forked from ufal/udpipe
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathMANUAL
2647 lines (1961 loc) · 114 KB
/
MANUAL
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
UDPipe
Version 1.2.1-devel
Introduction
============
UDPipe is a trainable pipeline for tokenization, tagging, lemmatization and
dependency parsing of CoNLL-U files. UDPipe is language-agnostic and can be
trained given annotated data in CoNLL-U format
(http://universaldependencies.org/format.html). Trained models are provided for
nearly all UD treebanks (http://universaldependencies.org). UDPipe is available
as a binary for Linux/Windows/OS X, as a library for C++, Python, Perl, Java,
C#, and as a web service. Third-party R CRAN package
(https://CRAN.R-project.org/package=udpipe) also exists.
UDPipe is a free software distributed under the Mozilla Public License 2.0
(http://www.mozilla.org/MPL/2.0/) and the linguistic models are free for
non-commercial use and distributed under the CC BY-NC-SA
(http://creativecommons.org/licenses/by-nc-sa/4.0/) license, although for some
models the original data used to create the model may impose additional
licensing conditions. UDPipe is versioned using Semantic Versioning
(http://semver.org/).
Copyright 2017 by Institute of Formal and Applied Linguistics, Faculty of
Mathematics and Physics, Charles University, Czech Republic.
Online Web Application and Web Service
======================================
UDPipe Web Application is available at
http://lindat.mff.cuni.cz/services/udpipe/ using LINDAT/CLARIN infrastructure
(http://lindat.cz).
UDPipe REST Web Service is also available, with the API documentation available
at http://lindat.mff.cuni.cz/services/udpipe/api-reference.php.
Release
=======
Download
--------
UDPipe releases are available on GitHub (http://github.com/ufal/udpipe), both as
source code and as a pre-compiled binary package. The binary package contains
Linux, Windows and OS X binaries, Java bindings binary, C# bindings binary, and
source code of UDPipe and all language bindings). While the binary packages do
not contain compiled Python or Perl bindings, packages for those languages are
available in standard package repositories, i.e. on PyPI
(https://pypi.python.org/pypi/ufal.udpipe/) and CPAN
(https://metacpan.org/pod/Ufal::UDPipe).
- Latest release (http://github.com/ufal/udpipe/releases/latest)
- All releases (http://github.com/ufal/udpipe/releases), Changelog
(https://github.com/ufal/udpipe/blob/master/CHANGES)
Language Models
---------------
To use UDPipe, a language model is needed. The language models are available
from LINDAT/CLARIN (http://www.lindat.cz) infrastructure and described further
in the UDPipe User's Manual. Currently, the following language models are
available:
- Universal Dependencies 2.0 Models: udpipe-ud2.0-170801
(http://hdl.handle.net/11234/1-2364) (documentation
(http://ufal.mff.cuni.cz/udpipe/users-manual#universal_dependencies_20_models))
- CoNLL17 Shared Task Baseline UD 2.0 Models: udpipe-ud2.0-conll17-170315
(http://hdl.handle.net/11234/1-1990) (documentation
(http://ufal.mff.cuni.cz/udpipe/users-manual#conll17_shared_task_baseline_ud_20_models))
- Universal Dependencies 1.2 Models: udpipe-ud1.2-160523
(http://hdl.handle.net/11234/1-1659) (documentation
(http://ufal.mff.cuni.cz/udpipe/users-manual#universal_dependencies_12_models))
License
-------
UDPipe is an open-source project and is freely available for non-commercial
purposes. The library is distributed under Mozilla Public License 2.0
(http://www.mozilla.org/MPL/2.0/) and the associated models and data under CC
BY-NC-SA (http://creativecommons.org/licenses/by-nc-sa/4.0/), although for some
models the original data used to create the model may impose additional
licensing conditions.
If you use this tool for scientific work, please give credit to us by
referencing Straka et al. 2016 and the UDPipe website
(http://ufal.mff.cuni.cz/udpipe).
UDPipe Installation
===================
UDPipe releases are available on GitHub (http://github.com/ufal/udpipe), either
as a pre-compiled binary package, or source code only. The binary package
contains Linux, Windows and OS X binaries, Java bindings binary, C# bindings
binary, and source code of UDPipe and all language bindings. While the binary
packages do not contain compiled Python or Perl bindings, packages for those
languages are available in standard package repositories, i.e. on PyPI and CPAN.
To use UDPipe, a language model is needed. Here is a list of available language
models (http://ufal.mff.cuni.cz/udpipe#language_models).
If you want to compile UDPipe manually, sources are available on on GitHub
(http://github.com/ufal/udpipe), both in the pre-compiled binary package
releases (http://github.com/ufal/udpipe/releases) and in the repository itself.
Requirements
------------
- g++ 4.7 or newer, clang 3.2 or newer, Visual C++ 2015 or newer
- make
- SWIG 3.0.8 or newer for language bindings other than C++
Compilation
-----------
To compile UDPipe, run make in the src directory.
Make targets and options:
- exe: compile the binaries (default)
- server: compile the REST server
- lib: compile the static library
- BITS=32 or BITS=64: compile for specified 32-bit or 64-bit architecture
instead of the default one
- MODE=release: create release build which statically links the C++ runtime and
uses LTO
- MODE=debug: create debug build
- MODE=profile: create profile build
Platforms
---------
Platform can be selected using one of the following options:
- PLATFORM=linux, PLATFORM=linux-gcc: gcc compiler on Linux operating system,
default on Linux
- PLATFORM=linux-clang: clang compiler on Linux, must be selected manually
- PLATFORM=osx, PLATFORM=osx-clang: clang compiler on OS X, default on OS X;
BITS=32+64 enables multiarch build
- PLATFORM=win, PLATFORM=win-gcc: gcc compiler on Windows (TDM-GCC is well
tested), default on Windows
- PLATFORM=win-vs: Visual C++ 2015 compiler on Windows, must be selected
manually; note that the cl.exe compiler must be already present in PATH and
corresponding BITS=32 or BITS=64 must be specified
Either POSIX shell or Windows CMD can be used as shell, it is detected
automatically.
Further Details
---------------
UDPipe uses C++ BuilTem system (http://github.com/ufal/cpp_builtem), please
refer to its manual if interested in all supported options.
