forked from TurkuNLP/Finnish-dep-parser
-
Notifications
You must be signed in to change notification settings - Fork 1
/
resolve_readings.py
264 lines (247 loc) · 11.5 KB
/
resolve_readings.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
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
# -*- coding: utf-8 -*-
##TODO: Warning: if there are words with +-character and if that word gets more than one reading, I don't know what happens.
import re
import sys
import omorfi_pos
import omorfi_wrapper
import os
SCRIPTDIR=os.path.dirname(os.path.abspath(__file__))
OmorTransducer=omorfi_wrapper.OmorfiWrapper(os.path.join(SCRIPTDIR,"model/generation.finntreebank.hfstol"))
atRe=re.compile(ur"@[A-Za-z.]+@",re.U) #The @....@ tags in the transducer output, want to get rid of them
############################################################################################
# SUPPORT FUNCTIONS:
#True, if one or more readings is compound
def isCompound(readings):
for r in readings:
if u"<Cmpnd>" in r: return True
elif u"+#" in r: return True
return False
#True, if one or more readings is derivation
def isDerivation(readings):
for r in readings:
if u"<Der_" in r: return True
return False
#Takes one reading and returns the reading of last member if compound and reading without lemma if normal
def give_last_one(r):
if u"+" in r:
columns=r.split(u"+")
ending=columns[len(columns)-1]
else: ending=r
index=0
for char in ending:
if char!=u"<": index+=1
else: break
ending=ending[index:len(ending)]
return ending
#Makes all necessary der changes to reading r
def der_changes(r):
regex="(<Der_[A-Za-z]+>)"
if u"<Der_" in r:
m=re.findall(regex,r)
Der=m[len(m)-1]
columns=r.split(Der)
r=columns[len(columns)-1]
if Der==u"<Der_u>": r=u"<N>"+r #because of missing <N>-tag
elif Der==u"<Der_llinen>": r=re.sub(u"<A>",u"<A><Pos>",r) #because of missing <Pos>-tag
elif Der==u"<Der_ton>": r=re.sub(u"<A>",u"<A><Pos>",r)
elif Der==u"<Der_tse>": r=re.sub(u"<Adv><Prl>",u"<Adv>",r)
elif Der==u"<Der_ttain>": r=re.sub(u"<Adv><Dis>",u"<Adv>",r)
elif Der==u"<Der_sti>": r=re.sub(u"<Adv><Comp>",u"<Adv>",r)
if u"[DRV=UUS]" in r: r=re.sub(u"\[DRV=UUS\]",u"",r) #deletes DRV=UUS-tag
r=re.sub(u"<cap>",u"",r)
r=re.sub(u"<Cap>",u"",r)
r=re.sub(u"<CAP>",u"",r)
return r
#Takes one COMPOUND-reading, returns lemma (like "raja jääkäri pataljoona")
def give_cmpnd_lemma(r):
lemma=u""
columns=r.split(u"+")
for col in columns:
if u"<" in col:
c=col.split(u"<")
lemma+=c[0]+u" "
else: lemma+=col+u" "
lemma=lemma.strip()
lemma=re.sub(u"#",u"",lemma)
return lemma
##############################################################################################
# COMPARING FUNCTIONS
#Takes two der-readings, returns True if considered as same
def der_vs_der(r1,r2):
regex="(<Der_[A-Za-z]+>)"
m1=re.findall(regex,r1)
m2=re.findall(regex,r2)
if len(m1)>len(m2):
Der=m1[0]
columns=r1.split(Der)
if len(columns)==2: r1_ending=columns[1]
else: return False
r2_ending=give_last_one(r2)
if r1_ending==r2_ending: return True
else: return False
elif len(r2)>len(r1):
Der=m2[0]
columns=r2.split(Der)
if len(columns)==2: r2_ending=columns[1]
else: return False
r1_ending=give_last_one(r1)
if r2_ending==r1_ending: return True
else: return False
else:
cols=r1.split(u"<")
lemma1=cols[0].lower()
cols=r2.split(u"<")
lemma2=cols[0].lower()
if lemma1==lemma2: return True
else: return False
#Takes two readings, checks if the readings of last members are same (this one just skips lemmas)
def isSameAnalysis(r1,r2):
r1_ending=give_last_one(r1)
r2_ending=give_last_one(r2)
r1_ending=der_changes(r1_ending)
r2_ending=der_changes(r2_ending)
if r1_ending==r2_ending: return True
else: return False
#Splits compounds to parts and compares those, returns True, if all parts match
def split_compounds(r1,r2):
parts1=r1.split(u"+")
parts2=r2.split(u"+")
if len(parts1)==len(parts2):
for i in range(0,len(parts1)):
part1=parts1[i]
part1=re.sub(u"#",u"",part1)
part1=re.sub(u"<Cmpnd>",u"",part1)
part2=parts2[i]
part2=re.sub(u"#",u"",part2)
part2=re.sub(u"<Cmpnd>",u"",part2)
if u"<" in part1 and u"<" in part2: #both have reading, no talvi+#uni cases
if u"<Der_" in part1 or u"<Der_" in part2: #if other have der-tag check also readings
if isSameLemma(part1,part2)==False or isSameAnalysis(part1,part2)==False: return False
else:
if not isSameLemma(part1,part2): return False
else: # other (or both) doesn't have tags, use omorfi generation if needed
if u"<" not in part1 and u"<" not in part2: #both #-tag compounds
if part1 != part2: return False
elif u"<" in part1: #part1 has reading
cols=part1.split(u"<")
lemma1=cols[0]
if lemma1==part2: continue #lemma is same, continue to next part
#different lemma, try omorfi generation
if OmorTransducer==None: raise ValueError("Omorfi is not loaded correctly, cannot do lookup")
wordforms=[atRe.sub(u"",unicode(r,"utf-8")) for r,score in OmorTransducer.lookup(part1.encode("utf-8"))]
