def parse_all(user_input: str, culture: str) -> List[ModelResult]:
return [
# Number recognizer - This function will find any number from the input
# E.g "I have two apples" will return "2".
Recognizers.recognize_number(user_input, culture),
# Ordinal number recognizer - This function will find any ordinal number
# E.g "eleventh" will return "11".
Recognizers.recognize_ordinal(user_input, culture),
# Percentage recognizer - This function will find any number presented as percentage
# E.g "one hundred percents" will return "100%"
Recognizers.recognize_percentage(user_input, culture),
# Age recognizer - This function will find any age number presented
# E.g "After ninety five years of age, perspectives change" will return "95 Year"
Recognizers.recognize_age(user_input, culture),
# Currency recognizer - This function will find any currency presented
# E.g "Interest expense in the 1988 third quarter was $ 75.3 million" will return "75300000 Dollar"
Recognizers.recognize_currency(user_input, culture),
# Dimension recognizer - This function will find any dimension presented
# E.g "The six-mile trip to my airport hotel that had taken 20 minutes earlier in the day took more than three hours." will return "6 Mile"
Recognizers.recognize_dimension(user_input, culture),
# Temperature recognizer - This function will find any temperature presented
# E.g "Set the temperature to 30 degrees celsius" will return "30 C"
Recognizers.recognize_temperature(user_input, culture),
# DateTime recognizer - This function will find any Date even if its write in colloquial language -
# E.g "I'll go back 8pm today" will return "2017-10-04 20:00:00"
Recognizers.recognize_datetime(user_input, culture)
]
def _parse_all_entities(user_input: str,
culture: str) -> List[Dict[Text, Any]]:
"""
This is the main method that does the entity extraction work.
For more details: https://github.com/Microsoft/Recognizers-Text/tree/master/Python#api-documentation
"""