Other language bindings
-----------------------
C#
--
Binary C# bindings are available in UDPipe binary packages.
To compile C# bindings manually, run make in the bindings/csharp directory,
optionally with the options described in UDPipe Installation.
Java
----
Binary Java bindings are available in UDPipe binary packages.
To compile Java bindings manually, run make in the bindings/java directory,
optionally with the options described in UDPipe Installation. Java 6 and newer
is supported.
The Java installation specified in the environment variable JAVA_HOME is used.
If the environment variable does not exist, the JAVA_HOME can be specified using
make JAVA_HOME=path_to_Java_installation
Perl
----
The Perl bindings are available as Ufal-UDPipe package on CPAN.
To compile Perl bindings manually, run make in the bindings/perl directory,
optionally with the options described in UDPipe Installation. Perl 5.10 and
later is supported.
Path to the include headers of the required Perl version must be specified in
the PERL_INCLUDE variable using
make PERL_INCLUDE=path_to_Perl_includes
Python
------
The Python bindings are available as ufal.udpipe package on PyPI.
To compile Python bindings manually, run make in the bindings/python directory,
optionally with options described in UDPipe Installation. Both Python 2.6+ and
Python 3+ are supported.
Path to the include headers of the required Python version must be specified in
the PYTHON_INCLUDE variable using
make PYTHON_INCLUDE=path_to_Python_includes
UDPipe User's Manual
====================
Like any supervised machine-learning tool, UDPipe needs a trained linguistic
model. This section describes the available language models and also the command
line tools and interfaces.
Running UDPipe
--------------
Probably the most common usage of UDPipe is to tokenize, tag and parse the input
using
udpipe --tokenize --tag --parse udpipe_model
The input is assumed to be in UTF-8 encoding and can be either already tokenized
and segmented, or it can be a plain text which will be tokenized and segmented
automatically.
Any number of input files can be specified after the udpipe_model and if no file
is given, the standard input is used. The output is by default saved to the
standard output, but if --outfile=name is used, it is saved to the given file
name. The output file name can contain a {}, which is replaced by a base name of
the processed file (i.e., without directories and an extension).
The full command syntax of running UDPipe is
Usage: udpipe [running_opts] udpipe_model [input_files]
udpipe --train [training_opts] udpipe_model [input_files]
udpipe --detokenize [detokenize_opts] raw_text_file [input_files]
Running opts: --accuracy (measure accuracy only)
--input=[conllu|generic_tokenizer|horizontal|vertical]
--immediate (process sentences immediately during loading)
--outfile=output file template
--output=[conllu|matxin|horizontal|plaintext|vertical]
--tokenize (perform tokenization)
--tokenizer=tokenizer options, implies --tokenize
--tag (perform tagging)
--tagger=tagger options, implies --tag
--parse (perform parsing)
--parser=parser options, implies --parse
Training opts: --method=[morphodita_parsito] which method to use
--heldout=heldout data file name
--tokenizer=tokenizer options
--tagger=tagger options
--parser=parser options
Detokenize opts: --outfile=output file template
Generic opts: --version
--help
Immediate Mode
--------------
By default UDPipe loads the whole input file into memory before starting to
process it. That allows to store the space markup (see the following Tokenizer
section) in most consistent way, i.e., store all spaces following a sentence in
the last token of that sentence.
However, sometimes it is desirable to process the input as soon as possible,
which can be achieved by specifying the --immediate option. In immediate mode,
the input is processed and printed as soon as a block of input guaranteed to
contain whole sentences is loaded. Specifically, for most input formats the
input is processed after loading an empty line (with the exception of horizontal
input format and presegmented tokenizer, where the input is processed after each
line).
Loading Model On Demand
-----------------------
Although a model for UDPipe always has to be specified, the model is loaded only
if really needed. It is therefore possible to use for example none as the model
in case it is not required for performing the requested operation (e.g.,
converting between formats or using a generic tokenizer).
Tokenizer
---------
If the --tokenize option is supplied, the input is assumed to be plain text and
is tokenized using model tokenizer. Additional arguments to the tokenizer might
be specified using --tokenizer=data option (which implies --tokenize), where
data is a semicolon-separated list of the following options:
- normalized_spaces: by default, UDPipe uses custom MISC fields to exactly
encode spaces in the original document (as described below). If the
normalized_spaces option is given, only the standard CoNLL-U v2 markup
(SpaceAfter=No and # newpar) is used.
- presegmented: the input file is assumed to be already segmented, with each
sentence on a separate line, and is only tokenized (respecting sentence
breaks)
- ranges: for each token, a range in the original document is stored in the
format described below.
- joint_with_parsing: an experimental mode performing sentence segmentation
jointly using the tokenizer and the parser (see Milan Straka and Jana
Straková: Tokenizing, POS Tagging, Lemmatizing and Parsing UD 2.0 with UDPipe
(http://ufal.mff.cuni.cz/~straka/papers/2017-conll_udpipe.pdf) paper for
details). The following options are utilized:
- joint_max_sentence_len (default 20): maximum sentence length
- joint_change_boundary_logprob (default -0.5): logprob of using sentence
boundary not generated by the tokenizer
- joint_sentence_logprob (default -0.5): additional logprob of every sentence
The logprob of a sentence is computed using logprob of its best dependency
parsing tree, together with joint_sentence_logprob and also
joint_change_boundary_logprob for every sentence boundary not returned by the
tokenizer (i.e., either 0, 1 or 2 times). The joint sentence segmentation
chooses such a segmentation, where every sentence has length at most
joint_max_sentence_len and the sum of logprobs of all sentences is as large as
possible.
Preserving Original Spaces
--------------------------
By default, UDPipe uses custom MISC fields to store all spaces in the original
document. This markup is backward compatible with CoNLL-U v2 SpaceAfter=No
feature. This markup can be utilized by the plaintext output format, which
allows reconstructing the original document.
Note that in theory not only spaces, but also other original content can be
saved in this way (for example XML tags if the input was encoded in a XML file).
