-
Notifications
You must be signed in to change notification settings - Fork 5
/
Copy pathcoptic_nlp.py
1349 lines (1191 loc) · 52.7 KB
/
coptic_nlp.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
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
#!/usr/bin/python
# -*- coding: utf-8 -*-
import re, sys, io, os, platform
import tempfile
import subprocess
from glob import glob
from _version import __version__
from lib.stacked_tokenizer import StackedTokenizer
from lib.order_meta_tag import reorder
from lib.reorder_sgml import reorder as reorder_sgml
from lib.ekthetic_para import ekthetic_to_para
from lib.tt2conll import conllize
from lib.depedit import DepEdit
from lib.binarize_tags import binarize
from lib.lang import lookup_lang
from lib.harvest_tt_sgml import harvest_tt
from lib.mwe import tag_mwes
from lib.marmot import tag_marmot
from lib.flair_pos_tagger import FlairTagger
from lib.lemmatize import Lemmatizer
from lib.heads import assign_entity_heads
PY3 = sys.version_info[0] > 2
if PY3:
from builtins import type
inp = input if PY3 else raw_input
script_dir = os.path.dirname(os.path.realpath(__file__))
lib_dir = script_dir + os.sep + "lib" + os.sep
bin_dir = script_dir + os.sep + "bin" + os.sep
data_dir = script_dir + os.sep + "data" + os.sep
parser_path = bin_dir + "maltparser-1.8" + os.sep
tt_path = bin_dir + "TreeTagger" + os.sep + "bin" + os.sep
ud_coptic_path = script_dir + os.sep + "UD_Coptic-Scriptorium" + os.sep # Optional path to UD_Coptic-Scriptorium - use to cache gold parses
from diaparser.parsers.parser import Parser as NeuralParser
neural_model = script_dir + os.sep + "lib" + os.sep + "cop.diaparser"
from lib.quiet import suppress_stdout_stderr # Context to suppress stderr messages from imported libraries
# Global cache of gold syntax trees, if UD_Coptic-Scriptorium data is available
gold_trees = {}
# Place holder for entity linking module
identifier = None
def get_gold_trees():
def conll2mapping(conllu):
out_dict = {}
sents = conllu.strip().replace("\r","").split("\n\n")
for sent in sents:
out_sent = []
tokens = []
for line in sent.split("\n"):
if "\t" in line:
fields = line.split("\t")
if "-" not in fields[0]:
tokens.append(fields[1])
fields[5], fields[-2], fields[-1] = "_", "_", "_" # Kill MISC, FEATS
fields[3] = fields[4] # Kill UPOS
out_sent.append("\t".join(fields))
out_dict[" ".join(tokens)] = "\n".join(out_sent)
return out_dict
global ud_coptic_path
global gold_trees
partition_files = ["train","test","dev"]
if os.path.exists(ud_coptic_path + 'cop_scriptorium-ud-train.conllu'):
partition_files = [ud_coptic_path + "cop_scriptorium-ud-" + f + ".conllu" for f in partition_files]
for f in partition_files:
gold_trees.update(conll2mapping(io.open(f,encoding="utf8").read()))
def replace_trees(conllu, gold_trees):
sents = conllu.strip().replace("\r", "").split("\n\n")
output = []
for sent in sents:
tokens = []
for line in sent.split("\n"):
if "\t" in line:
fields = line.split("\t")
if "-" not in fields[0]:
tokens.append(fields[1])
plain = " ".join(tokens)
if plain in gold_trees:
output.append(gold_trees[plain])
else:
output.append(sent)
return "\n\n".join(output) + "\n\n"
# Global lookup lemmatizer
lemmatizer = None
flair_tagger = None
def detect_lang(text):