return [
# Number recognizer - This function will find any number from the input
# E.g "I have two apples" will return "2".
Recognizers.recognize_number(user_input, culture),
# Ordinal number recognizer - This function will find any ordinal number
# E.g "eleventh" will return "11".
Recognizers.recognize_ordinal(user_input, culture),
# Percentage recognizer - This function will find any number presented as percentage
# E.g "one hundred percents" will return "100%"
Recognizers.recognize_percentage(user_input, culture),
# Age recognizer - This function will find any age number presented
# E.g "After ninety five years of age, perspectives change" will return
# "95 Year"
Recognizers.recognize_age(user_input, culture),
# Currency recognizer - This function will find any currency presented
# E.g "Interest expense in the 1988 third quarter was $ 75.3 million"
# will return "75300000 Dollar"
Recognizers.recognize_currency(user_input, culture),
# Temperature recognizer - This function will find any temperature presented
# E.g "Set the temperature to 30 degrees celsius" will return "30 C"
Recognizers.recognize_temperature(user_input, culture),
# DateTime recognizer - This function will find any Date even if its write in colloquial language -
# E.g "I'll go back 8pm today" will return "2017-10-04 20:00:00"
Recognizers.recognize_datetime(user_input, culture),
# PhoneNumber recognizer will find any phone number presented
# E.g "My phone number is ( 19 ) 38294427."
Recognizers.recognize_phone_number(user_input, culture),
# Email recognizer will find any phone number presented
# E.g "Please write to me at [email protected] for more information on task
# #A1"
Recognizers.recognize_email(user_input, culture),
]
def time_delta(hour_minute, delta_time):
if not re.match('[0-2][0-9]:[0-5][0-9]', hour_minute):
result = Recognizers.recognize_datetime(hour_minute, Culture.Chinese)
if not result or not result[0].resolution['values'] or result[
0].resolution['values'][0]['type'] != 'time':
hour_minute = '12:00'
else:
hour_minute = result[0].resolution['values'][0]['value'][:5]
h_str_src, m_str_src = hour_minute.split(':')
h_int_src, m_int_src = int(h_str_src), int(m_str_src)
h_int_dst, m_int_dst = h_int_src + delta_time[0], m_int_src + delta_time[1]
if m_int_dst >= 60:
m_int_dst -= 60
h_int_dst += 1
elif m_int_dst < 0:
m_int_dst += 60
h_int_dst -= 1
if h_int_dst >= 24:
h_int_dst -= 24
elif h_int_dst < 0:
h_int_dst += 24
return "%02d:%02d" % (h_int_dst, m_int_dst)
def get_real_slot_value(slot_name, slot_value):
if slot_name == '评分':
f_rating = f_rating1 = f_rating2 = 4.0
b_about = False
digit_items = c2d.takeNumberFromString(slot_value)['digitsStringList']
if len(digit_items) == 0 or len(digit_items) > 2:
return '[{:.2f}, ∞)'.format(f_rating)
elif len(digit_items) == 1:
f_rating = float(digit_items[0])
else:
b_about = True
f_rating = float(digit_items[0]), float(digit_items[1])
f_rating1, f_rating2 = min(f_rating[0], f_rating[1]), max(
f_rating[0], f_rating[1])
if not b_about:
f_rating = revise_value(f_rating, slot_value, b_check_half=True)
if re.search('最少|至少|以上|超过|超出', slot_value):
return '[{:.2f}, ∞)'.format(f_rating)
if re.search('最多|最高|以下|不超过|以内|之内', slot_value):
return '[0.0, {:.2f}]'.format(f_rating)
if re.search('左右|上下|差不多|大概', slot_value):
return '[{:.2f}, {:.2f}]'.format(f_rating * 0.95,
f_rating * 1.05)
return '{:.2f}'.format(f_rating)
else:
obj = re.search('至|到|-|—', slot_value)
if obj:
raw_value1, raw_value2 = slot_value[:obj.start(
)], slot_value[obj.end():]
f_rating1 = revise_value(f_rating1,
raw_value1,
b_check_half=True)
f_rating2 = revise_value(f_rating2,
raw_value2,
b_check_half=True)
if abs(f_rating1 - f_rating2) <= 0.01:
return '{:.2f}'.format((f_rating1 + f_rating2) / 2)
else:
return '[{:.2f}, {:.2f}]'.format(f_rating1, f_rating2)
if slot_name == '价格':
f_price1 = f_price2 = 100
b_about = False
digit_items = c2d.takeNumberFromString(slot_value)['digitsStringList']
if len(digit_items) == 0 or len(digit_items) > 2:
return slot_value
elif len(digit_items) == 1:
f_price = abs(float(digit_items[0]))
else:
b_about = True
f_price = abs(float(digit_items[0])), abs(float(digit_items[1]))
f_price1, f_price2 = min(f_price[0],
f_price[1]), max(f_price[0], f_price[1])
if not b_about:
f_price = revise_value(f_price, slot_value)
if re.search('最少|至少|以上|超过|超出', slot_value):
return '[{:.2f}, ∞)'.format(f_price)
if re.search('最多|最高|以下|不超过|以内|之内', slot_value):
return '[0.0, {:.2f}]'.format(f_price)
if re.search('左右|上下|差不多|大概', slot_value):
return '[{:.2f}, {:.2f}]'.format(f_price * 0.8, f_price * 1.2)
return '{:.2f}'.format(f_price)
else:
obj = re.search('至|到|-|—', slot_value)
if obj:
raw_value1, raw_value2 = slot_value[:obj.start(
)], slot_value[obj.end():]
f_price1 = revise_value(f_price1, raw_value1)
f_price2 = revise_value(f_price2, raw_value2)
if abs(f_price1 - f_price2) <= 1.0:
return '{:.2f}'.format((f_price1 + f_price2) / 2)
else:
return '[{:.2f}, {:.2f}]'.format(f_price1, f_price2)
if slot_name == '时长':
f_hour1 = f_hour2 = 2
b_about = False
digit_items = c2d.takeNumberFromString(slot_value)['digitsStringList']
if len(digit_items) == 0 or len(digit_items) > 2:
if '半' in slot_value:
f_hour = 0.
