/
autograder.py
244 lines (191 loc) · 8.35 KB
/
autograder.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
'''
Version 0.7
Python 3
'''
import sys
import json
import difflib
from pprint import pprint
from collections import Counter
from nltk.metrics import edit_distance
import gg_api
global toMovie
toMovie = {'johann johannsson': 'the theory of everything', 'alexandre desplat': 'the imitation game', 'trent reznor and atticus ross': 'gone girl', 'antonio sanchez': 'birdman', 'hans zimmer': 'interstellar', 'glory': 'selma', 'big eyes': 'big eyes', 'mercy is': 'noah',
'opportunity': 'annie', 'yellow flicker beat': 'the hunger games mockingjay part 1', 'alejandro gonzalez inarritu': 'birdman', 'wes anderson': 'the grand budapest hotel', 'gillian flynn': 'gone girl', 'richard linklater': 'boyhood', 'graham moore': 'the imitation game'}
def norm_text(textstring):
"""Takes a string of text and returns a string of normalized text."""
return "".join([c.lower() for c in textstring if c.isalnum() or c.isspace()])
def text(resultstr, answerstr):
"""Accepts two normalized texts, as output by the norm_text
function, and returns a score based on the match length relative
to the longest text length."""
result = resultstr.split()
answer = answerstr.split()
len_result = len(result)
len_answer = len(answer)
if (resultstr in answerstr) or (answerstr in resultstr):
textscore = min(len_result, len_answer) / \
float(max(len_result, len_answer))
else:
s = difflib.SequenceMatcher(None, result, answer)
longest = s.find_longest_match(0, len_result, 0, len_answer)
longest = longest.size/float(max(len_result, len_answer))
if longest > 0.3:
matchlen = sum([m[2] for m in s.get_matching_blocks() if m[2] > 1])
textscore = float(matchlen)/max(len_result, len_answer)
else:
textscore = longest
return textscore
def spell_check(r, a, s, scores, weight=1):
change = weight*(1-(edit_distance(r, a)/float(max(len(r), len(a)))))
if s in scores:
# penalty for returning multiple of the same result when
# one instance is incorrectly spelled
return (scores[s] + change)/2.0
else:
return change
def calc_translation(result, answer):
'''Accepts two lists of strings, determines the best matches
between them, and returns a translation dictionary and
score.'''
resultmap = {norm_text(r): r for r in result}
answermap = {norm_text(a): a for a in answer}
result = set(resultmap.keys())
answer = set(answermap.keys())
intersection = result.intersection(answer)
translation = {resultmap[i]: answermap[i] for i in intersection}
scores = dict(list(zip(list(translation.values()), [1]*len(intersection))))
score_by_results = {}
score_by_answers = {}
# loop through results that didn't have a perfect match
# and get a score for each of them.
comp = list(result - intersection)
for r in comp:
score_by_results[r] = Counter()
for a in answer:
if a not in score_by_answers:
score_by_answers[a] = Counter()
score_by_results[r][a] = text(r, a)
score_by_answers[a][r] = score_by_results[r][a]
for r in score_by_results:
cnt = 0
ranking = score_by_results[r].most_common()
flag = True
while flag:
# The answer that best matches the result
answer_match = ranking[cnt][0]
# The top result matching that answer
max_result = score_by_answers[answer_match].most_common(1)[0]
if score_by_results[r][answer_match] < 0.45:
bestAnswer = False
score = 0
# Unacceptably low score.
# Check if we have a case of returning the movie instead
# of the person, or vice versa.
for ha in toMovie:
tempScore = text(r, ha)
if tempScore > score:
score = tempScore
bestAnswer = ha
if bestAnswer and score > 0.45:
translation[resultmap[r]] = toMovie[ha]
scores[toMovie[ha]] = spell_check(
r, ha, toMovie[ha], scores, 0.5)
flag = False
elif (max_result[0] == r) or (score_by_results[r][answer_match] > score_by_answers[answer_match][max_result[0]]):
# if the top result matching that answer is our current result or
# if the current result's score is greater than the previous top result
translation[resultmap[r]] = answermap[answer_match]
scores[answermap[answer_match]] = spell_check(
r, answer_match, answer_match, scores)
flag = False
cnt += 1
if cnt == len(ranking):
flag = False
if scores:
return sum(scores.values())/float(len(scores)), translation
else:
return 0, translation
def calc_score(result, answer):
result = set(result)
intersection = result.intersection(answer)
len_intersection = len(intersection)
len_union = len(result.union(answer))
len_result = len(result)
len_answer = len(answer)
if len_union == 0:
return 0
elif len_result == len_answer and len_intersection == len_answer:
m = 1.0
elif len_intersection == len_result:
# all results correspond to a correct answer, but some
# answers are missing
m = 0.95
elif len_intersection == len_answer:
# all answers correspond to a result, but there are
# some extra results as well
m = 0.9
elif len_intersection > 0:
# there is some post-translation intersection between
# results and answers.
m = 0.85
else:
return 0
return (len_intersection / float(len_union)) * m
def score_structured(year, answers, info_type):
# c_score is the completeness score
spelling_score = 0
c_score = 0
results = getattr(gg_api, 'get_%s' % info_type)(year)
length = 26
if info_type == "nominees":
tempans = answers['award_data']['cecil b. demille award']
del answers['award_data']['cecil b. demille award']
tempres = results['cecil b. demille award']
del results['cecil b. demille award']
length = 25
for a in answers['award_data']:
if info_type == 'winner':
temp_spelling, translation = calc_translation(
[results[a]], [answers['award_data'][a][info_type]])
else:
temp_spelling, translation = calc_translation(
results[a], answers['award_data'][a][info_type])
c_score += calc_score([translation[res] if res in translation else res for res in results[a]],
answers['award_data'][a][info_type])
spelling_score += temp_spelling
if info_type == "nominees":
answers['award_data']['cecil b. demille award'] = tempans
results['cecil b. demille award'] = tempres
return spelling_score/length, c_score/length
def score_unstructured(year, answers, info_type):
results = getattr(gg_api, 'get_%s' % info_type)(year)
spelling_score, translation = calc_translation(results, answers[info_type])
c_score = calc_score(
[translation[res] if res in translation else res for res in results], answers[info_type])
return spelling_score, c_score
def main(grading):
years = ['2013']
types = ['spelling', 'completeness']
scores = {y: {g: {t: 0 for t in types} for g in grading} for y in years}
for y in years:
with open('gg%sanswers.json' % y, 'r') as f:
answers = json.load(f)
answers['awards'] = list(answers['award_data'].keys())
for g in grading:
if g in ['hosts', 'awards']:
scores[y][g]['spelling'], scores[y][g]['completeness'] = score_unstructured(
y, answers, g)
else:
scores[y][g]['spelling'], scores[y][g]['completeness'] = score_structured(
y, answers, g)
if "winner" in grading:
del scores[y]['winner']['completeness']
pprint(scores)
if __name__ == '__main__':
grading = ["hosts", "awards", "nominees", "presenters", "winner"]
if len(sys.argv) > 1:
newg = [g for g in grading if g in sys.argv]
if len(newg) > 0:
grading = newg
main(grading)