Hugging Face's logo

flair
/
frame-english

Token Classification Flair PyTorch
English sequence-tagger-model
Model card Files Community
1
alanakbik commited on
Commit
b610528
•
1 Parent(s): 9aa54d5

initial model commit

Browse files
Files changed (4) hide show
  1. README.md +140 -0
  2. loss.tsv +151 -0
  3. pytorch_model.bin +3 -0
  4. training.log +0 -0
README.md ADDED
@@ -0,0 +1,140 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ tags:
3
+ - flair
4
+ - token-classification
5
+ - sequence-tagger-model
6
+ language: en
7
+ datasets:
8
+ - ontonotes
9
+ inference: false
10
+ ---
11
+
12
+ ## English Verb Disambiguation in Flair (default model)
13
+
14
+ This is the standard verb disambiguation model for English that ships with [Flair](https://github.com/flairNLP/flair/).
15
+
16
+ F1-Score: **89,34** (Ontonotes) - predicts [Proposition Bank verb frames](http://verbs.colorado.edu/propbank/framesets-english-aliases/).
17
+
18
+ Based on [Flair embeddings](https://www.aclweb.org/anthology/C18-1139/) and LSTM-CRF.
19
+
20
+ ---
21
+
22
+ ### Demo: How to use in Flair
23
+
24
+ Requires: **[Flair](https://github.com/flairNLP/flair/)** (`pip install flair`)
25
+
26
+ ```python
27
+ from flair.data import Sentence
28
+ from flair.models import SequenceTagger
29
+
30
+ # load tagger
31
+ tagger = SequenceTagger.load("flair/pos-english")
32
+
33
+ # make example sentence
34
+ sentence = Sentence("I love Berlin.")
35
+
36
+ # predict NER tags
37
+ tagger.predict(sentence)
38
+
39
+ # print sentence
40
+ print(sentence)
41
+
42
+ # print predicted NER spans
43
+ print('The following NER tags are found:')
44
+ # iterate over entities and print
45
+ for entity in sentence.get_spans('pos'):
46
+ print(entity)
47
+
48
+ ```
49
+
50
+ This yields the following output:
51
+ ```
52
+ Span [1]: "I" [− Labels: PRP (1.0)]
53
+ Span [2]: "love" [− Labels: VBP (1.0)]
54
+ Span [3]: "Berlin" [− Labels: NNP (0.9999)]
55
+ Span [4]: "." [− Labels: . (1.0)]
56
+
57
+ ```
58
+
59
+ So, the word "*I*" is labeled as a **pronoun** (PRP), "*love*" is labeled as a **verb** (VBP) and "*Berlin*" is labeled as a **proper noun** (NNP) in the sentence "*TheI love Berlin*".
60
+
61
+
62
+ ---
63
+
64
+ ### Training: Script to train this model
65
+
66
+ The following Flair script was used to train this model:
67
+
68
+ ```python
69
+ from flair.data import Corpus
70
+ from flair.datasets import ColumnCorpus
71
+ from flair.embeddings import WordEmbeddings, StackedEmbeddings, FlairEmbeddings
72
+
73
+ # 1. load the corpus (Ontonotes does not ship with Flair, you need to download and reformat into a column format yourself)
74
+ corpus = ColumnCorpus(
75
+ "resources/tasks/srl", column_format={1: "text", 11: "frame"}
76
+ )
77
+
78
+
79
+ # 2. what tag do we want to predict?
80
+ tag_type = 'frame'
81
+
82
+ # 3. make the tag dictionary from the corpus
83
+ tag_dictionary = corpus.make_tag_dictionary(tag_type=tag_type)
84
+
85
+ # 4. initialize each embedding we use
86
+ embedding_types = [
87
+
88
+ BytePairEmbeddings("en"),
89
+
90
+ FlairEmbeddings("news-forward-fast"),
91
+
92
+ FlairEmbeddings("news-backward-fast"),
93
+ ]
94
+
95
+ # embedding stack consists of Flair and GloVe embeddings
96
+ embeddings = StackedEmbeddings(embeddings=embedding_types)
97
+
98
+ # 5. initialize sequence tagger
99
+ from flair.models import SequenceTagger
100
+
101
+ tagger = SequenceTagger(hidden_size=256,
102
+ embeddings=embeddings,
103
+ tag_dictionary=tag_dictionary,
104
+ tag_type=tag_type)
105
+
106
+ # 6. initialize trainer
107
+ from flair.trainers import ModelTrainer
108
+
109
+ trainer = ModelTrainer(tagger, corpus)
110
+
111
+ # 7. run training
112
+ trainer.train('resources/taggers/frame-english',
113
+ train_with_dev=True,
114
+ max_epochs=150)
115
+ ```
116
+
117
+
118
+
119
+ ---
120
+
121
+ ### Cite
122
+
123
+ Please cite the following paper when using this model.