# Split text into units agnostic of whether it's XML, TT SGML, plain text with pipes...
if "|" in text:
text = text.replace("|", "") # Ensure we can see whole bound groups
units = re.split(r'[\s"_-]+', text)
units = set(units)
if "ϧ" in text:
if "ⲡⲛⲟⲩⲧⲉ" not in units and "ⲁⲩⲱ" not in units:
return "bohairic"
elif "ⲟⲩⲟϩ" in units and "ⲁⲩⲱ" not in units:
return "bohairic"
elif "ⲫⲛⲟⲩϯ" in units and "ⲡⲛⲟⲩⲧⲉ" not in units:
return "bohairic"
return "sahidic"
def log_tasks(opts):
sys.stderr.write("\nRunning standard tasks:\n" +"="*20 + "\n")
if opts.dialect == "auto":
sys.stderr.write("o Auto detect dialect\n")
if opts.unary:
sys.stderr.write("o Binarize unary XML milestone tags\n")
if opts.para:
sys.stderr.write("o Paragraph detection\n")
if opts.meta:
sys.stderr.write("o Metadata insertion\n")
if not opts.no_tok and not opts.parse_only and not opts.merge_parse:
if opts.from_pipes:
sys.stderr.write("o Tokenization (from pipes)\n")
else:
sys.stderr.write("o Tokenization\n")
if opts.norm:
sys.stderr.write("o Normalization\n")
if opts.tag:
sys.stderr.write("o POS tagging\n")
if opts.lemma:
sys.stderr.write("o Lemmatization\n")
if opts.etym:
sys.stderr.write("o Language of origin detection\n")
if opts.multiword:
sys.stderr.write("o Multiword expression recognition\n")
if opts.sent is not None:
sys.stderr.write("o Splitting sentences based on tag: "+opts.sent+"\n")
if opts.parse or opts.parse_only or opts.merge_parse:
sys.stderr.write("o Dependency parsing\n")
if opts.recognize_entities:
sys.stderr.write("o Entity recognition\n")
if opts.identities:
sys.stderr.write("o Wikification\n")
special_tasks = []
if opts.space:
special_tasks.append("o Space out punctuation")
if opts.detokenize > 0:
if opts.detokenize == 2:
special_tasks.append("o Detokenization (a.k.a. 'Laytonization') - aggressive")
elif opts.detokenize == 3:
special_tasks.append("o Detokenization (a.k.a. 'Laytonization') - smart")
else:
special_tasks.append("o Detokenization (a.k.a. 'Laytonization') - conservative")
if opts.segment_merged:
special_tasks.append("o Insert boundary between merged groups")
if opts.breaklines:
special_tasks.append("o Add line tags to preserve line breaks")
if opts.merge_parse:
special_tasks.append("o Merge parse into SGML file")
if opts.para:
special_tasks.append('o Add paragraphs based on <hi rend="ekthetic">')
if opts.parse_only:
special_tasks.append("o Parse to CoNLL file")
if len(special_tasks) > 0:
sys.stderr.write("\nRunning special tasks:\n" + "=" * 20 + "\n")
sys.stderr.write("\n".join(special_tasks) + "\n")
sys.stderr.write("\n")
def groupify(output,anno):
groups = ""
current_group = ""
for line in output.split("\n"):
if " "+anno+"=" in line:
current_group += re.search(anno + r'="([^"]*)"',line).group(1)
if line.startswith("</") and "_group" in line:
groups += current_group + "\n"
current_group = ""
return groups
def remove_nesting_attr(data, nester, nested, attr="xml:lang"):
"""
Removes attribute on nesting element if a nested element includes it
:param data: SGML input
:param nester: nesting tag, e.g. "norm"
:param nested: nested tag, e.g. "morph"
:param attr: attribute, e.g. "lang"
:return: cleaned SGML
"""
if attr not in data:
return data
flagged = []
in_attr_nester = False
last_nester = -1
lines = data.split("\n")
for i, line in enumerate(lines):
if nester + "=" in line and attr+"=" in line:
in_attr_nester = True
last_nester = i
if "</" + nester + ">" in line:
in_attr_nester = False
if nested in line and attr+"=" in line and in_attr_nester and last_nester > -1:
flagged.append(last_nester)
in_attr_nester = False
for i in flagged:
lines[i] = re.sub(' '+attr+'="[^"]+"','',lines[i])
return "\n".join(lines)
def tok_from_norm(data):
"""
Takes TT-SGML, extracts norm attribute, and replaces existing tokens with norm values while retaining SGML tags.
Used to feed parser norms while retaining SGML sentence separators.
:param data: TTSGML with <norm norm=...> and raw tokens to replace
:return: TTSGML with tags preserved and tokens replace by norm attribute values
"""
outdata = []
norm = ""
for line in data.replace("\r","").split("\n"):
if line.startswith("<"):
m = re.search(r'norm="([^"]*)"',line)
if m is not None:
norm = m.group(1)
outdata.append(line)
else:
if norm != "":
outdata.append(norm)
norm=""
return "\n".join(outdata) + "\n"
def read_attributes(input,attribute_name):
out_stream =""
for line in input.split('\n'):
if attribute_name + '="' in line:
m = re.search(attribute_name+r'="([^"]*)"',line)
if m is None:
print("ERR: cant find " + attribute_name + " in line: " + line)
attribute_value = ""
else:
attribute_value = m.group(1)
if len(attribute_value)==0:
attribute_value = "_warn:empty_"+attribute_name+"_"
out_stream += attribute_value +"\n"
return out_stream
def merge_into_tag(tag_to_kill, tag_to_merge_into,stream):
vals = []
cleaned_stream = ""
for line in stream.split("\n"):
if " "+tag_to_kill + "=" in line:
val = re.search(" " + tag_to_kill+'="([^"]*)"',line).group(1)
vals.append(val)
elif "</" + tag_to_kill + ">" in line:
pass
else:
cleaned_stream += line + "\n"
injected = inject(tag_to_kill,"\n".join(vals).strip(),tag_to_merge_into,cleaned_stream)
return injected
def exec_via_temp(input_text, command_params, workdir=""):
temp = tempfile.NamedTemporaryFile(delete=False)
exec_out = ""
try:
if PY3:
temp.write(input_text.encode("utf8"))
else:
temp.write(input_text)
temp.close()
command_params = [x if x != 'tempfilename' else temp.name for x in command_params]
if workdir == "":
proc = subprocess.Popen(command_params, stdout=subprocess.PIPE,stdin=subprocess.PIPE,stderr=subprocess.PIPE)
(stdout, stderr) = proc.communicate()
else:
proc = subprocess.Popen(command_params, stdout=subprocess.PIPE,stdin=subprocess.PIPE,stderr=subprocess.PIPE,cwd=workdir)
(stdout, stderr) = proc.communicate()
exec_out = stdout
