Spaces:
Runtime error
Runtime error
Upload utils.py
Browse files
utils.py
ADDED
@@ -0,0 +1,149 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import inflect
|
2 |
+
from nltk.corpus import stopwords
|
3 |
+
from nltk.tag import pos_tag
|
4 |
+
from nltk.tokenize import word_tokenize
|
5 |
+
import re
|
6 |
+
import string
|
7 |
+
import wikipedia as wiki
|
8 |
+
|
9 |
+
|
10 |
+
# cleans last sentence of Quiz Bowl question
|
11 |
+
def clean_last_sent(text):
|
12 |
+
cleaned = text.replace('FTP,', 'For 10 points,')
|
13 |
+
cleaned = re.sub('(?i)for 10 points,', '', cleaned)
|
14 |
+
|
15 |
+
return cleaned
|
16 |
+
|
17 |
+
|
18 |
+
# gives the question a more "question-like" ending if the question hasn't reached the last sentence yet
|
19 |
+
def add_proper_tail(text):
|
20 |
+
if 'name this' in text.lower() or 'this' not in text.lower():
|
21 |
+
return text
|
22 |
+
|
23 |
+
if text[-1] == '.':
|
24 |
+
beginning = ' name this '
|
25 |
+
else:
|
26 |
+
beginning = '. name this '
|
27 |
+
words = text.split()
|
28 |
+
words = [word.lower() for word in words]
|
29 |
+
idx_of_this = words.index('this')
|
30 |
+
tail = beginning + words[idx_of_this + 1] + '.'
|
31 |
+
new_text = text + tail
|
32 |
+
|
33 |
+
return new_text
|
34 |
+
|
35 |
+
|
36 |
+
# returns the words in the text excluding stop words and punctuation
|
37 |
+
def get_filtered_words(text):
|
38 |
+
cleaned = clean_last_sent(text)
|
39 |
+
stop_words = set(stopwords.words('english'))
|
40 |
+
words = word_tokenize(cleaned)
|
41 |
+
filtered = [word for word in words if not word.lower() in stop_words]
|
42 |
+
filtered = [word for word in filtered if word in string.punctuation]
|
43 |
+
|
44 |
+
return filtered
|
45 |
+
|
46 |
+
|
47 |
+
# gets valid query using proper nouns in reverse order
|
48 |
+
def get_nnp_query(text):
|
49 |
+
# take out words in quotes
|
50 |
+
text = re.sub('"(.*?)"', '', text)
|
51 |
+
|
52 |
+
# find all proper nouns
|
53 |
+
tagged_sent = pos_tag(word_tokenize(text))
|
54 |
+
proper_nouns = [word for word, pos in tagged_sent if 'NNP' in pos]
|
55 |
+
proper_nouns.reverse()
|
56 |
+
|
57 |
+
query = ''
|
58 |
+
for nnp in proper_nouns:
|
59 |
+
test_query = query + nnp
|
60 |
+
results = wiki.search(test_query)
|
61 |
+
if len(results) == 0:
|
62 |
+
continue
|
63 |
+
query += nnp + ' '
|
64 |
+
|
65 |
+
return query
|
66 |
+
|
67 |
+
|
68 |
+
# gets valid query using nouns in reverse order
|
69 |
+
def get_nn_query(text):
|
70 |
+
# take out words in quotes
|
71 |
+
text = re.sub('"(.*?)"', '', text)
|
72 |
+
|
73 |
+
# find all types of nouns
|
74 |
+
tagged_sent = pos_tag(word_tokenize(text))
|
75 |
+
nouns = [word for word, pos in tagged_sent if 'NN' in pos]
|
76 |
+
nouns.reverse()
|
77 |
+
|
78 |
+
query = ''
|
79 |
+
for nn in nouns:
|
80 |
+
test_query = query + nn
|
81 |
+
results = wiki.search(test_query)
|
82 |
+
if len(results) == 0:
|
83 |
+
continue
|
84 |
+
query += nn + ' '
|
85 |
+
|
86 |
+
return query
|
87 |
+
|
88 |
+
|
89 |
+
# helper func to allow use of lambda in map
|
90 |
+
def lower(s):
|
91 |
+
return s.lower()
|
92 |
+
|
93 |
+
|
94 |
+
# checks if either of the texts are subsets of the other
|
95 |
+
def is_either_text_subset(text1, text2):
|
96 |
+
# tokenize words, lower() them, and get the unique words in a set
|
97 |
+
text1_set = set(map(lambda word: lower(word), word_tokenize(text1)))
|
98 |
+
text2_set = set(map(lambda word: lower(word), word_tokenize(text2)))
|
99 |
+
|
100 |
+
if text1_set.issubset(text2_set) or text2_set.issubset(text1_set):
|
101 |
+
return True
|
102 |
+
|
103 |
+
return False
|
104 |
+
|
105 |
+
|
106 |
+
# checks if the text is a wikipedia page
|
107 |
+
def has_wiki_page(text):
|
108 |
+
results = wiki.search(text)
|
109 |
+
if not results:
|
110 |
+
return False
|
111 |
+
|
112 |
+
title = results[0]
|
113 |
+
p = inflect.engine()
|
114 |
+
singular = p.singular_noun(text)
|
115 |
+
singular = singular if singular else text
|
116 |
+
|
117 |
+
if text.lower() == title.lower() or singular.lower() == title.lower():
|
118 |
+
return True
|
119 |
+
|
120 |
+
return False
|
121 |
+
|
122 |
+
|
123 |
+
# uses the 3 functions above to filter the answer set
|
124 |
+
def filter_answers(set_, text):
|
125 |
+
# make sure the answers are more than two characters to avoid random letter(s) which sometime appear
|
126 |
+
set_ = {tup for tup in set_ if len(tup[0]) > 2}
|
127 |
+
# make sure the answers are not in the question
|
128 |
+
set_ = {tup for tup in set_ if not is_either_text_subset(tup[0], text)}
|
129 |
+
# make sure the all answers have their own Wikipedia pages
|
130 |
+
set_ = {tup for tup in set_ if has_wiki_page(tup[0])}
|
131 |
+
return set_
|
132 |
+
|
133 |
+
|
134 |
+
# gets the text of the first result of the given query in a Wikipedia search
|
135 |
+
def get_wiki_text(query):
|
136 |
+
results = wiki.search(query)
|
137 |
+
top_page = wiki.page(results[0], auto_suggest=False)
|
138 |
+
text = top_page.content
|
139 |
+
|
140 |
+
return text
|
141 |
+
|
142 |
+
|
143 |
+
# splits the text into "chunks" of the requested size
|
144 |
+
def get_text_chunks(text, size):
|
145 |
+
splits = []
|
146 |
+
for i in range(0, len(text), size):
|
147 |
+
splits.append(text[i: i + size])
|
148 |
+
|
149 |
+
return splits
|