+)( *)(foot|feet|ft|'')\.?", r"\1ft. ", s)
s = re.sub(r"([0-9]+)( *)(pounds|pound|lbs|lb)\.?", r"\1lb. ", s)
s = re.sub(r"([0-9]+)( *)(square|sq) ?\.?(feet|foot|ft)\.?", r"\1sq.ft. ", s)
s = re.sub(r"([0-9]+)( *)(cubic|cu) ?\.?(feet|foot|ft)\.?", r"\1cu.ft. ", s)
s = re.sub(r"([0-9]+)( *)(gallons|gallon|gal)\.?", r"\1gal. ", s)
s = re.sub(r"([0-9]+)( *)(ounces|ounce|oz)\.?", r"\1oz. ", s)
s = re.sub(r"([0-9]+)( *)(centimeters|cm)\.?", r"\1cm. ", s)
s = re.sub(r"([0-9]+)( *)(milimeters|mm)\.?", r"\1mm. ", s)
s = s.replace("°"," degrees ")
s = re.sub(r"([0-9]+)( *)(degrees|degree)\.?", r"\1deg. ", s)
s = s.replace(" v "," volts ")
s = re.sub(r"([0-9]+)( *)(volts|volt)\.?", r"\1volt. ", s)
s = re.sub(r"([0-9]+)( *)(watts|watt)\.?", r"\1watt. ", s)
s = re.sub(r"([0-9]+)( *)(amperes|ampere|amps|amp)\.?", r"\1amp. ", s)
s = s.replace(" "," ")
s = s.replace(" . "," ")
s = s.replace("toliet","toilet")
s = s.replace("airconditioner","air conditioner")
s = s.replace("vinal","vinyl")
s = s.replace("vynal","vinyl")
s = s.replace("skill","skil")
s = s.replace("snowbl","snow bl")
s = s.replace("plexigla","plexi gla")
s = s.replace("rustoleum","rust-oleum")
s = s.replace("whirpool","whirlpool")
s = s.replace("whirlpoolga", "whirlpool ga")
s = s.replace("whirlpoolstainless","whirlpool stainless")
s = (" ").join([str(strNum[z]) if z in strNum else z for z in s.split(" ")])
s = tokenizer.tokenize(s)
s = [word for word in s if word not in stopwords.words('english')]
return s
def correct_spelling(line):
return ' '.join([spell_checker.correct(word) for word in line.strip().split()])
def str_correcter(s):
return " ".join(
[spell_checker.correct(word) for word in s.lower().split()])