except Exception as e:
print(e)
finally:
os.remove(temp.name)
if PY3:
exec_out = exec_out.decode("utf8")
return exec_out
def get_origs(data):
"""
Harvests orig from plain tokens (non-tag lines of TT SGML), grouped by norm spans
:param data: TT SGML with unnormalized tokens in non-tag lines and <norm.. tags indicating norm/orig borders
:return: string containing one reconstituted orig unit per line
"""
origs = []
current = ""
for line in data.split("\n"):
if "</norm>" in line:
origs.append(current)
current = ""
if not line.startswith("<"): # Token line
current += line
return "\n".join(origs)
def inject(attribute_name, contents, at_attribute,into_stream,replace=True):
"""
Inject new attributes and values into a specified SGML element in TT SGML format
:param attribute_name: name of the attribute to insert
:param contents: string, one value per line to insert as attribute values
:param at_attribute: attribute in target element to insert this attribute before (can be same attribute name, replace in place)
:param into_stream: TT SGML string with elements to add attributes to
:param replace: boolean, whether to replace existing values of attribute_name when already present
:return: TT SGML with added attributes
"""
insertions = contents.split('\n')
injected = ""
i=0
for line in into_stream.split("\n"):
if at_attribute + "=" in line:
if i >= len(insertions):
raise Exception("Error out of bounds at element " + str(i) + " in document beginning " + into_stream[:1000])
if len(insertions[i])>0:
if at_attribute == attribute_name: # Replace old value of attribute with new one
line = re.sub(attribute_name+'="[^"]*"',attribute_name+'="'+insertions[i]+'"',line)
else: # Place before specific at_attribute
if replace and ' ' + attribute_name + '=' in line:
line = re.sub(' ' + attribute_name + '="[^"]*"','',line) # Remove old value
line = re.sub(at_attribute+"=",attribute_name+'="'+insertions[i]+'" '+at_attribute+"=",line)
i += 1
injected += line + "\n"
return injected
def extract_conll(conll_string, mark_new_sent=True):
conll_string = conll_string.replace("\r","").strip()
sentences = conll_string.split("\n\n")
ids = ""
funcs = ""
parents = ""
id_counter = 0
offset = 0
new_sents = ""
for sentence in sentences:
tokens = sentence.split("\n")
for token in tokens:
if "\t" in token:
id_counter +=1
ids += "u"+ str(id_counter) + "\n"
cols = token.split("\t")
funcs += cols[7].replace("ROOT","root") +"\n"
if cols[6] == "0":
parents += "#u0\n"
else:
parents += "#u" + str(int(cols[6])+offset)+"\n"
if cols[0] == "1" and mark_new_sent:
new_sents += "true\n"
else:
new_sents += "false\n"
offset = id_counter
return ids, funcs, parents, new_sents
def parse2conllu(parser_output, tagged):
output = []
try:
tags = [l.split("\t")[1] for l in tagged.split("\n") if "\t" in l]
lemmas = [l.split("\t")[2] for l in tagged.split("\n") if "\t" in l]
except:
sys.stderr.write(tagged)
raise IOError("\nUnexpected number of tabs in tagged data line\n")
toknum = -1
for sent in parser_output.sentences:
tid = -1
for position in sorted(list(sent.annotations.keys())):
fields = sent.annotations[position]
if "\t" in fields:
toknum += 1
tid += 1
fields = fields.split("\t")
fields[3] = fields[4] = tags[toknum]
fields[2] = lemmas[toknum]
fields[6] = str(sent.values[6][tid])
fields[7] = sent.values[7][tid]
fields = "\t".join(fields)
output.append(fields)
output.append("")
return "\n".join(output)
def space_punct(input_text):
punct = set(["·",".","·","ⲵ",",",":",";","ʼ","„","“","{","}"])
if not PY3:
punct = set([unicode(p) for p in punct])
outstr = ""
textmode = True
for c in input_text:
if c == "<":
textmode = False
if c in punct and textmode:
outstr = "".join([outstr," " + c + " "])
else:
outstr = "".join([outstr,c])
if c == ">":
textmode = True
outstr = re.sub(" +", " ", outstr) # Kill double spaces
return outstr
def inject_with_nesting(in_sgml,insertion_specs,around_tag="norm",inserted_tag="entity"):
"""
Inject possibly nesting annotations around existing tags
:param in_sgml: input SGML stream including tags to surround with new tags
:param insertion_specs: list of triples (start, end, value), where start/end correspond to positions of around_tag
:param around_tag: tag of span to surround by insertion
:param inserted_tag: tag and attribute name to wrap inserted values in
:return: modified SGML stream
"""
lines = in_sgml.split("\n")
open_positions = [i for i in range(len(lines)) if lines[i].startswith("<" + around_tag+" ")]
for insertion in insertion_specs[::-1]: # Insert opening tags at desired indices in reverse
lines.insert(open_positions[insertion[0]],'<'+inserted_tag+' '+inserted_tag+'="' + insertion[2] + '">')
close_positions = [i for i in range(len(lines)) if lines[i] == "</" + around_tag + ">"]
insertion_specs.sort(key=lambda x:x[1]) # Sort by closing index
for insertion in insertion_specs[::-1]: # Insert opening tags at desired indices in reverse
lines.insert(close_positions[insertion[1]]+1,'</'+inserted_tag+'>')
return "\n".join(lines)
def get_entity_offsets(sgml):
lines = sgml.split("\n")
started = []
entities = []
toknum = 0
for line in lines:
if 'entity="' in line:
entity_type = re.search(r' entity="([^"]*)"',line).group(1)
started.append((toknum,entity_type))
elif '</referent>' in line:
start, entity_type = started.pop()
entities.append((start,toknum-1,entity_type))
elif not line.startswith("<") and not line.endswith(">") and len(line)>0: # Token
toknum += 1
return sorted(entities,key=lambda x: (x[0],-x[1]))
def analyze_entities(conll_parse,sgml_so_far,preloaded,outmode,do_identities=False,docname=None,dialect="sahidic",ot=False):
global identifier
if preloaded is None:
preloaded = {"stk":None,"xrenner":None,"parser":None,"tagger":None,"lemmatizer":None, "dialect":dialect}
if preloaded["xrenner"] is None:
from xrenner import Xrenner # lib.