The markup uses the following MISC fields on tokens (not words in multi-word
tokens):
- SpacesBefore=content (by default empty): spaces/other content preceding the
token
- SpacesAfter=content (by default a space if SpaceAfter=No feature is not
present, empty otherwise): spaces/other content following the token
- SpacesInToken=content (by default equal to the FORM of the token): FORM of
the token including original spaces (this is needed only if tokens are allowed
to contain spaces and a token contains a tab or newline characters)
The content of all the three fields must be escaped to allow storing tabs and
newlines. The following C-like schema is used:
- \s: space
- \t: tab
- \r: CR character
- \n: LF character
- \p: | (pipe character)
- \\: \ (backslash character)
Preserving Token Ranges
-----------------------
When the ranges tokenizer option is used, the range of each token in the
original document is stored in the TokenRange MISC field.
The format of the TokenRange field (inspired by Python) is TokenRange=start:end,
where start is a zero-based document-level index of the start of the token
(counted in Unicode characters) and end is a zero-based document-level index of
the first character following the token (i.e., the length of the token is
end-start).
Input Formats
-------------
If the tokenizer is not used, the input format can be specified using the
--input option. The individual input formats can be parametrized in the same way
a tokenizer is, by using format=data syntax. Currently supported input formats
are:
- conllu (default): the CoNLL-U format
(http://universaldependencies.org/docs/format.html). Supported options:
- v2 (default): use CoNLL-U v2
- v1: allow loading only CoNLL-U v1 (i.e., no empty nodes and no spaces in
forms and lemmas)
- generic_tokenizer: generic tokenizer for English-like languages (with spaces
separating tokens and English-like punctuation). The tokenizer is rule-based
and needs no trained model. It supports the same options as a model tokenizer,
i.e., normalized_spaces, presegmented and ranges.
- horizontal: each sentence on a separate line, with tokens separated by
spaces. In order to allow spaces in tokens, Unicode character 'NO-BREAK SPACE'
(U+00A0) is considered part of token and converted to a space during loading.
- vertical: each token on a separate line, with an empty line denoting end of
sentence; only the first tab-separated word is used as a token, the rest of
the line is ignored.
Note that a model tokenizer can be specified using the --input option too, by
using the tokenizer input format, for example using --input tokenizer=ranges.
Tagger
------
If the --tag option is supplied, the input is POS tagged and lemmatized using
the model tagger. Additional arguments to the tagger might be specified using
the --tagger=data option (which implies --tag).
Dependency Parsing
------------------
If the --parse option is supplied, the input is parsed using the model
dependency parser. Additional arguments to the parser might be specified using
the --parser=data option (which implies --parse).
Output Formats
--------------
The output format is specified using the --output option. The individual output
formats can be parametrized in the same way as input formats, by using the
format=data syntax. Currently supported output formats are:
- conllu (default): the CoNLL-U format
(http://universaldependencies.org/docs/format.html) Supported options:
- v2 (default): use CoNLL-U v2
- v1: produce output in CoNLL-U v1 format. Note that this is a lossy process,
as empty nodes are ignored and spaces in forms and lemmas are converted to
underscores.
- matxin: the Matxin format (http://wiki.apertium.org/wiki/Matxin)
- horizontal: writes the words (in the UD sense) in horizontal format, that is,
each sentence is on a separate line, with words separated by a single space.
Because words can contain spaces in CoNLL-U v2, the spaces in words are
converted to Unicode character 'NO-BREAK SPACE' (U+00A0). Supported options:
- paragraphs: an empty line is printed after the end of a paragraph or a
document (recognized by # newpar or # newdoc comments)
- plaintext: writes the tokens (in the UD sense) using original spacing. By
default, UDPipe's custom MISC features (SpacesBefore, SpacesAfter and
SpacesInToken, see the description in the Tokenizer section) are used to
reconstruct the exact original spaces. However, if the document does not
contain these features or if you want only normalized spacing, you can use the
following option:
- normalized_spaces: write one sentence on a line, and either one or no space
between tokens according to the SpaceAfter=No feature
- vertical: each word on a separate line, with an empty line denoting the end
of sentence. Supported options:
- paragraphs: an empty line is printed after the end of a paragraph or a
document (recognized by # newpar or # newdoc comments)
Running the UDPipe REST Server
------------------------------
UDPipe also provides a REST server binary called udpipe_server. The binary uses
MicroRestD (http://github.com/ufal/microrestd) as a REST server implementation
and provides UDPipe REST API
(http://lindat.mff.cuni.cz/services/udpipe/api-reference.php).
The full command syntax of udpipe_server is
udpipe_server [options] port default_model (model_ids model_file acknowledgements)*
Options: --concurrent_models=maximum concurrently loaded models (default 10)
--daemon (daemonize after start)
--no_check_models_loadable (do not check models are loadable)
--no_preload_default (do not preload default model)
The udpipe_server can run either in foreground or in background (when --daemon
is used).
For every model, you specify three arguments:
- model_ids: possibly multiple model ids, separated by a semicolon. Note that
apart from splitting on semicolon, every hyphen-separated prefix of any id is
also considered an id
- model_file: filename containing the model
- acknowledgements: if non-empty, it is added to a list of acknowledgements
returned when using this model
Since UDPipe 1.1.1, the models are loaded on demand, so that at most
concurrent_models (default 10) are kept in memory at the same time. The model
files are opened during start and never closed until the server stops. Unless
no_check_models_loadable is specified, the model files are also checked to be
loadable during start. Note that the default model is preloaded and never
released, unless no_preload_default is given. (Before UDPipe 1.1.1, specified
model files were loaded during start and kept in memory all the time.)
Training UDPipe Models
----------------------
Custom UDPipe models can be trained using the following syntax:
udpipe --train model.output [--heldout=heldout_data] training_file ...
The training data should be in the CoNLL-U format
(http://universaldependencies.org/docs/format.html).
By default, three model components are trained - tokenizer, tagger and parser.
Any subset of the model components can be trained and a model component may be
copied from an existing model.
The training options are specified for each model component separately using the
--tokenizer, --tagger and --parser options. If a model component should not be
trained, value none should be used (e.g., --tagger=none).
The options are name=value pairs separated by a semicolon. The value can be
either a simple string value (ending by a semicolon), file content specified as
name=file:filename, or an arbitrary string value specified as
name=data:length:value, where the value is exactly length bytes long.
Reusing Components from Existing Models
---------------------------------------
The model components (tagger, parser or tagger) can be reused from existing
models, by specifying the from_model=file:filename option.
Random Hyperparameter Search
----------------------------
The default values of hyperparameters are set to the values which were used the
most during UD 1.2 models training, but if you want to reach best performance,
the hyperparameters must be tuned.