124
+
125
+ ```
126
+ @inproceedings{akbik2019flair,
127
+ title={FLAIR: An easy-to-use framework for state-of-the-art NLP},
128
+ author={Akbik, Alan and Bergmann, Tanja and Blythe, Duncan and Rasul, Kashif and Schweter, Stefan and Vollgraf, Roland},
129
+ booktitle={{NAACL} 2019, 2019 Conference of the North American Chapter of the Association for Computational Linguistics (Demonstrations)},
130
+ pages={54--59},
131
+ year={2019}
132
+ }
133
+
134
+ ```
135
+
136
+ ---
137
+
138
+ ### Issues?
139
+
140
+ The Flair issue tracker is available [here](https://github.com/flairNLP/flair/issues/).
loss.tsv ADDED
@@ -0,0 +1,151 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ EPOCH TIMESTAMP BAD_EPOCHS LEARNING_RATE TRAIN_LOSS
2
+ 0 22:16:56 0 0.1000 1.0039893659218302
3
+ 1 22:19:32 0 0.1000 0.7061636147521577
4
+ 2 22:22:09 0 0.1000 0.5969159854133174
5
+ 3 22:24:45 0 0.1000 0.5266432488077092
6
+ 4 22:27:20 0 0.1000 0.4787272334942278
7
+ 5 22:29:57 0 0.1000 0.4413028061221231
8
+ 6 22:32:31 0 0.1000 0.4140418250718207
9
+ 7 22:35:07 0 0.1000 0.38958123512706666
10
+ 8 22:37:44 0 0.1000 0.3678484100721917
11
+ 9 22:40:17 0 0.1000 0.35009735131601116
12
+ 10 22:42:53 0 0.1000 0.33932022991607774
13
+ 11 22:45:29 0 0.1000 0.3243708434419812
14
+ 12 22:48:05 0 0.1000 0.31249178839742014
15
+ 13 22:50:41 0 0.1000 0.30411046091678007
16
+ 14 22:53:17 0 0.1000 0.29482137056735325
17
+ 15 22:55:52 0 0.1000 0.2861803585853217
18
+ 16 22:58:28 0 0.1000 0.2790562824398842
19
+ 17 23:01:05 0 0.1000 0.2705823567116036
20
+ 18 23:03:41 0 0.1000 0.26671336980923166
21
+ 19 23:06:18 0 0.1000 0.2593297838124464
22
+ 20 23:08:54 0 0.1000 0.25446213510520055
23
+ 21 23:11:30 0 0.1000 0.24768655170809548
24
+ 22 23:14:06 0 0.1000 0.24429248825559077
25
+ 23 23:16:43 0 0.1000 0.24085550292482916
26
+ 24 23:19:19 0 0.1000 0.23584571266792856
27
+ 25 23:21:53 0 0.1000 0.23129866656829726
28
+ 26 23:24:19 0 0.1000 0.22962194557741003
29
+ 27 23:26:57 0 0.1000 0.22643692269921303
30
+ 28 23:29:34 0 0.1000 0.22286132424226346
31
+ 29 23:32:12 0 0.1000 0.22026822692661915
32
+ 30 23:34:49 0 0.1000 0.21729991109866018
33
+ 31 23:37:26 0 0.1000 0.2145115195748941
34
+ 32 23:40:07 0 0.1000 0.2117686057807702
35
+ 33 23:42:44 0 0.1000 0.20961582855795916
36
+ 34 23:45:20 0 0.1000 0.20412308041920077
37
+ 35 23:47:56 0 0.1000 0.20272977979678028
38
+ 36 23:50:33 0 0.1000 0.2013391835748587
39
+ 37 23:53:05 0 0.1000 0.19829643093190102
40
+ 38 23:55:41 1 0.1000 0.19841094032211123
41
+ 39 23:58:17 0 0.1000 0.19503913455695476
42
+ 40 00:00:52 0 0.1000 0.19209832764880838
43
+ 41 00:03:28 0 0.1000 0.19079046875519573