if dialect == "bohairic":
xrenner = Xrenner(model=lib_dir + "boh.xrm")
else:
xrenner = Xrenner(model=lib_dir + "cop.xrm")
preloaded["xrenner"] = xrenner
else:
xrenner = preloaded["xrenner"]
if docname is None:
xrenner.docname = "_"
# Make sure we don't have stale entity annotations in input
sgml_so_far = re.sub(r'</?entity[^\n]+\n','',sgml_so_far)
if outmode == "sgml":
ents = xrenner.analyze(conll_parse, "sgml") # "conll_sent")
insertion_specs = get_entity_offsets(ents) # Get [(start, end, entity_type),...]
output = inject_with_nesting(sgml_so_far, insertion_specs, around_tag="norm", inserted_tag="entity")
# Ensure entity nests norm
output = reorder_sgml(output.strip(),priorities=["entity", "orig_group", "norm_group", "norm", "orig"])
output = assign_entity_heads(output)
if do_identities:
if identifier is None:
from lib.identify import Identifier
identifier = Identifier(old_testament=ot)
identifier.read_words(output)
output = identifier.predict_sgml(output, docname=docname)
return output
elif outmode == "conllu":
ents = xrenner.analyze(conll_parse, "conll_sent")
ents = ents.replace("\n\n", "\n").strip().split("\n")
ents = [line.split("\t")[-1] for line in ents if "\t" in line]
counter = 0
out_conll = []
for line in conll_parse.split("\n"):
if "\t" in line:
fields = line.split("\t")
fields[-1] = ents[counter]
line = "\t".join(fields)
counter += 1
out_conll.append(line)
return "\n".join(out_conll)
def postag(indata, full_sgml, tagger="flair",notokens=False,sent=None,tabular=False,postprocess=True,
preloaded=None,dialect="sahidic",with_sgml=False):
def get_bg_position(sgml):
lines = sgml.split("\n")
positions = []
open = 0
for line in lines:
if '</norm>' in line:
if open == 0:
positions.append("B")
else:
positions.append("I")
open += 1
if '</norm_group>' in line:
if open == 0:
positions[-1] = "S"
else:
positions[-1] = "E"
open = 0
return positions
bg_positions = get_bg_position(full_sgml)
if tagger=="treetagger":
if dialect == "bohairic":
tag = [tt_path + 'tree-tagger', tt_path + 'coptic.b_fine.par', '-lemma', '-no-unknown', '-sgml'] # no -token
else:
tag = [tt_path + 'tree-tagger', tt_path+'coptic_fine.par', '-lemma','-no-unknown', '-sgml'] #no -token
if not notokens:
tag += ['-token']
tag += ['tempfilename']
tagged = exec_via_temp(indata,tag)
if notokens:
return tagged
elif tagger=="marmot":
tagged = tag_marmot(indata, sent=sent, dialect=dialect)
tagged = lemmatizer.lemmatize(tagged, bg_positions)
elif tagger=="flair":
if preloaded["tagger"] is None:
with suppress_stdout_stderr():
preloaded["tagger"] = FlairTagger()
sentstr = "translation" if sent is None else sent
tagged = preloaded["tagger"].predict(full_sgml, in_format="sgml", out_format="tt", sent=sentstr, as_text=True)
tagged = lemmatizer.lemmatize(tagged, bg_positions)
spl = [line.split("\t") for line in tagged.strip().split("\n")]
words, tags, lemmas = zip(*spl)
if postprocess: # Replace implausible word+tag combinations
tags = list(tags)
tagtab = io.open(data_dir+"postprocess_tagger.tab", encoding="utf8").read().replace("\r", "").strip().split("\n")
mapping = dict(((line.split("\t")[0], line.split("\t")[1]), line.split("\t")[2]) for line in tagtab)
for i, word in enumerate(words):
tag = tags[i]
if (word, tag) in mapping:
tags[i] = mapping[(word, tag)]
elif (word, "*") in mapping:
tags[i] = mapping[(word, "*")]
if tabular and not notokens:
if with_sgml:
output = []
toknum = 0
for line in indata.strip().split("\n"):
if not (line.startswith("<") and line.endswith(">")):
output.append(words[toknum] + "\t" + tags[toknum] + "\t" + lemmas[toknum])
toknum += 1
else:
output.append(line)
return "\n".join(output)
else:
return "\n".join([words[i]+"\t"+tags[i]+"\t"+lemmas[i] for i in range(len(tags))])
elif notokens:
return "\n".join([tags[i]+"\t"+lemmas[i] for i in range(len(tags))])
tagged = inject("pos","\n".join(tags),"norm",indata)
tagged = inject("lemma","\n".join(lemmas),"norm",tagged)
tagged = re.sub('\r','',tagged)
return tagged
def predict_sents(sgml):
from lib.flair_sent_splitter import FlairSentSplitter
with suppress_stdout_stderr():