Apart from manual grid search, UDPipe can perform a simple random search. You
can perform the random search by repeatedly training UDPipe (preferably in
parallel, most likely on different computers) while specifying different
training run number - some of the hyperparameters (chosen by us; you can of
course override their value by specifying it on the command line) change their
values in different training runs. The pseudorandom sequences of hyperparameters
are of course deterministic.
The training run can be specified by providing the run=number option to a model
component. The run number 1 is the default one (with the best hyperparameters
for the UD 1.2 models), run numbers 2 and more randomize the hyperparameters.
Tokenizer
---------
The tokenizer is trained using the SpaceAfter=No features in the CoNLL-U files.
If the feature is not present, a detokenizer can be used to guess the
SpaceAfter=No features according to a supplied plain text (which typically does
not overlap with the texts in the CoNLL-U files).
In order to use the detokenizer, use the
detokenizer=file:filename_with_plaintext option. In UD 1.2 models, the optimal
performance is achieved with very small plain texts - only 500kB.
The tokenizer recognizes the following options:
- tokenize_url (default 1): tokenize URLs and emails using a manually
implemented recognizer
- allow_spaces (default 1 if any token contains a space, 0 otherwise): allow
tokens to contain spaces
- dimension (default 24): dimension of character embeddings and of the
per-character bidirectional GRU. Note that inference time is quadratic in this
parameter. Supported values are only 16, 24 and 64, with 64 needed only for
languages with complicated tokenization like Japanese, Chinese or Vietnamese.
- epochs (default 100): the number of epochs to train the tokenizer for
- batch_size (default 50): batch size used during tokenizer training
- learning_rate (default 0.005): the learning rate used during tokenizer
training
- dropout (default 0.1): dropout used during tokenizer training
- early_stopping (default 1 if heldout is given, 0 otherwise): perform early
stopping, choosing training iteration maximizing sentences F1 score plus
tokens F1 score on heldout data
During random hyperparameter search, batch_size is chosen uniformly from
{50,100} and learning_rate logarithmically from <0.0005, 0.01).
Detokenizing CoNLL-U Files
--------------------------
The --detokenizer option allows generating the SpaceAfter=No features
automatically from a given plain text. Even if the current algorithm is very
simple and makes quite a lot of mistakes, the tokenizer trained on generated
features is very close to a tokenizer trained on gold SpaceAfter=No features
(the difference in token F1 score is usually one or two tenths of percent).
The generated SpaceAfter=No features are only used during tokenizer training,
not printed. However, if you would like to obtain the CoNLL-U files with
automatic detokenization (generated SpaceAfter=No features), you can run UDPipe
with the --detokenize option. In this case, you have to supply plain text in the
given language (usually the best results are achieved with just 500kB or 1MB of
text) and UDPipe then detokenizes all the given CoNLL-U files.
The complete usage of the --detokenize option is:
udpipe --detokenize [detokenize_opts] raw_text_file [input_files]
Detokenize opts: --outfile=output file template
Tagger
------
The tagging is currently performed using MorphoDiTa
(http://ufal.mff.cuni.cz/morphodita). The UDPipe tagger consists of possibly
several MorphoDiTa models, each tagging some of the POS tags and/or lemmas.
By default, only one model is constructed, which generates all available tags
(UPOS, XPOS, Feats and Lemma). However, we found out during the UD 1.2 models
training that performance improves if one model tags the UPOS, XPOS and Feats
tags, while the other is performing lemmatization. Therefore, if you utilize two
MorphoDiTa models, by default the first one generates all tags (except lemmas)
and the second one performs lemmatization.
The number of MorphoDiTa models can be specified using the models=number
parameter. All other parameters may be either generic for all models
(guesser_suffix_rules=5), or specific for a given model
(guesser_suffix_rules_2=6), including the from_model option (therefore,
MorphoDiTa models can be trained separately and then combined together into one
UDPipe model).
Every model utilizes UPOS for disambiguation and the first model is the one
producing the UPOS tags on output.
The tagger recognizes the following options:
- use_lemma (default for the second model and also if there is only one model):
use the lemma field internally to perform disambiguation; the lemma may be not
outputted
- provide_lemma (default for the second model and also if there is only one
model): produce the disambiguated lemma on output
- use_xpostag (default for the first model): use the XPOS tags internally to
perform disambiguation; it may not be outputted
- provide_xpostag (default for the first model): produce the disambiguated XPOS
tag on output
- use_feats (default for the first model): use the Feats internally to perform
disambiguation; it may not be outputted
- provide_feats (default for the first model): produce the disambiguated Feats
field on output
- dictionary_max_form_analyses (default 0 - unlimited): the maximum number of
(most frequent) form analyses from UD training data that are to be kept in the
morphological dictionary
- dictionary_file (default empty): use a given custom morphological dictionary,
where each line contains 5 tab-separated fields FORM, LEMMA, UPOSTAG, XPOSTAG
and FEATS. Note that this dictionary data is appended to the dictionary
created from the UD training data, not replacing it.
- guesser_suffix_rules (default 8): number of rules generated for every suffix
- guesser_prefixes_max (default 4 if ``provide_lemma`, 0 otherwise): maximum
number of form-generating prefixes to use in the guesser
- guesser_prefix_min_count (default 10): minimum number of occurrences of
form-generating prefix to consider using it in the guesser
- guesser_enrich_dictionary (default 6 if no dictionary_file is passed, 0
otherwise): number of rules generated for forms present in training data
(assuming that the analyses from the training data may not be all)
- iterations (default 20): number of training iterations to perform
- early_stopping (default 1 if heldout is given, 0 otherwise): perform early
stopping, choosing training iteration maximizing tagging accuracy on the
heldout data
- templates (default lemmatizer for second model, tagger otherwise): MorphoDiTa
feature templates to use, either lemmatizer which focuses more on lemmas, or
tagger which focuses more on UPOS/XPOS/FEATS
During random hyperparameter search, guesser_suffix_rules is chosen uniformly
from {5,6,7,8,9,10,11,12} and guesser_enrich_dictionary is chosen uniformly from
{3,4,5,6,7,8,9,10}.
Parser
------
The parsing is performed using Parsito (http://ufal.mff.cuni.cz/parsito), which
is a transition-based parser using a neural-network classifier.