44
+ 42 00:06:04 0 0.1000 0.18920110067387796
45
+ 43 00:08:41 0 0.1000 0.18687667162069735
46
+ 44 00:11:17 0 0.1000 0.18621931438879022
47
+ 45 00:13:53 0 0.1000 0.18598864254383546
48
+ 46 00:16:24 0 0.1000 0.18228494104349388
49
+ 47 00:18:57 0 0.1000 0.18072697211267813
50
+ 48 00:21:33 0 0.1000 0.1804653215984691
51
+ 49 00:24:09 0 0.1000 0.17930122343999036
52
+ 50 00:26:43 0 0.1000 0.17748751340609675
53
+ 51 00:29:19 0 0.1000 0.1752133839085417
54
+ 52 00:31:55 1 0.1000 0.17594264423509814
55
+ 53 00:34:31 0 0.1000 0.17262986932723026
56
+ 54 00:37:03 0 0.1000 0.17199531916318075
57
+ 55 00:39:34 0 0.1000 0.1710775262061155
58
+ 56 00:42:11 0 0.1000 0.16777184028670472
59
+ 57 00:44:47 1 0.1000 0.1685719361957514
60
+ 58 00:47:23 0 0.1000 0.16631305848090153
61
+ 59 00:50:00 0 0.1000 0.1660435242725993
62
+ 60 00:52:37 0 0.1000 0.1637892189354829
63
+ 61 00:55:13 1 0.1000 0.16440932420486548
64
+ 62 00:57:49 0 0.1000 0.16141063866609673
65
+ 63 01:00:25 1 0.1000 0.16214315994871112
66
+ 64 01:03:01 0 0.1000 0.15980435563848827
67
+ 65 01:05:37 1 0.1000 0.1604709279059239
68
+ 66 01:08:14 0 0.1000 0.15847808923361437
69
+ 67 01:10:50 0 0.1000 0.157500858904337
70
+ 68 01:13:26 0 0.1000 0.15638931530826497
71
+ 69 01:16:03 0 0.1000 0.1555610438323808
72
+ 70 01:18:39 0 0.1000 0.1532352480587532
73
+ 71 01:21:15 1 0.1000 0.15329415766416857
74
+ 72 01:23:47 2 0.1000 0.1536654862269478
75
+ 73 01:26:24 0 0.1000 0.1525104184111334
76
+ 74 01:29:00 0 0.1000 0.15176699819148712
77
+ 75 01:31:35 0 0.1000 0.15075179436999672
78
+ 76 01:34:11 0 0.1000 0.14847835810960464
79
+ 77 01:36:48 1 0.1000 0.14985399553517126
80
+ 78 01:39:25 0 0.1000 0.14741164753740688
81
+ 79 01:42:01 1 0.1000 0.14828819069519358
82
+ 80 01:44:37 0 0.1000 0.1462389815936111
83
+ 81 01:47:13 1 0.1000 0.14666952050378862
84
+ 82 01:49:49 2 0.1000 0.1462645870208178
85
+ 83 01:52:26 0 0.1000 0.14490419472866464
86
+ 84 01:55:02 1 0.1000 0.1463355064799763
87
+ 85 01:57:38 0 0.1000 0.1448573852583485
88
+ 86 02:00:11 0 0.1000 0.14278261882094842
89
+ 87 02:02:46 1 0.1000 0.1432769181270082
90
+ 88 02:05:23 0 0.1000 0.14183188567605784
91
+ 89 02:07:56 0 0.1000 0.14052370639201606
92
+ 90 02:10:32 1 0.1000 0.14088499827486164
93
+ 91 02:13:08 0 0.1000 0.1392386663128745
94
+ 92 02:15:44 1 0.1000 0.13959584156859595
95
+ 93 02:18:21 0 0.1000 0.13846903681333336
96
+ 94 02:20:57 1 0.1000 0.13860615584929034
97
+ 95 02:23:33 0 0.1000 0.13762011011113537
98
+ 96 02:26:09 0 0.1000 0.13614174911436044
99
+ 97 02:28:45 1 0.1000 0.13641496248121532
100
+ 98 02:31:22 0 0.1000 0.13551912194855933
101
