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
|
"""Information Content is a corpus-based metrics of synset or sense
specificity.
"""
from collections import Counter
from collections.abc import Callable, Iterable, Iterator
from math import log
from pathlib import Path
from typing import TextIO, TypeAlias
from wn import Synset, Wordnet
from wn._types import AnyPath
from wn.constants import ADJ, ADJ_SAT, ADV, NOUN, VERB
from wn.util import synset_id_formatter
# Just use a subset of all available parts of speech
IC_PARTS_OF_SPEECH = frozenset((NOUN, VERB, ADJ, ADV))
Freq: TypeAlias = dict[str, dict[str | None, float]]
def information_content(synset: Synset, freq: Freq) -> float:
"""Calculate the Information Content value for a synset.
The information content of a synset is the negative log of the
synset probability (see :func:`synset_probability`).
"""
return -log(synset_probability(synset, freq))
def synset_probability(synset: Synset, freq: Freq) -> float:
"""Calculate the synset probability.
The synset probability is defined as freq(ss)/N where freq(ss) is
the IC weight for the synset and N is the total IC weight for all
synsets with the same part of speech.
Note: this function is not generally used directly, but indirectly
through :func:`information_content`.
"""
pos_freq = freq[synset.pos]
return pos_freq[synset.id] / pos_freq[None]
def _initialize(
wordnet: Wordnet,
smoothing: float,
) -> Freq:
"""Populate an Information Content weight mapping to a smoothing value.
All synsets in *wordnet* are inserted into the dictionary and
mapped to *smoothing*.
"""
freq: Freq = {
pos: {synset.id: smoothing for synset in wordnet.synsets(pos=pos)}
for pos in IC_PARTS_OF_SPEECH
}
# pretend ADJ_SAT is just ADJ
for synset in wordnet.synsets(pos=ADJ_SAT):
freq[ADJ][synset.id] = smoothing
# also initialize totals (when synset is None) for each part-of-speech
for pos in IC_PARTS_OF_SPEECH:
freq[pos][None] = smoothing
return freq
def compute(
corpus: Iterable[str],
wordnet: Wordnet,
distribute_weight: bool = True,
smoothing: float = 1.0,
) -> Freq:
"""Compute Information Content weights from a corpus.
Arguments:
corpus: An iterable of string tokens. This is a flat list of
words and the order does not matter. Tokens may be single
words or multiple words separated by a space.
wordnet: An instantiated :class:`wn.Wordnet` object, used to
look up synsets from words.
distribute_weight: If :python:`True`, the counts for a word
are divided evenly among all synsets for the word.
smoothing: The initial value given to each synset.
Example:
>>> import wn, wn.ic, wn.morphy
>>> ewn = wn.Wordnet("ewn:2020", lemmatizer=wn.morphy.morphy)
>>> freq = wn.ic.compute(["Dogs", "run", ".", "Cats", "sleep", "."], ewn)
>>> dog = ewn.synsets("dog", pos="n")[0]
>>> cat = ewn.synsets("cat", pos="n")[0]
>>> frog = ewn.synsets("frog", pos="n")[0]
>>> freq["n"][dog.id]
1.125
>>> freq["n"][cat.id]
1.1
>>> freq["n"][frog.id] # no occurrence; smoothing value only
1.0
>>> carnivore = dog.lowest_common_hypernyms(cat)[0]
>>> freq["n"][carnivore.id]
1.3250000000000002
"""
freq = _initialize(wordnet, smoothing)
counts = Counter(corpus)
hypernym_cache: dict[Synset, list[Synset]] = {}
for word, count in counts.items():
synsets = wordnet.synsets(word)
num = len(synsets)
if num == 0:
continue
weight = float(count / num if distribute_weight else count)
for synset in synsets:
pos = synset.pos
if pos == ADJ_SAT:
pos = ADJ
if pos not in IC_PARTS_OF_SPEECH:
continue
freq[pos][None] += weight
# The following while-loop is equivalent to:
#
# freq[pos][synset.id] += weight
# for path in synset.hypernym_paths():
# for ss in path:
# freq[pos][ss.id] += weight
#
# ...but it caches hypernym lookups for speed
agenda: list[tuple[Synset, set[Synset]]] = [(synset, set())]
while agenda:
ss, seen = agenda.pop()
# avoid cycles
if ss in seen:
continue
freq[pos][ss.id] += weight
if ss not in hypernym_cache:
hypernym_cache[ss] = ss.hypernyms()
agenda.extend((hyp, seen | {ss}) for hyp in hypernym_cache[ss])
return freq
def load(
source: AnyPath,
wordnet: Wordnet,
get_synset_id: Callable | None = None,
) -> Freq:
"""Load an Information Content mapping from a file.
Arguments:
source: A path to an information content weights file.
wordnet: A :class:`wn.Wordnet` instance with synset
identifiers matching the offsets in the weights file.
get_synset_id: A callable that takes a synset offset and part
of speech and returns a synset ID valid in *wordnet*.
Raises:
:class:`wn.Error`: If *wordnet* does not have exactly one
lexicon.
Example:
>>> import wn, wn.ic
>>> pwn = wn.Wordnet("pwn:3.0")
>>> path = "~/nltk_data/corpora/wordnet_ic/ic-brown-resnik-add1.dat"
>>> freq = wn.ic.load(path, pwn)
"""
source = Path(source).expanduser().resolve(strict=True)
assert len(wordnet.lexicons()) == 1
lexid = wordnet.lexicons()[0].id
if get_synset_id is None:
get_synset_id = synset_id_formatter(prefix=lexid)
freq = _initialize(wordnet, 0.0)
with source.open() as icfile:
for offset, pos, weight, is_root in _parse_ic_file(icfile):
ssid = get_synset_id(offset=offset, pos=pos)
# synset = wordnet.synset(ssid)
freq[pos][ssid] = weight
if is_root:
freq[pos][None] += weight
return freq
def _parse_ic_file(icfile: TextIO) -> Iterator[tuple[int, str, float, bool]]:
"""Parse the Information Content file.
A sample of the format is::
wnver::eOS9lXC6GvMWznF1wkZofDdtbBU
1740n 1915712 ROOT
1930n 859272
2137n 1055337
"""
next(icfile) # skip header
for line in icfile:
ssinfo, value, *isroot = line.split()
yield (int(ssinfo[:-1]), ssinfo[-1], float(value), bool(isroot))
|