The transition-based systems can be configured by the following options:
- transition_system (default projective): which transition system to use for
parsing (language dependent, you can choose according to language properties
or try all and choose the best one)
- projective: projective stack-based arc standard system with shift, left_arc
and right_arc transitions
- swap: fully non-projective system which extends projective system by adding
the swap transition
- link2: partially non-projective system which extends projective system by
adding left_arc2 and right_arc2 transitions
- transition_oracle (default dynamic/static_lazy_static whichever first is
applicable): which transition oracle to use for the chosen transition_system:
- transition_system=projective: available oracles are static and dynamic
(dynamic usually gives better results, but training time is slower)
- transition_system=swap: available oracles are static_eager and static_lazy
(static_lazy almost always gives better results)
- transition_system=link2: only available oracle is static
- structured_interval (default 8): use search-based oracle in addition to the
translation_oracle specified. This almost always gives better results, but
makes training 2-3 times slower. For details, see the paper Straka et al.
2015: Parsing Universal Dependency Treebanks using Neural Networks and
Search-Based Oracle
- single_root (default 1): allow only single root when parsing, and make sure
only the root node has the root deprel (note that training data are checked to
be in this format)
The Lemmas/UPOS/XPOS/FEATS used by the parser are configured by:
- use_gold_tags (default 0): if false and a tagger exists, the
Lemmas/UPOS/XPOS/FEATS for both the training and heldout data are generated by
the tagger, otherwise they are taken from the gold data
The embeddings used by the parser can be specified as follows:
- embedding_upostag (default 20): the dimension of the UPos embedding used in
the parser
- embedding_feats (default 20): the dimension of the Feats embedding used in
the parser
- embedding_xpostag (default 0): the dimension of the XPos embedding used in
the parser
- embedding_form (default 50): the dimension of the Form embedding used in the
parser
- embedding_lemma (default 0): the dimension of the Lemma embedding used in the
parser
- embedding_deprel (default 20): the dimension of the Deprel embedding used in
the parser
- embedding_form_file: pre-trained word embeddings in word2vec textual format
- embedding_lemma_file: pre-trained lemma embeddings in word2vec textual format
- embedding_form_mincount (default 2): for forms not present in the pre-trained
embeddings, generate random embeddings if the form appears at least this
number of times in the trainig data (forms not present in the pre-trained
embeddings and appearing less number of times are considered OOV)
- embedding_lemma_mincount (default 2): for lemmas not present in the
pre-trained embeddings, generate random embeddings if the lemma appears at
least this number of times in the trainig data (lemmas not present in the
pre-trained embeddings and appearing less number of times are considered OOV)
The neural-network training options:
- iterations (default 10): number of training iterations to use
- hidden_layer (default 200): the size of the hidden layer
- batch_size (default 10): batch size used during neural-network training
- learning_rate (default 0.02): the learning rate used during neural-network
training
- learning_rate_final (0.001): the final learning rate used during
neural-network training
- l2 (0.5): the L2 regularization used during neural-network training
- early_stopping (default 1 if heldout is given, 0 otherwise): perform early
stopping, choosing training iteration maximizing LAS on heldout data
During random hyperparameter search, structured_interval is chosen uniformly
from {0,8,10}, learning_rate is chosen logarithmically from <0.005,0.04) and l2
is chosen uniformly from <0.2,0.6).
Pre-trained Word Embeddings
---------------------------
The pre-trained word embeddings for forms and lemmas can be specified in the
word2vec textual format using the embedding_form_file and embedding_lemma_file
options.
Note that pre-training word embeddings even on the UD data itself improves the
accuracy (we use word2vec with -cbow 0 -size 50 -window 10 -negative 5 -hs 0
-sample 1e-1 -threads 12 -binary 0 -iter 15 -min-count 2 options to pre-train on
the UD data after converting it to the horizontal format using udpipe
--output=horizontal).
Forms and lemmas can contain spaces in CoNLL-U v2, so these spaces are converted
to a Unicode character 'NO-BREAK SPACE' (U+00A0) before performing the embedding
lookup, because spaces are usually used to delimit tokens in word embedding
generating software (both word2vec and glove use spaces to separate words on
input and on output). When using UDPipe to generate plain texts from CoNLL-U
format using --output=horizontal, this space replacing happens automatically.
When looking up an embedding for a given word, the following possibilities are
tried in the following order until a match is found (or an embedding for unknown
word is returned):
- original word
- all but the first character lowercased
- all characters lowercased
- if the word contains only digits, just the first digit is tried
Measuring Model Accuracy
------------------------
Measuring custom model accuracy can be performed by running:
udpipe --accuracy [udpipe_options] udpipe_model file ...
The command syntax is similar to the regular UDPipe operation, only the input
must be always in the CoNLL-U format
(http://universaldependencies.org/docs/format.html) and the --input and --output
options are ignored.
Three different settings (depending on --tokenize(r), --tag(ger) and --parse(r))
can be evaluated:
- --tokenize(r) [--tag(ger) [--parse(r)]]: Tokenizer is used to segment and
tokenize plain text (obtained by SpaceAfter=No features and # newdoc and #
newpar comments in the input file). Optionally, a tagger is used on the
resulting data to obtain Lemma/UPOS/XPOS/Feats columns and eventually a parser
can be used to parse the results.
The tokenizer is evaluated using F1-score on tokens, multi-word tokens,
sentences and words. The words are aligned using a word-alignment algorithm
described in the CoNLL 2017 Shared Task in UD Parsing
(http://universaldependencies.org/conll17). The tagger and parser are
evaluated on aligned words, resulting in F1 scores of
Lemmas/UPOS/XPOS/Feats/UAS/LAS.
- --tag(ger) [--parse(r)]: The gold segmented and tokenized input is tagged
(and then optionally parsed using the tagger outputs) and then evaluated.
- --parse(r): The gold segmented and tokenized input is parsed using gold
morphology (Lemmas/UPOS/XPOS/Feats) and evaluated.
Universal Dependencies 2.0 Models
---------------------------------
Universal Dependencies 2.0 Models are distributed under the CC BY-NC-SA
(http://creativecommons.org/licenses/by-nc-sa/4.0/) licence. The models are
based solely on Universal Dependencies 2.0 (http://hdl.handle.net/11234/1-1983)
treebanks. The models work in UDPipe version 1.2 and later.