else:
return slot_value
elif len(digit_items) == 1:
f_hour = float(digit_items[0])
elif re.search('[1-9]小时[1-9][0-9]分钟', slot_value):
f_hour = float(digit_items[0]) + float(digit_items[1]) / 60.0
else:
b_about = True
f_hour = float(digit_items[0]), float(digit_items[1])
f_hour1, f_hour2 = min(f_hour[0],
f_hour[1]), max(f_hour[0], f_hour[1])
if not b_about:
f_hour = revise_value(f_hour,
slot_value,
b_check_half=True,
b_time=True)
if re.search('最少|至少|以上|超过|超出', slot_value):
return '[{:.2f}, ∞)'.format(f_hour)
if re.search('最多|最高|以下|不超过|以内|之内', slot_value):
return '[0.0, {:.2f}]'.format(f_hour)
if re.search('左右|差不多|大概', slot_value):
return '[{:.2f}, {:.2f}]'.format(f_hour * 0.8, f_hour * 1.2)
return '{:.2f}'.format(f_hour)
else:
obj = re.search('至|到|-|—', slot_value)
if obj:
raw_value1, raw_value2 = slot_value[:obj.start(
)], slot_value[obj.end():]
f_hour1 = revise_value(f_hour1,
raw_value1,
b_check_half=True,
b_time=True)
f_hour2 = revise_value(f_hour2,
raw_value2,
b_check_half=True,
b_time=True)
if abs(f_hour1 - f_hour2) <= 0.1:
return '{:.2f}'.format((f_hour1 + f_hour2) / 2)
else:
return '[{:.2f}, {:.2f}]'.format(f_hour1, f_hour2)
if slot_name == '准点率':
obj = re.search('至|到|-|—', slot_value)
if obj and '至少' not in slot_value:
value1, value2 = slot_value[:obj.start()], slot_value[obj.end():]
f_punctuality1, f_punctuality2 = retrieve_punctuality(
value1), retrieve_punctuality(value1)
if f_punctuality1 < 0.1 or f_punctuality2 < 0.1:
return slot_value
if abs(f_punctuality1 - f_punctuality2) <= 0.01:
return '{:.2f}'.format((f_punctuality1 + f_punctuality2) / 2)
else:
return '[{:.2f}, {:.2f}]'.format(f_punctuality1,
f_punctuality2)
f_punctuality = retrieve_punctuality(slot_value)
if f_punctuality < 0.1:
return slot_value
if re.search('最低|至少|最少|以上|不低于', slot_value):
return '[{:.2f}, ∞)'.format(f_punctuality)
if re.search('大概|左右|上下|差不多', slot_value):
return '[{:.2f}, {:.2f}]'.format(f_punctuality * 0.8,
f_punctuality * 1.2)
return slot_value
if slot_name in ['出发时间', '到达时间']:
result = Recognizers.recognize_datetime(slot_value, Culture.Chinese)
if not result or not result[-1].resolution['values']:
return slot_value
str_type = result[-1].resolution['values'][0]['type']
if str_type == 'time':
time_str = result[-1].resolution['values'][0]['value'][:5]
if re.search('大概|左右|前后|差不多', slot_value):
time_left, time_right = time_delta(time_str,
[0, -10]), time_delta(
time_str, [0, 10])
return '[{}, {}]'.format(time_left, time_right)
if re.search('最早|后|以后|之后', slot_value):
return '[{}, ∞)'.format(time_str)
if re.search('最迟|最晚|前|以前|之前', slot_value):
return '(-∞, {}]'.format(time_str)
return time_str
elif str_type == 'timerange':
time_left = result[-1].resolution['values'][0]['start'][:5]
time_right = result[-1].resolution['values'][0]['end'][:5]
return '[{}, {}]'.format(time_left, time_right)
return slot_value
if slot_name in ['时间', '开始时间', '结束时间']:
result = Recognizers.recognize_datetime(slot_value, Culture.Chinese)
if not result or not result[-1].resolution['values'] or result[
-1].resolution['values'][0]['type'] != 'time':
return slot_value
return result[-1].resolution['values'][0]['value'][:5]
if slot_name in ['人数', '天数', '数量', '距离', 'choice']:
if slot_name == 'choice' and '两' in slot_value:
return '2'
digit_items = c2d.takeNumberFromString(slot_value)['digitsStringList']
if len(digit_items) == 0 or len(digit_items) > 2:
return slot_value
return str(digit_items[0]).split('.')[0]
return slot_value