+ 99 02:33:57 1 0.1000 0.13605132245230225
102
+ 100 02:36:34 0 0.1000 0.13385561471277813
103
+ 101 02:39:10 1 0.1000 0.13484313386931734
104
+ 102 02:41:46 0 0.1000 0.13224975113036497
105
+ 103 02:44:22 1 0.1000 0.13391570887827087
106
+ 104 02:46:59 0 0.1000 0.13182711181002404
107
+ 105 02:49:35 1 0.1000 0.13251979998822483
108
+ 106 02:52:12 2 0.1000 0.13202438766523353
109
+ 107 02:54:48 0 0.1000 0.13128281369805336
110
+ 108 02:57:24 0 0.1000 0.12939676740540648
111
+ 109 03:00:00 1 0.1000 0.12954153502113977
112
+ 110 03:02:36 2 0.1000 0.13075922897964154
113
+ 111 03:05:13 3 0.1000 0.13048652087320697
114
+ 112 03:07:49 0 0.1000 0.12890468911812553
115
+ 113 03:10:25 1 0.1000 0.12892177955019024
116
+ 114 03:13:01 0 0.1000 0.128635391510039
117
+ 115 03:15:36 0 0.1000 0.12629613022478123
118
+ 116 03:18:12 1 0.1000 0.12783387429590495
119
+ 117 03:20:49 2 0.1000 0.12665061510255876
120
+ 118 03:23:19 3 0.1000 0.1267600216526749
121
+ 119 03:25:55 0 0.1000 0.12617559208200788
122
+ 120 03:28:32 0 0.1000 0.12521145474798273
123
+ 121 03:31:08 0 0.1000 0.12497721069053097
124
+ 122 03:33:44 1 0.1000 0.12529846722646704
125
+ 123 03:36:20 0 0.1000 0.1248895901722728
126
+ 124 03:38:57 0 0.1000 0.12346008466170082
127
+ 125 03:41:33 1 0.1000 0.1235640672074174
128
+ 126 03:44:10 0 0.1000 0.12268940063369162
129
+ 127 03:46:46 0 0.1000 0.12130651440822854
130
+ 128 03:49:22 1 0.1000 0.12257342219774453
131
+ 129 03:51:58 2 0.1000 0.12226920841086024
132
+ 130 03:54:35 3 0.1000 0.12219730263821921
133
+ 131 03:57:11 4 0.1000 0.12310085796030625
134
+ 132 03:59:42 0 0.0500 0.12108087065506656
135
+ 133 04:02:18 0 0.0500 0.11954833609737316
136
+ 134 04:04:55 1 0.0500 0.11963284577120026
137
+ 135 04:07:31 0 0.0500 0.11788792767192957
138
+ 136 04:10:07 1 0.0500 0.11916132699417056
139
+ 137 04:12:43 2 0.0500 0.11934914865372878
140
+ 138 04:15:16 0 0.0500 0.11769732458130369
141
+ 139 04:17:49 0 0.0500 0.11766204663464483
142
+ 140 04:20:25 0 0.0500 0.11757042552291784
143
+ 141 04:23:01 0 0.0500 0.11738244868252637
144
+ 142 04:25:37 0 0.0500 0.11603592486454631
145
+ 143 04:28:14 0 0.0500 0.11534709937167617
146
+ 144 04:30:50 1 0.0500 0.11650450829627379
147
+ 145 04:33:26 2 0.0500 0.11535481170803871
148
+ 146 04:36:02 3 0.0500 0.11545257576934571
149
+ 147 04:38:38 0 0.0500 0.1152307657166472
150
+ 148 04:41:13 1 0.0500 0.11627023742165206
151
+ 149 04:43:46 0 0.0500 0.1151900436061452
pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cd17c26020bd000092df4319d51962ab2e74d15e3bb43205271fd143ef36cfbc
3
+ size 290521503
training.log ADDED
The diff for this file is too large to render. See raw diff