Universal Dependencies 2.0 Models are versioned according to the date released
in the format YYMMDD, where YY, MM and DD are two-digit representation of year,
month and day, respectively. The latest version is 170801.
Download
--------
The latest version 170801 of the Universal Dependencies 2.0 models can be
downloaded from LINDAT/CLARIN repository (http://hdl.handle.net/11234/1-2364).
Acknowledgements
----------------
This work has been partially supported and has been using language resources and
tools developed, stored and distributed by the LINDAT/CLARIN project of the
Ministry of Education, Youth and Sports of the Czech Republic (project
LM2015071). The wark was also partially supported by OP VVV projects
CZ.02.1.01/0.0/0.0/16\_013/0001781 and CZ.02.2.69/0.0/0.0/16\_018/0002373, and
by SVV project number 260 453.
The models were trained on Universal Dependencies 2.0
(http://hdl.handle.net/11234/1-1983) treebanks.
For the UD treebanks which do not contain original plain text version, raw text
is used to train the tokenizer instead. The plain texts were taken from the W2C
- Web to Corpus (http://hdl.handle.net/11858/00-097C-0000-0022-6133-9).
Publications
------------
- (Straka et al. 2017) Milan Straka and Jana Straková. Tokenizing, POS
Tagging, Lemmatizing and Parsing UD 2.0 with UDPipe
(http://ufal.mff.cuni.cz/~straka/papers/2017-conll_udpipe.pdf). In Proceedings
of the CoNLL 2017 Shared Task: Multilingual Parsing from Raw Text to Universal
Dependencies, Vancouver, Canada, August 2017.
- (Straka et al. 2016) Straka Milan, Hajič Jan, Straková Jana. UDPipe:
Trainable Pipeline for Processing CoNLL-U Files Performing Tokenization,
Morphological Analysis, POS Tagging and Parsing
(http://ufal.mff.cuni.cz/~straka/papers/2016-lrec_udpipe.pdf). In Proceedings
of the Tenth International Conference on Language Resources and Evaluation
(LREC 2016), Portorož, Slovenia, May 2016.
Model Description
-----------------
The Universal Dependencies 2.0 models contain 68 models of 50 languages, each
consisting of a tokenizer, tagger, lemmatizer and dependency parser, all trained
using the UD data. Note that we use custom train-dev split, by moving sentences
from the beginning of dev data to the end of train data, until the training data
is at least 9 times the dev data.
The tokenizer is trained using the SpaceAfter=No features. If the features are
not present in the data, they can be filled in using raw text in the language in
question.
The tagger, lemmatizer and parser are trained using gold UD data.
Details about model architecture and training process can be found in the
(Straka et al. 2017) paper.
Reproducible Training
---------------------
In case you want to train the same models, scripts for downloading and
resplitting UD 2.0 data, precomputed word embedding, raw texts for tokenizers,
all hyperparameter values and training scripts are available in the second
archive on the model download page (http://hdl.handle.net/11234/1-2364).
Model Performance
-----------------
We present the tagger, lemmatizer and parser performance, measured on the
testing portion of the data, evaluated in three different settings: using raw
text only, using gold tokenization only, and using gold tokenization plus gold
morphology (UPOS, XPOS, FEATS and Lemma).
|| Treebank | Mode | Words | Sents | UPOS | XPOS | Feats | AllTags | Lemma | UAS | LAS |
| Ancient Greek | Raw text | 100.0% | 98.7% | 82.4% | 72.3% | 85.8% | 72.3% | 82.6% | 64.4% | 57.8% |
| Ancient Greek | Gold tok | - | - | 82.4% | 72.4% | 85.8% | 72.3% | 82.7% | 64.6% | 57.9% |
| Ancient Greek | Gold tok+morph | - | - | - | - | - | - | - | 69.2% | 64.4% |
| Ancient Greek-PROIEL | Raw text | 100.0% | 47.2% | 95.8% | 96.0% | 88.6% | 87.2% | 92.6% | 71.8% | 67.1% |
| Ancient Greek-PROIEL | Gold tok | - | - | 95.8% | 96.1% | 88.7% | 87.2% | 92.8% | 77.2% | 72.3% |
| Ancient Greek-PROIEL | Gold tok+morph | - | - | - | - | - | - | - | 79.7% | 76.1% |
| Arabic | Raw text | 93.8% | 83.1% | 88.4% | 83.4% | 83.5% | 82.3% | 87.5% | 71.7% | 65.8% |
| Arabic | Gold tok | - | - | 94.4% | 89.5% | 89.6% | 88.3% | 92.6% | 81.3% | 74.3% |
| Arabic | Gold tok+morph | - | - | - | - | - | - | - | 82.9% | 77.9% |
| Basque | Raw text | 100.0% | 99.5% | 93.2% | - | 87.6% | - | 93.8% | 75.8% | 70.7% |