match=False
for form in wordforms:
form=re.sub(u"-",u"",form)
form=re.sub(u"\u2010",u"",form)
if form==part2:
match=True
break
if not match: return False #no match using omorfi generation
else:#same but part2 has reading
cols=part2.split(u"<")
lemma2=cols[0]
if lemma2==part1: continue
if OmorTransducer==None: raise ValueError("Omorfi is not loaded correctly, cannot do lookup")
wordforms=[atRe.sub(u"",unicode(r,"utf-8")) for r,score in OmorTransducer.lookup(part2.encode("utf-8"))]
match=False
for form in wordforms:
form=re.sub(u"-",u"",form)
form=re.sub(u"\u2010",u"",form)
if form==part1:
match=True
break
if not match: return False
#same len and all part matches --> is same
return True
else: return False # different len, can't be same
#Takes two readings and checks if lemmas are same
#NORMAL CMPND: Lemmas has to be exactly same, or omorfi generated
#NORMAL Der_: 1) If der vs. normal, returns always True (because lemmas are different, can't know for real) 2) If der vs der, calls der_vs_der()
#BOTH CMPND AND DER: splits to sub words and compare those
def isSameLemma(r1,r2):
if u"Cmpnd" in r1 or u"+#" in r1:
if u"Cmpnd" in r2 or u"+#" in r2:
if u"Der_" in r1 or u"Der_" in r2: return split_compounds(r1,r2) #BOTH COMPOUNDS + DERIVATION AT LEAST IN OTHER
else: #both are COMPOUNDS WITHOUT DERIVATIONS
if give_cmpnd_lemma(r1).lower()==give_cmpnd_lemma(r2).lower(): return True
else: return split_compounds(r1,r2) #try generated forms
else: #r1 is compound, r2 isn't --> if lemmas are same, true --> isoisä vs. iso|isä
compound=give_cmpnd_lemma(r1).lower()
compound_lemma=re.sub(u"\s",u"",compound)
compound_hash=re.sub(u"\s",u"-",compound) #try also put "-" between, beacause omorfi drops those
columns=r2.split("<")
lemma=columns[0].lower()
if compound_lemma==lemma or compound_hash==lemma: return True
else: return False
elif u"Cmpnd" in r2 or u"+#" in r2: #r2 is compound, r1 isn't --> if lemmas are same, true
compound=give_cmpnd_lemma(r2).lower()
compound_lemma=re.sub(u"\s",u"",compound)
compound_hash=re.sub(u"\s",u"-",compound)
columns=r1.split("<")
lemma=columns[0].lower()
if compound_lemma==lemma or compound_hash==lemma: return True
else: return False
elif u"Der_" in r1 or u"Der_" in r2: #JUST DERIVATIONS:
if u"Der_" in r1 and u"Der_" in r2: return der_vs_der(r1,r2) #both are derivations
else: return True
else:#both are NORMALS
columns=r1.split("<")
lemma1=columns[0].lower()
columns=r2.split("<")
lemma2=columns[0].lower()
if lemma1==lemma2: return True
else: return False
#compare 1vs1,return better or None if equal, better is the one with less Der_-tags or if same number, more #-tags
def compare(r1,r2):
if isSameLemma(r1,r2) and isSameAnalysis(r1,r2):
compound_count1=re.findall(u"\+",r1)
compound_count2=re.findall(u"\+",r2)
der_count1=re.findall(u"Der_",r1)
der_count2=re.findall(u"Der_",r2)
if len(compound_count1) < len(compound_count2): return r1
elif len(compound_count2) < len(compound_count1): return r2
elif len(der_count1) < len(der_count2): return r1
elif len(der_count2) < len(der_count1): return r2
else:
hash_count1=re.findall(u"#",r1)
hash_count2=re.findall(u"#",r2)
if len(hash_count2)>len(hash_count1): return r2
else: return r1
else: return None
####################################################################################################
# MAINS, GO TROUGHT READINGS
#Takes readings (set) and compares those 1vs1 --> if same, marks notwanted with "???". Returns readings (set) with no "???" in them
def handle_cmpnds_and_ders(readings):
readings=list(readings) #convert set to list, because we need indices.
for index in range(0,len(readings)):
r=re.sub(u"\+\+#",u"+#",readings[index]) #because some adj-der-cmpnd:s have ++# instead of +#
#example: pidempijaksoinen pitkä<A><Comp><Sg><Nom><Cmpnd>++#jaksoinen<A><Pos><Sg><Nom><cap>
if r.startswith(u"???"): r=re.sub(u"\?\?\?",u"",r)
for i in range(index+1,len(readings)):
if readings[index].startswith(u"???") and readings[i].startswith(u"???"):
continue #both discarded --> no need to compare berween these
else:
other=re.sub(u"\+\+#",u"+#",readings[i])
other=re.sub(u"\?\?\?",u"",other)
better=compare(r,other)
if better!=None: #mark the reading we don't want choose
if better==r:
if not readings[i].startswith(u"???"):
readings[i]=u"???"+readings[i]
else:
if not readings[index].startswith(u"???"):
readings[index]=u"???"+readings[index]
left_readings=set()
for r in readings:
if not r.startswith(u"???"): left_readings.add(r)
assert len(left_readings)!=0, u"Every reading discarded"
return left_readings
# main function, calls handle_cmpnds_and_ders() if needed, otherwise just returns same set
def main(readings):
if len(readings)>1:
if isCompound(readings) or isDerivation(readings):
return handle_cmpnds_and_ders(readings)
else: #all readings are normal, return just same list
return readings
else: #just one reading, just return it
return readings