s = FlairSentSplitter(train=False)
preds = s.predict(sgml,as_text=True,as_bound_groups=True)
output = []
bg_counter = 0
last_break = -1
first_sent = True
trans_open = False
# place translation on outermost group type
elem = "orig_group" if "</norm_group>\n</orig_group>" in sgml else "norm_group"
lines = sgml.strip().split("\n")
for l, line in enumerate(lines):
if "</chapter" in line:
preds[bg_counter] = 1
if elem + " " in line or "</chapter" in line:
if preds[bg_counter] == 1 and last_break < bg_counter:
if not first_sent:
output.append('</translation>')
trans_open = False
if any([len(l) > 0 and not l.startswith("<") for l in lines[l:]]): # Check that we still have tokens coming
output.append('<translation translation="...">')
trans_open = True
first_sent = False
last_break = bg_counter
if '</translation>' in line:
trans_open = False
output.append(line)
if "</"+elem+">" in line:
bg_counter += 1
if trans_open:
output.append("</translation>")
return "\n".join(output)
def predict_sents_xgboost(sgml,genre="other",force_break_on_chapter=True):
from lib.sentencer import BGSentencer
s = BGSentencer()
preds = s.predict(sgml,as_text=True,genre=genre)
preds[0] = 1 # Force first group to be a new sentence
output = []
bg_counter = 0
last_break = -1
first_sent = True
# place translation on outermost group type
elem = "orig_group" if "</norm_group>\n</orig_group>" in sgml else "norm_group"
for line in sgml.strip().split("\n"):
if elem + " " in line or "</chapter" in line:
if preds[bg_counter] == 1 and last_break < bg_counter:
if not first_sent:
output.append('</translation>')
output.append('<translation translation="...">')
first_sent = False
last_break = bg_counter
output.append(line)
if "</"+elem+">" in line:
bg_counter += 1
output.append("</translation>")
return "\n".join(output)
def check_requirements(require_tt=False, require_malt=False, require_marmot=False):
marmot_OK = True
tt_OK = True
malt_OK = True
foma_OK = True
tt = "tree-tagger"
if platform.system() == "Windows":
tt+=".exe"
if not os.path.exists(tt_path + tt) and require_tt:
sys.stderr.write("! TreeTagger not found at ./bin/\n")
tt_OK = False
if not os.path.exists(parser_path+"maltparser-1.8.jar") and require_malt:
sys.stderr.write("! Malt Parser 1.8 not found at ./bin/\n")
malt_OK = False
if not os.path.exists(bin_dir+"marmot"+os.sep+"marmot.jar") and require_marmot:
sys.stderr.write("! Marmot not found at ./bin/marmot/\n")
marmot_OK = False
if not (os.path.exists(bin_dir + "foma" + os.sep + "flookup") or os.path.exists(bin_dir + "foma" + os.sep + "flookup.exe")):
sys.stderr.write("! Foma flookup not found at ./bin/foma/\n")
foma_OK = False
return tt_OK, malt_OK, foma_OK, marmot_OK
def download_requirements(tt_ok=True, malt_ok=True, foma_ok=True, marmot_ok=True, require_tt=False, require_malt=False):
import requests, zipfile, shutil, tarfile
if not PY3:
import StringIO
urls = []
if not foma_ok:
if platform.system() == "Windows":
with zipfile.ZipFile(bin_dir+"foma"+os.sep+"foma_win.zip", 'r') as z:
z.extractall(bin_dir+"foma"+os.sep)
elif platform.system() == "Darwin":
with zipfile.ZipFile(bin_dir+"foma"+os.sep+"foma_osx.zip", 'r') as z:
z.extractall(bin_dir+"foma"+os.sep)
else: # Linux
sys.stderr.write("! Need to compile foma on Linux and place flookup in bin/foma/\n! See bin/foma/README.md \n")
sys.exit(0)
if not marmot_ok:
if not os.path.exists(bin_dir + "marmot"):
os.makedirs(bin_dir + "marmot")
marmot_base_url = "http://cistern.cis.lmu.de/marmot/bin/CURRENT/"
marmot_current = requests.get(marmot_base_url).text
files = re.findall(r'href="((?:marmot|trove)[^"]+jar)"',marmot_current)
marmot_file = ""
trove_file = ""
for f in files:
if f.startswith("marmot"):
marmot_file = f
elif f.startswith("trove"):
trove_file = f
urls.append(marmot_base_url + marmot_file)
urls.append(marmot_base_url + trove_file)
if not malt_ok and require_malt:
urls.append("http://maltparser.org/dist/maltparser-1.8.tar.gz")
if not tt_ok and require_tt:
if platform.system() == "Windows":
u = "http://www.cis.uni-muenchen.de/~schmid/tools/TreeTagger/data/tree-tagger-windows-3.2.1.zip"