| Basque | Gold tok | - | - | 93.3% | - | 87.7% | - | 93.9% | 75.9% | 70.8% |
| Basque | Gold tok+morph | - | - | - | - | - | - | - | 82.3% | 78.4% |
| Belarusian | Raw text | 99.4% | 76.8% | 88.2% | 85.6% | 71.7% | 68.6% | 81.3% | 68.0% | 60.6% |
| Belarusian | Gold tok | - | - | 88.7% | 85.7% | 72.4% | 69.2% | 81.5% | 69.4% | 61.9% |
| Belarusian | Gold tok+morph | - | - | - | - | - | - | - | 76.8% | 74.0% |
| Bulgarian | Raw text | 99.9% | 93.9% | 97.6% | 94.6% | 95.6% | 94.0% | 94.6% | 88.8% | 84.8% |
| Bulgarian | Gold tok | - | - | 97.7% | 94.7% | 95.6% | 94.1% | 94.7% | 89.5% | 85.5% |
| Bulgarian | Gold tok+morph | - | - | - | - | - | - | - | 92.6% | 89.1% |
| Catalan | Raw text | 100.0% | 99.2% | 98.0% | 98.0% | 97.1% | 96.5% | 97.9% | 88.8% | 85.7% |
| Catalan | Gold tok | - | - | 98.0% | 98.0% | 97.2% | 96.5% | 97.9% | 88.8% | 85.8% |
| Catalan | Gold tok+morph | - | - | - | - | - | - | - | 91.1% | 88.7% |
| Chinese | Raw text | 90.2% | 98.8% | 84.0% | 83.8% | 89.0% | 82.7% | 90.2% | 62.9% | 58.7% |
| Chinese | Gold tok | - | - | 92.2% | 92.0% | 98.7% | 90.8% | 100.0% | 75.6% | 70.1% |
| Chinese | Gold tok+morph | - | - | - | - | - | - | - | 84.1% | 81.4% |
| Coptic | Raw text | 65.8% | 35.7% | 62.6% | 62.1% | 65.7% | 62.1% | 64.6% | 41.1% | 39.3% |
| Coptic | Gold tok | - | - | 95.1% | 94.3% | 99.7% | 94.2% | 96.2% | 83.2% | 79.2% |
| Coptic | Gold tok+morph | - | - | - | - | - | - | - | 88.1% | 84.9% |
| Croatian | Raw text | 99.9% | 97.0% | 95.9% | - | 84.3% | - | 94.4% | 83.6% | 77.9% |
| Croatian | Gold tok | - | - | 96.0% | - | 84.4% | - | 94.4% | 83.9% | 78.1% |
| Croatian | Gold tok+morph | - | - | - | - | - | - | - | 87.1% | 83.2% |
| Czech | Raw text | 99.9% | 91.6% | 98.3% | 92.8% | 92.1% | 91.7% | 97.8% | 86.8% | 83.2% |
| Czech | Gold tok | - | - | 98.4% | 92.9% | 92.2% | 91.9% | 97.9% | 87.7% | 84.1% |
| Czech | Gold tok+morph | - | - | - | - | - | - | - | 90.2% | 87.5% |
| Czech-CAC | Raw text | 100.0% | 99.8% | 98.1% | 90.6% | 89.4% | 89.1% | 97.0% | 86.9% | 82.7% |
| Czech-CAC | Gold tok | - | - | 98.1% | 90.7% | 89.5% | 89.1% | 97.1% | 87.0% | 82.8% |
| Czech-CAC | Gold tok+morph | - | - | - | - | - | - | - | 89.7% | 86.6% |
| Czech-CLTT | Raw text | 99.5% | 92.3% | 96.5% | 87.5% | 87.8% | 87.3% | 96.8% | 80.2% | 76.6% |
| Czech-CLTT | Gold tok | - | - | 97.0% | 87.9% | 88.3% | 87.7% | 97.2% | 81.0% | 77.6% |
| Czech-CLTT | Gold tok+morph | - | - | - | - | - | - | - | 83.8% | 80.8% |
| Danish | Raw text | 99.8% | 77.9% | 95.2% | - | 94.2% | - | 94.9% | 78.4% | 74.7% |
| Danish | Gold tok | - | - | 95.5% | - | 94.5% | - | 95.0% | 80.4% | 76.6% |
| Danish | Gold tok+morph | - | - | - | - | - | - | - | 85.6% | 82.7% |
| Dutch | Raw text | 99.8% | 77.6% | 91.4% | 88.1% | 89.3% | 87.0% | 89.9% | 75.4% | 69.6% |
| Dutch | Gold tok | - | - | 91.8% | 88.8% | 89.9% | 87.7% | 90.1% | 77.0% | 71.2% |
| Dutch | Gold tok+morph | - | - | - | - | - | - | - | 82.9% | 79.4% |
| Dutch-LassySmall | Raw text | 100.0% | 80.4% | 97.6% | - | 97.2% | - | 98.1% | 84.4% | 82.0% |
| Dutch-LassySmall | Gold tok | - | - | 97.7% | - | 97.4% | - | 98.2% | 87.5% | 85.0% |
| Dutch-LassySmall | Gold tok+morph | - | - | - | - | - | - | - | 89.7% | 87.4% |
| English | Raw text | 99.0% | 76.6% | 93.5% | 92.9% | 94.4% | 91.5% | 96.0% | 80.2% | 77.2% |
| English | Gold tok | - | - | 94.5% | 93.9% | 95.4% | 92.5% | 96.9% | 84.3% | 81.2% |
| English | Gold tok+morph | - | - | - | - | - | - | - | 87.8% | 86.0% |
| English-LinES | Raw text | 99.9% | 86.2% | 95.0% | 92.7% | - | - | - | 78.6% | 74.4% |
| English-LinES | Gold tok | - | - | 95.1% | 92.8% | - | - | - | 79.5% | 75.3% |
| English-LinES | Gold tok+morph | - | - | - | - | - | - | - | 84.1% | 81.1% |
| English-ParTUT | Raw text | 99.6% | 97.5% | 94.2% | 94.0% | 93.3% | 92.0% | 96.9% | 81.6% | 77.9% |
| English-ParTUT | Gold tok | - | - | 94.6% | 94.4% | 93.6% | 92.3% | 97.3% | 82.1% | 78.4% |
| English-ParTUT | Gold tok+morph | - | - | - | - | - | - | - | 86.4% | 84.5% |
| Estonian | Raw text | 99.9% | 94.2% | 91.2% | 93.2% | 85.0% | 83.2% | 84.5% | 72.4% | 65.6% |
| Estonian | Gold tok | - | - | 91.3% | 93.2% | 85.0% | 83.3% | 84.5% | 72.8% | 66.0% |
| Estonian | Gold tok+morph | - | - | - | - | - | - | - | 83.1% | 79.6% |
| Finnish | Raw text | 99.7% | 86.7% | 94.5% | 95.7% | 91.5% | 90.3% | 86.5% | 80.5% | 76.9% |
