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
|
#-*- coding:utf-8 -*-
###############################################################
# PyNLPl - DutchSemCor
# by Maarten van Gompel (proycon)
# http://ilk.uvt.nl/~mvgompel
# Induction for Linguistic Knowledge Research Group
# Universiteit van Tilburg
#
# Licensed under GPLv3
#
# Modified by Ruben Izquierdo
# We need also to store the TIMBL distance to the nearest neighboor
#
# Collection of formats for the DutchSemCor project
#
###############################################################
from __future__ import print_function
from __future__ import unicode_literals
from __future__ import division
from __future__ import absolute_import
from pynlpl.common import u
import sys
if sys.version < '3':
from codecs import getwriter
stderr = getwriter('utf-8')(sys.stderr)
stdout = getwriter('utf-8')(sys.stdout)
else:
stderr = sys.stderr
stdout = sys.stdout
from pynlpl.formats.timbl import TimblOutput
from pynlpl.statistics import Distribution
import io
class WSDSystemOutput(object):
def __init__(self, filename = None):
self.data = {}
self.distances={}
self.maxDistance=1
if filename:
self.load(filename)
def append(self, word_id, senses,distance=0):
# Commented by Ruben, there are some ID's that are repeated in all sonar test files...
#assert (not word_id in self.data)
if isinstance(senses, Distribution):
self.data[word_id] = ( (x,y) for x,y in senses ) #PATCH UNDONE (#TODO: this is a patch, something's not right in Distribution?)
self.distances[word_id]=distance
if distance > self.maxDistance:
self.maxDistance=distance
return
else:
assert isinstance(senses, list) and len(senses) >= 1
self.distances[word_id]=distance
if distance > self.maxDistance:
self.maxDistance=distance
if len(senses[0]) == 1:
#not a (sense_id, confidence) tuple! compute equal confidence for all elements automatically:
confidence = 1 / float(len(senses))
self.data[word_id] = [ (x,confidence) for x in senses ]
else:
fulldistr = True
for sense, confidence in senses:
if confidence == None:
fulldistr = False
break
if fulldistr:
self.data[word_id] = Distribution(senses)
else:
self.data[word_id] = senses
def getMaxDistance(self):
return self.maxDistance
def __iter__(self):
for word_id, senses in self.data.items():
yield word_id, senses,self.distances[word_id]
def __len__(self):
return len(self.data)
def __getitem__(self, word_id):
"""Returns the sense distribution for the given word_id"""
return self.data[word_id]
def load(self, filename):
f = io.open(filename,'r',encoding='utf-8')
for line in f:
fields = line.strip().split(" ")
word_id = fields[0]
if len(fields[1:]) == 1:
#only one sense, no confidence expressed:
self.append(word_id, [(fields[1],None)])
else:
senses = []
distance=-1
for i in range(1,len(fields),2):
if i+1==len(fields):
#The last field is the distance
if fields[i][:4]=='+vdi': #Support for previous format of wsdout
distance=float(fields[i][4:])
else:
distance=float(fields[i])
else:
if fields[i+1] == '?': fields[i+1] = None
senses.append( (fields[i], fields[i+1]) )
self.append(word_id, senses,distance)
f.close()
def save(self, filename):
f = io.open(filename,'w',encoding='utf-8')
for word_id, senses,distance in self:
f.write(word_id)
for sense, confidence in senses:
if confidence == None: confidence = "?"
f.write(" " + str(sense) + " " + str(confidence))
if word_id in self.distances.keys():
f.write(' '+str(self.distances[word_id]))
f.write("\n")
f.close()
def out(self, filename):
for word_id, senses,distance in self:
print(word_id,distance,end="")
for sense, confidence in senses:
if confidence == None: confidence = "?"
print(" " + sense + " " + str(confidence),end="")
print()
def senses(self, bestonly=False):
"""Returns a list of all predicted senses"""
l = []
for word_id, senses,distance in self:
for sense, confidence in senses:
if not sense in l: l.append(sense)
if bestonly:
break
return l
def loadfromtimbl(self, filename):
timbloutput = TimblOutput(io.open(filename,'r',encoding='utf-8'))
for i, (features, referenceclass, predictedclass, distribution, distance) in enumerate(timbloutput):
if distance != None:
#distance='+vdi'+str(distance)
distance=float(distance)
if len(features) == 0:
print("WARNING: Empty feature vector in " + filename + " (line " + str(i+1) + ") skipping!!",file=stderr)
continue
word_id = features[0] #note: this is an assumption that must be adhered to!
if distribution:
self.append(word_id, distribution,distance)
def fromTimblToWsdout(self,fileTimbl,fileWsdout):
timbloutput = TimblOutput(io.open(fileTimbl,'r',encoding='utf-8'))
wsdoutfile = io.open(fileWsdout,'w',encoding='utf-8')
for i, (features, referenceclass, predictedclass, distribution, distance) in enumerate(timbloutput):
if len(features) == 0:
print("WARNING: Empty feature vector in " + fileTimbl + " (line " + str(i+1) + ") skipping!!",file=stderr)
continue
word_id = features[0] #note: this is an assumption that must be adhered to!
if distribution:
wsdoutfile.write(word_id+' ')
for sense, confidence in distribution:
if confidence== None: confidence='?'
wsdoutfile.write(sense+' '+str(confidence)+' ')
wsdoutfile.write(str(distance)+'\n')
wsdoutfile.close()
class DataSet(object): #for testsets/trainingsets
def __init__(self, filename):
self.sense = {} #word_id => (sense_id, lemma,pos)
self.targetwords = {} #(lemma,pos) => [sense_id]
f = io.open(filename,'r',encoding='utf-8')
for line in f:
if len(line) > 0 and line[0] != '#':
fields = line.strip('\n').split('\t')
word_id = fields[0]
sense_id = fields[1]
lemma = fields[2]
pos = fields[3]
self.sense[word_id] = (sense_id, lemma, pos)
if not (lemma,pos) in self.targetwords:
self.targetwords[(lemma,pos)] = []
if not sense_id in self.targetwords[(lemma,pos)]:
self.targetwords[(lemma,pos)].append(sense_id)
f.close()
def __getitem__(self, word_id):
return self.sense[self._sanitize(word_id)]
def getsense(self, word_id):
return self.sense[self._sanitize(word_id)][0]
def getlemma(self, word_id):
return self.sense[self._sanitize(word_id)][1]
def getpos(self, word_id):
return self.sense[self._sanitize(word_id)][2]
def _sanitize(self, word_id):
return u(word_id)
def __contains__(self, word_id):
return (self._sanitize(word_id) in self.sense)
def __iter__(self):
for word_id, (sense, lemma, pos) in self.sense.items():
yield (word_id, sense, lemma, pos)
def senses(self, lemma, pos):
return self.targetwords[(lemma,pos)]
|