elif platform.system() == "Darwin":
u = "http://www.cis.uni-muenchen.de/~schmid/tools/TreeTagger/data/tree-tagger-MacOSX-3.2.tar.gz"
else:
if "Red Hat Enterprise" in platform.linux_distribution()[0] and platform.linux_distribution()[1].startswith("6."):
# Use older kernel version of TreeTagger for RHEL 6
u = "http://www.cis.uni-muenchen.de/~schmid/tools/TreeTagger/data/tree-tagger-linux-3.2-old5.tar.gz"
else:
u = "http://www.cis.uni-muenchen.de/~schmid/tools/TreeTagger/data/tree-tagger-linux-3.2.1.tar.gz"
if platform.system() != "Windows": # Create TreeTagger subdirectory
os.makedirs(bin_dir + "TreeTagger")
urls.append(u)
for u in urls:
sys.stderr.write("o Downloading from " + u + "\n")
r = requests.get(u, stream=True)
if PY3:
file_contents = io.BytesIO(r.content)
else:
file_contents = StringIO.StringIO(r.content)
if u.endswith("zip"):
z = zipfile.ZipFile(file_contents)
elif u.endswith("jar"):
if "trove" in u:
with open(bin_dir + "marmot" + os.sep + "trove.jar", 'wb') as f:
f.write(r.content)
elif "marmot" in u:
with open(bin_dir + "marmot" + os.sep + "marmot.jar", 'wb') as f:
f.write(r.content)
continue
else:
z = tarfile.open(fileobj=file_contents, mode="r:gz")
os_suf = ""
if "tree" in u and platform.system() != "Windows":
os_suf = "TreeTagger" + os.sep
z.extractall(path=bin_dir + os_suf)
if not malt_OK and require_malt:
shutil.copyfile(bin_dir+"coptic.mco",bin_dir+"maltparser-1.8" + os.sep + "coptic.mco")
if not tt_ok and require_tt:
shutil.copyfile(bin_dir+"coptic_fine.par",bin_dir+"TreeTagger" + os.sep + "bin" + os.sep + "coptic_fine.par")
def nlp_coptic(input_data, lb=False, parse_only=False, do_tok=True, do_norm=True, do_mwe=True, do_tag=True, do_lemma=True, do_lang=True,
do_milestone=True, do_parse=True, sgml_mode="sgml", tok_mode="auto", old_tokenizer=False, sent_tag=None,
preloaded=None, pos_spans=False, merge_parse=False, detokenize=0, segment_merged=False, gold_parse="",
tagger="flair", parser="diaparser", do_entities=False, no_gold_parse=False, mark_new_sent=True, do_identities=False,
docname=None,dialect="auto",ot=False):
global data_dir
global lemmatizer
global neural_model
if dialect.lower().startswith("auto"):
dialect = detect_lang(input_data)
if dialect == "bohairic":
data_dir = data_dir.replace("data" + os.sep, "data.b" + os.sep)
neural_model = neural_model.replace("cop.diaparser", "boh.diaparser")
tagger = "treetagger" if not os.path.exists(lib_dir + "boh.marmot") else "marmot"
if detokenize > 1:
detokenize = 1
do_mwe = False
if preloaded is None:
with suppress_stdout_stderr():
preloaded = {"stk":None,"xrenner":None,"parser":NeuralParser.load(neural_model),"tagger":FlairTagger(),
"lemmatizer":Lemmatizer(data_dir + "copt_lemma_lex.tab", no_unknown=True, dialect=dialect),
"dialect":dialect}
elif preloaded["lemmatizer"] is None and do_lemma:
preloaded["lemmatizer"] = Lemmatizer(data_dir + "copt_lemma_lex.tab", no_unknown=True, dialect=dialect)
lemmatizer = preloaded["lemmatizer"]
if docname is not None:
docname = docname.replace(".tt","").replace(".sgml","").replace(".xml","").replace(".txt","")
with suppress_stdout_stderr():
if do_parse or parse_only or merge_parse:
neural_parser = preloaded["parser"] if preloaded["parser"] is not None else NeuralParser.load(neural_model)
data = input_data.replace("\t","")
data = data.replace("\r","")
if preloaded["stk"] is not None:
stk = preloaded["stk"]
elif do_tok:
stk = StackedTokenizer(pipes=sgml_mode != "sgml", lines=lb, tokenized=tok_mode=="from_pipes",
detok=detokenize, segment_merged=segment_merged, ambig=data_dir + "ambig.tab",
dialect=dialect)
if preloaded["tagger"] is None and tagger=="flair" and do_tag:
with suppress_stdout_stderr():
preloaded["tagger"] = FlairTagger(train=False)
if do_milestone:
data = binarize(data)
if do_entities and not do_parse and not parse_only and not merge_parse:
do_parse = True
if do_tok:
if old_tokenizer:
tokenize = ['perl', lib_dir + 'tokenize_coptic.pl', '-n']
if lb:
tokenize.append('-l')
if sgml_mode == "pipes":
tokenize.append('-p')
if tok_mode == "from_pipes":