| Finnish | Gold tok | - | - | 94.9% | 96.0% | 91.8% | 90.7% | 86.8% | 82.0% | 78.4% |
| Finnish | Gold tok+morph | - | - | - | - | - | - | - | 86.9% | 84.7% |
| Finnish-FTB | Raw text | 100.0% | 86.4% | 92.0% | 91.0% | 92.5% | 89.2% | 88.9% | 80.1% | 75.7% |
| Finnish-FTB | Gold tok | - | - | 92.2% | 91.3% | 92.7% | 89.5% | 88.9% | 81.7% | 77.3% |
| Finnish-FTB | Gold tok+morph | - | - | - | - | - | - | - | 88.8% | 86.5% |
| French | Raw text | 98.9% | 94.6% | 95.4% | - | 95.5% | - | 96.6% | 84.2% | 80.7% |
| French | Gold tok | - | - | 96.5% | - | 96.5% | - | 97.6% | 85.4% | 82.0% |
| French | Gold tok+morph | - | - | - | - | - | - | - | 88.4% | 86.0% |
| French-ParTUT | Raw text | 99.0% | 97.8% | 94.5% | 94.2% | 91.9% | 90.8% | 94.3% | 82.9% | 78.7% |
| French-ParTUT | Gold tok | - | - | 95.6% | 95.3% | 92.7% | 91.6% | 95.2% | 84.1% | 80.2% |
| French-ParTUT | Gold tok+morph | - | - | - | - | - | - | - | 88.1% | 85.3% |
| French-Sequoia | Raw text | 99.1% | 84.0% | 95.9% | - | 95.1% | - | 96.8% | 83.2% | 80.6% |
| French-Sequoia | Gold tok | - | - | 96.8% | - | 96.0% | - | 97.7% | 85.1% | 82.7% |
| French-Sequoia | Gold tok+morph | - | - | - | - | - | - | - | 88.7% | 87.4% |
| Galician | Raw text | 99.9% | 95.8% | 97.2% | 96.7% | 99.7% | 96.4% | 97.1% | 81.0% | 77.8% |
| Galician | Gold tok | - | - | 97.2% | 96.8% | 99.8% | 96.4% | 97.1% | 81.2% | 77.9% |
| Galician | Gold tok+morph | - | - | - | - | - | - | - | 83.1% | 80.5% |
| Galician-TreeGal | Raw text | 98.7% | 86.7% | 91.1% | 87.8% | 89.9% | 87.0% | 92.6% | 71.5% | 66.3% |
| Galician-TreeGal | Gold tok | - | - | 92.4% | 88.8% | 91.0% | 88.0% | 93.7% | 74.4% | 68.7% |
| Galician-TreeGal | Gold tok+morph | - | - | - | - | - | - | - | 81.5% | 77.1% |
| German | Raw text | 99.7% | 79.3% | 90.7% | 94.7% | 80.5% | 76.3% | 95.4% | 74.0% | 68.6% |
| German | Gold tok | - | - | 91.2% | 95.0% | 80.9% | 76.7% | 95.6% | 76.5% | 70.7% |
| German | Gold tok+morph | - | - | - | - | - | - | - | 84.7% | 82.2% |
| Gothic | Raw text | 100.0% | 29.5% | 94.2% | 94.8% | 87.6% | 85.6% | 92.9% | 69.7% | 63.5% |
| Gothic | Gold tok | - | - | 94.8% | 95.3% | 88.0% | 86.5% | 92.9% | 78.8% | 72.6% |
| Gothic | Gold tok+morph | - | - | - | - | - | - | - | 82.2% | 78.3% |
| Greek | Raw text | 99.9% | 88.2% | 95.8% | 95.8% | 90.3% | 89.1% | 94.5% | 84.2% | 80.4% |
| Greek | Gold tok | - | - | 96.0% | 96.0% | 90.5% | 89.3% | 94.6% | 85.0% | 81.1% |
| Greek | Gold tok+morph | - | - | - | - | - | - | - | 87.9% | 85.9% |
| Hebrew | Raw text | 85.2% | 100.0% | 80.9% | 80.9% | 77.6% | 76.8% | 79.6% | 62.2% | 57.9% |
| Hebrew | Gold tok | - | - | 95.1% | 95.1% | 91.3% | 90.5% | 93.2% | 84.5% | 78.9% |
| Hebrew | Gold tok+morph | - | - | - | - | - | - | - | 87.8% | 84.3% |
| Hindi | Raw text | 100.0% | 99.1% | 95.8% | 94.9% | 90.3% | 87.7% | 98.0% | 91.3% | 87.3% |
| Hindi | Gold tok | - | - | 95.8% | 94.9% | 90.3% | 87.7% | 98.0% | 91.4% | 87.3% |
| Hindi | Gold tok+morph | - | - | - | - | - | - | - | 93.9% | 91.0% |
| Hungarian | Raw text | 99.8% | 96.2% | 91.6% | - | 70.5% | - | 89.3% | 74.1% | 68.1% |
| Hungarian | Gold tok | - | - | 91.8% | - | 70.6% | - | 89.5% | 74.5% | 68.5% |
| Hungarian | Gold tok+morph | - | - | - | - | - | - | - | 81.2% | 78.5% |
| Indonesian | Raw text | 100.0% | 92.0% | 93.5% | - | 99.5% | - | - | 80.6% | 74.3% |
| Indonesian | Gold tok | - | - | 93.5% | - | 99.6% | - | - | 80.8% | 74.5% |
| Indonesian | Gold tok+morph | - | - | - | - | - | - | - | 83.1% | 79.1% |
| Irish | Raw text | 99.4% | 94.3% | 88.0% | 86.9% | 75.1% | 72.7% | 85.5% | 72.5% | 62.4% |
| Irish | Gold tok | - | - | 88.5% | 87.4% | 75.5% | 73.1% | 86.0% | 73.3% | 63.1% |
| Irish | Gold tok+morph | - | - | - | - | - | - | - | 78.1% | 71.4% |
| Italian | Raw text | 99.8% | 97.1% | 97.2% | 97.0% | 97.0% | 96.1% | 97.3% | 88.8% | 86.1% |
| Italian | Gold tok | - | - | 97.4% | 97.2% | 97.2% | 96.3% | 97.5% | 89.3% | 86.6% |
| Italian | Gold tok+morph | - | - | - | - | - | - | - | 91.3% | 89.7% |
| Japanese | Raw text | 91.9% | 95.1% | 89.1% | - | 91.8% | - | 91.1% | 78.0% | 76.6% |
| Japanese | Gold tok | - | - | 96.6% | - | 100.0% | - | 99.0% | 93.4% | 91.5% |
| Japanese | Gold tok+morph | - | - | - | - | - | - | - | 95.6% | 95.0% |
| Kazakh | Raw text | 94.0% | 84.9% | 52.0% | 52.1% | 47.2% | 40.0% | 59.2% | 40.2% | 23.9% |