/
SemioticSquare.py
213 lines (177 loc) · 7.48 KB
/
SemioticSquare.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
import numpy as np
import pandas as pd
from scattertext.termranking import AbsoluteFrequencyRanker
from scattertext.termscoring.RankDifference import RankDifference
class EmptyNeutralCategoriesError(Exception): pass
'''
!!! Need to properly segregate interfaces
'''
class SemioticSquareBase(object):
def get_labels(self):
raise NotImplementedError()
def get_axes(self, **kwargs):
raise NotImplementedError()
def get_lexicons(self, num_terms=10):
raise NotImplementedError()
class SemioticSquare(SemioticSquareBase):
'''
Create a visualization of a semiotic square. Requires Corpus to have
at least three categories.
>>> newsgroups_train = fetch_20newsgroups(subset='train',
... remove=('headers', 'footers', 'quotes'))
>>> vectorizer = CountVectorizer()
>>> X = vectorizer.fit_transform(newsgroups_train.data)
>>> corpus = st.CorpusFromScikit(
... X=X,
... y=newsgroups_train.target,
... feature_vocabulary=vectorizer.vocabulary_,
... category_names=newsgroups_train.target_names,
... raw_texts=newsgroups_train.data
... ).build()
>>> semseq = SemioticSquare(corpus,
... category_a = 'alt.atheism',
... category_b = 'soc.religion.christian',
... neutral_categories = ['talk.religion.misc']
... )
>>> # A simple HTML table
>>> html = SemioticSquareViz(semseq).to_html()
>>> # The table with an interactive scatterplot below it
>>> html = st.produce_semiotic_square_explorer(semiotic_square,
... x_label='More Atheism, Less Xtnity',
... y_label='General Religious Talk')
'''
def __init__(self,
term_doc_matrix,
category_a,
category_b,
neutral_categories,
labels=None,
term_ranker=AbsoluteFrequencyRanker,
scorer=None):
'''
Parameters
----------
term_doc_matrix : TermDocMatrix
TermDocMatrix (or descendant) which will be used in constructing square.
category_a : str
Category name for term A
category_b : str
Category name for term B (in opposition to A)
neutral_categories : list[str]
List of category names that A and B will be contrasted to. Should be in same domain.
labels : dict
None by default. Labels are dictionary of {'a_and_b': 'A and B', ...} to be shown
above each category.
term_ranker : TermRanker
Class for returning a term-frequency convention_df
scorer : termscoring class, optional
Term scoring class for lexicon mining. Default: `scattertext.termscoring.ScaledFScore`
'''
assert category_a in term_doc_matrix.get_categories()
assert category_b in term_doc_matrix.get_categories()
for category in neutral_categories:
assert category in term_doc_matrix.get_categories()
if len(neutral_categories) == 0:
raise EmptyNeutralCategoriesError()
self.category_a_ = category_a
self.category_b_ = category_b
self.neutral_categories_ = neutral_categories
self._build_square(term_doc_matrix, term_ranker, labels, scorer)
def _build_square(self, term_doc_matrix, term_ranker, labels, scorer):
self.term_doc_matrix_ = term_doc_matrix
self.term_ranker = term_ranker(term_doc_matrix)
self.scorer = RankDifference() \
if scorer is None else scorer
self.axes = self._build_axes(scorer)
self.lexicons = self._build_lexicons()
self._labels = labels
def get_axes(self, scorer=None):
'''
Returns
-------
pd.DataFrame
'''
if scorer:
return self._build_axes(scorer)
return self.axes
def get_lexicons(self, num_terms=10):
'''
Parameters
----------
num_terms, int
Returns
-------
dict
'''
return {k: v.index[:num_terms]
for k, v in self.lexicons.items()}
def get_labels(self):
a = self._get_default_a_label()
b = self._get_default_b_label()
default_labels = {'a': a,
'not_a': 'Not ' + a,
'b': b,
'not_b': 'Not ' + b,
'a_and_b': a + ' + ' + b,
'not_a_and_not_b': 'Not ' + a + ' + Not ' + b,
'a_and_not_b': a + ' + Not ' + b,
'b_and_not_a': 'Not ' + a + ' + ' + b}
labels = self._labels
if labels is None:
labels = {}
return {name + '_label': labels.get(name, default_labels[name])
for name in default_labels}
def _get_default_b_label(self):
return self.category_b_
def _get_default_a_label(self):
return self.category_a_
def _build_axes(self, scorer):
if scorer is None:
scorer = self.scorer
tdf = self._get_term_doc_count_df()
counts = tdf.sum(axis=1)
tdf['x'] = self._get_x_axis(scorer, tdf)
tdf['x'][np.isnan(tdf['x'])] = self.scorer.get_default_score()
tdf['y'] = self._get_y_axis(scorer, tdf)
tdf['y'][np.isnan(tdf['y'])] = self.scorer.get_default_score()
tdf['counts'] = counts
return tdf[['x', 'y', 'counts']]
def _get_x_axis(self, scorer, tdf):
return scorer.get_scores(
tdf[self.category_a_ + ' freq'],
tdf[self.category_b_ + ' freq']
)
def _get_y_axis(self, scorer, tdf):
return scorer.get_scores(
tdf[[t + ' freq' for t in [self.category_a_, self.category_b_]]].sum(axis=1),
tdf[[t + ' freq' for t in self.neutral_categories_]].sum(axis=1)
)
def _get_term_doc_count_df(self):
return (self.term_ranker.get_ranks()
[[t + ' freq' for t in self._get_all_categories()]])
def _get_all_categories(self):
return [self.category_a_, self.category_b_] + self.neutral_categories_
def _build_lexicons(self):
self.lexicons = {}
ax = self.axes
x_max = ax['x'].max()
y_max = ax['y'].max()
x_min = ax['x'].min()
y_min = ax['y'].min()
x_baseline = self._get_x_baseline()
y_baseline = self._get_y_baseline()
def dist(candidates, x_bound, y_bound):
return ((x_bound - candidates['x']) ** 2 + (y_bound - candidates['y']) ** 2).sort_values()
self.lexicons['a'] = dist(ax[(ax['x'] > x_baseline) & (ax['y'] > y_baseline)], x_max, y_max)
self.lexicons['not_a'] = dist(ax[(ax['x'] < x_baseline) & (ax['y'] < y_baseline)], x_min, y_min)
self.lexicons['b'] = dist(ax[(ax['x'] < x_baseline) & (ax['y'] > y_baseline)], x_min, y_max)
self.lexicons['not_b'] = dist(ax[(ax['x'] > x_baseline) & (ax['y'] < y_baseline)], x_max, y_min)
self.lexicons['a_and_b'] = dist(ax[(ax['y'] > y_baseline)], x_baseline, y_max)
self.lexicons['not_a_and_not_b'] = dist(ax[(ax['y'] < y_baseline)], x_baseline, y_min)
self.lexicons['a_and_not_b'] = dist(ax[(ax['x'] > x_baseline)], x_max, y_baseline)
self.lexicons['b_and_not_a'] = dist(ax[(ax['x'] < x_baseline)], x_min, y_baseline)
return self.lexicons
def _get_y_baseline(self):
return self.scorer.get_default_score()
def _get_x_baseline(self):
return self.scorer.get_default_score()