tokenize.append('-t')
tokenize += ['-d', data_dir + 'copt_lemma_lex.tab', '-s', data_dir + 'segmentation_table.tab', '-m', data_dir + 'morph_table.tab', 'tempfilename']
tokenized = exec_via_temp(data,tokenize)
tokenized = tokenized.replace('\r','').strip()
tokenized = re.sub(r'_$','',tokenized)
else:
tokenized = stk.analyze(data)
if not lb and sgml_mode == "pipes":
tokenized = tokenized.replace("\n","")
if sgml_mode == "pipes":
return tokenized
else:
tokenized = data
if sgml_mode == "sgml" and "norm=" not in tokenized:
# Assume raw one token per line, wrap everything in norm tags
tok_lines = []
for line in tokenized.split("\n"):
if not line.startswith("<"): # Leave XML tags alone
line = '<norm_group norm_group="' + line + '">\n<norm norm="'+ line +'">\n' + line + '\n</norm>\n</norm_group>'
tok_lines.append(line)
tokenized = "\n".join(tok_lines)
tokenized = tokenized.replace('\r','').strip()
output = tokenized
norms = read_attributes(tokenized,"norm")
if do_norm:
from lib.auto_norm import normalize
if dialect == "bohairic":
norms = normalize(norms, table_file=data_dir + "norm_table.tab", method="none")
else:
norms = normalize(norms,table_file=data_dir + "norm_table.tab")
output = inject("norm", norms, "norm", output)
if parse_only or merge_parse:
if not do_tag and (parse_only or merge_parse) and not pos_spans:
if "\t" not in input_data and 'pos="' not in input_data:
sys.stderr.write("! You selected parsing without tagging (-t) and your data format appears to contain no POS tag column.\n")
resp = inp("! Would you like to add POS tagging to the job profile? [Y]es/[N]o/[A]bort ")
if resp.lower() == "y":
do_tag = True
elif resp.lower() == "a":
sys.exit(0)
if do_tag and not pos_spans:
tagged = postag(norms, output, tagger=tagger,sent=sent_tag, preloaded=preloaded, dialect=dialect)
else: # Assume data is already tagged, in TT SGML format
if pos_spans:
tagged = harvest_tt(input_data, keep_sgml=True)
else:
tagged = input_data
if PY3:
tagged = input_data.encode("utf8") # Handle non-UTF-8 when calling TT from subprocess in Python 3
if gold_parse == "":
# NB if element is present for conllize it supercedes the POS tag for sentence splitting
if parser != "malt":
conllized = conllize(tagged, tag="PUNCT", element=sent_tag, no_zero=True, ten_cols=True)
for_parser = []
for s in conllized.strip().split("\n\n"):
for_parser.append([l.split("\t")[1] for l in s.split("\n") if "\t" in l])
with suppress_stdout_stderr():
parsed = neural_parser.predict(for_parser)
parsed = parse2conllu(parsed,tagged)
else:
conllized = conllize(tagged, tag="PUNCT", element=sent_tag, no_zero=True)
deped = DepEdit(io.open(data_dir + "add_ud_and_flat_morph.ini",encoding="utf8"),options=type('', (), {"quiet":True, "kill":None})())
depedited = deped.run_depedit(conllized.split("\n"))
#depedited = conllized
parse_coptic = ['java','-mx512m','-jar',"maltparser-1.8.jar",'-c','coptic','-i','tempfilename','-m','parse']
if not os.path.exists(bin_dir+"maltparser-1.8" + os.sep + "coptic.mco"):
sys.stderr.write("! can't find coptic.mco parser model in " + bin_dir+"maltparser-1.8" + os.sep + "coptic.mco")
sys.exit(0)
parsed = exec_via_temp(depedited,parse_coptic,parser_path)
deped = DepEdit(io.open(data_dir + "postprocess_parser.ini",encoding="utf8"),options=type('', (), {"quiet":True, "kill":None})())
depedited = deped.run_depedit(parsed.split("\n"))
if len(gold_trees) == 0 and not no_gold_parse:
get_gold_trees()
if len(gold_trees) > 0:
# Replace automatic parses with cached trees from UD_Coptic if available
depedited = replace_trees(depedited, gold_trees)
else: # A cached gold parse has been specified by the user
depedited = gold_parse
norm_count = len(re.findall(r'(\n|^)[0-9]+\t',depedited))
input_norms = input_data.count(" norm=")
if norm_count != input_norms:
dped = re.findall(r'(?:\n|^)[0-9]+\t([^\n\t]+)',depedited)
nrms = re.findall(r' norm="([^"]*)"',input_data)
mismatch = ""
for i, nrm in enumerate(nrms):
if nrm != dped[i]:
mismatch = nrm +"!="+dped[i] + " at " + str(i) + " after " + nrms[i-4] + " " + nrms[i-3] + " " + nrms[i-2] + " " + nrms[i-1]
break
raise IOError("Mismatch in word count: " + str(norm_count) + " in gold parse but " + str(input_norms) + " in SGML file\nMismatch: "+ mismatch)
if do_entities:
if do_identities:
pass
#sys.stderr.write("! ignoring Wikification task due to parser merge workflow")
if " func=" not in output: # SGML has no merged syntax
ids, funcs, parents, new_sents = extract_conll(depedited.strip(), mark_new_sent=mark_new_sent)
output = inject("xml:id", ids, "norm", output)
output = inject("func", funcs, "norm", output)
output = inject("head", parents, "norm", output)
output = inject("new_sent", new_sents, "norm", output)
output = output.replace(' head="#u0"', "").replace(' new_sent="false"', "")
output = merge_into_tag("pos", "norm", output)
output = merge_into_tag("lemma", "norm", output).strip()
ents = analyze_entities(depedited, output, preloaded, sgml_mode, do_identities=do_identities, docname=docname, dialect=dialect, ot=ot)
if sgml_mode == "conllu":
depedited = ents
else:
output = ents
else:
output = input_data
if parse_only: # Output parse in conll format
return depedited
elif merge_parse: # Insert parse into input SGML as attributes of <norm>
if "norm=" not in input_data:
sys.stderr.write('ERR: --merge_parse was selected but no <norm norm=".."> tags found in input\n')
sys.exit(0)
if sgml_mode == "conllu":
return depedited
if " func=" not in output:
ids, funcs, parents, new_sents = extract_conll(depedited.strip(), mark_new_sent=mark_new_sent)
output = inject("xml:id", ids, "norm", output)
output = inject("func", funcs, "norm", output)
output = inject("head", parents, "norm", output)
output = inject("new_sent", new_sents, "norm", output)
output = output.replace(' head="#u0"', "").replace(' new_sent="false"', "")
if '<norm pos=' not in output: # Move pos and lemma tags into norm elements, unless already there
output = merge_into_tag("pos", "norm", output)
output = merge_into_tag("lemma", "norm", output)
return output
elif sent_tag == "predict" and not do_parse and not do_tok and "pos=" in tokenized:
#sys.stderr.write("o Adding predicted sentence spans to input SGML\n")
output = predict_sents(tokenized)
return output
elif not do_parse:
tagged = postag(norms,output,tagger=tagger,sent=sent_tag,notokens=True, preloaded=preloaded, dialect=dialect).replace("\r","")
if do_parse:
if sent_tag is None and not pos_spans:
tagged = postag(norms,output,tagger=tagger,sent=sent_tag,tabular=True, preloaded=preloaded, dialect=dialect)
else:
norm_with_sgml = tok_from_norm(output)
tagged = postag(norm_with_sgml,output,tagger=tagger,sent=sent_tag,tabular=True, preloaded=preloaded, dialect=dialect, with_sgml=True)
if parser!="malt":
conllized = conllize(tagged, tag="PUNCT", element=sent_tag, no_zero=True, ten_cols=True)
for_parser = []
for s in conllized.strip().split("\n\n"):
for_parser.append([l.split("\t")[1] for l in s.split("\n") if "\t" in l])
with suppress_stdout_stderr():
parsed = neural_parser.predict(for_parser)
parsed = parse2conllu(parsed,tagged)
else:
conllized = conllize(tagged, tag="PUNCT", element=sent_tag, no_zero=True)
if not os.path.exists(bin_dir + "maltparser-1.8" + os.sep + "coptic.mco"):
sys.stderr.write(
"! can't find coptic.mco parser model in " + bin_dir + "maltparser-1.8" + os.sep + "coptic.mco")
sys.exit(0)
deped = DepEdit(io.open(data_dir + "add_ud_and_flat_morph.ini", encoding="utf8"),
options=type('', (), {"quiet": True, "kill": "supertoks"})())
depedited = deped.run_depedit(conllized.split("\n"))
# depedited = conllized
parse_coptic = ['java','-mx1g','-jar',"maltparser-1.8.jar",'-c','coptic','-i','tempfilename','-m','parse']
parsed = exec_via_temp(depedited,parse_coptic,parser_path)
deped = DepEdit(io.open(data_dir + "postprocess_parser.ini",encoding="utf8"),options=type('', (), {"quiet":True,"kill":"supertoks"})())
depedited = deped.run_depedit(parsed.split("\n"))
if len(gold_trees) > 0:
# Replace automatic parses with cached trees from UD_Coptic if available
depedited = replace_trees(depedited, gold_trees)
if sgml_mode == "conllu" and not do_entities: # Return conllu parse and finish