Hugging Face's logo

flair
/
ner-english-ontonotes-fast

Token Classification
Flair
PyTorch
English
sequence-tagger-model
Model card Files Community
1
alanakbik commited on
Commit
4e67b84
·
1 Parent(s): d5724d0

initial model commit

Browse files
Files changed (3) hide show
  1. README.md +164 -0
  2. loss.tsv +151 -0
  3. pytorch_model.bin +3 -0
README.md ADDED
@@ -0,0 +1,164 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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 NER in Flair (Ontonotes fast model)
13
+
14
+ This is the fast version of the 18-class NER model for English that ships with [Flair](https://github.com/flairNLP/flair/).
15
+
16
+ F1-Score: **89.3** (Ontonotes)
17
+
18
+ Predicts 18 tags:
19
+
20
+ | **tag** | **meaning** |
21
+ |---------------------------------|-----------|
22
+ | CARDINAL | cardinal value |
23
+ | DATE | date value |
24
+ | EVENT | event name |
25
+ | FAC | building name |
26
+ | GPE | geo-political entity |
27
+ | LANGUAGE | language name |
28
+ | LAW | law name |
29
+ | LOC | location name |
30
+ | MONEY | money name |
31
+ | NORP | affiliation |
32
+ | ORDINAL | ordinal value |
33
+ | ORG | organization name |
34
+ | PERCENT | percent value |
35
+ | PERSON | person name |
36
+ | PRODUCT | product name |
37
+ | QUANTITY | quantity value |
38
+ | TIME | time value |
39
+ | WORK_OF_ART | name of work of art |
40
+
41
+ Based on [Flair embeddings](https://www.aclweb.org/anthology/C18-1139/) and LSTM-CRF.
42
+
43
+ ---
44
+
45
+ ### Demo: How to use in Flair
46
+
47
+ Requires: **[Flair](https://github.com/flairNLP/flair/)** (`pip install flair`)
48
+
49
+ ```python
50
+ from flair.data import Sentence
51
+ from flair.models import SequenceTagger
52
+
53
+ # load tagger
54
+ tagger = SequenceTagger.load("flair/ner-english-ontonotes-fast")
55
+
56
+ # make example sentence
57
+ sentence = Sentence("On September 1st George Washington won 1 dollar.")
58
+
59
+ # predict NER tags
60
+ tagger.predict(sentence)
61
+
62
+ # print sentence
63
+ print(sentence)
64
+
65
+ # print predicted NER spans
66
+ print('The following NER tags are found:')
67
+ # iterate over entities and print
68
+ for entity in sentence.get_spans('ner'):
69
+ print(entity)
70
+
71
+ ```
72
+
73
+ This yields the following output:
74
+ ```
75
+ Span [2,3]: "September 1st" [− Labels: DATE (0.8824)]
76
+ Span [4,5]: "George Washington" [− Labels: PERSON (0.9604)]
77
+ Span [7,8]: "1 dollar" [− Labels: MONEY (0.9837)]
78
+ ```
79
+
80
+ So, the entities "*September 1st*" (labeled as a **date**), "*George Washington*" (labeled as a **person**) and "*1 dollar*" (labeled as a **money**) are found in the sentence "*On September 1st George Washington won 1 dollar*".
81
+
82
+
83
+ ---
84
+
85
+ ### Training: Script to train this model
86
+
87
+ The following Flair script was used to train this model:
88
+
89
+ ```python
90
+ from flair.data import Corpus
91
+ from flair.datasets import ColumnCorpus
92
+ from flair.embeddings import WordEmbeddings, StackedEmbeddings, FlairEmbeddings
93
+
94
+ # 1. load the corpus (Ontonotes does not ship with Flair, you need to download and reformat into a column format yourself)
95
+ corpus: Corpus = ColumnCorpus(
96
+ "resources/tasks/onto-ner",
97
+ column_format={0: "text", 1: "pos", 2: "upos", 3: "ner"},
98
+ tag_to_bioes="ner",
99
+ )
100
+
101
+ # 2. what tag do we want to predict?
102
+ tag_type = 'ner'
103
+
104
+ # 3. make the tag dictionary from the corpus
105
+ tag_dictionary = corpus.make_tag_dictionary(tag_type=tag_type)
106
+
107
+ # 4. initialize each embedding we use
108
+ embedding_types = [
109
+
110
+ # GloVe embeddings
111
+ WordEmbeddings('en-crawl'),
112
+
113
+ # contextual string embeddings, forward
114
+ FlairEmbeddings('news-forward-fast'),
115
+
116
+ # contextual string embeddings, backward
117
+ FlairEmbeddings('news-backward-fast'),
118
+ ]
119
+
120
+ # embedding stack consists of Flair and GloVe embeddings
121
+ embeddings = StackedEmbeddings(embeddings=embedding_types)
122
+
123
+ # 5. initialize sequence tagger
124
+ from flair.models import SequenceTagger
125
+
126
+ tagger = SequenceTagger(hidden_size=256,
127
+ embeddings=embeddings,
128
+ tag_dictionary=tag_dictionary,
129
+ tag_type=tag_type)
130
+
131
+ # 6. initialize trainer
132
+ from flair.trainers import ModelTrainer
133
+
134
+ trainer = ModelTrainer(tagger, corpus)
135
+
136
+ # 7. run training
137
+ trainer.train('resources/taggers/ner-english-ontonotes-fast',
138
+ train_with_dev=True,
139
+ max_epochs=150)
140
+ ```
141
+
142
+
143
+
144
+ ---
145
+
146
+ ### Cite
147
+
148
+ Please cite the following paper when using this model.
149
+
150
+ ```
151
+ @inproceedings{akbik2018coling,
152
+ title={Contextual String Embeddings for Sequence Labeling},
153
+ author={Akbik, Alan and Blythe, Duncan and Vollgraf, Roland},
154
+ booktitle = {{COLING} 2018, 27th International Conference on Computational Linguistics},
155
+ pages = {1638--1649},
156
+ year = {2018}
157
+ }
158
+ ```
159
+
160
+ ---
161
+
162
+ ### Issues?
163
+
164
+ 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 23:09:41 0 0.1000 3.2296528358391994
3
+ 1 23:22:27 0 0.1000 1.5920132506231093
4
+ 2 23:35:15 0 0.1000 1.3207480492007058
5
+ 3 23:48:00 0 0.1000 1.1772499416796667
6
+ 4 00:00:46 0 0.1000 1.075935570046587
7
+ 5 00:13:32 0 0.1000 1.0152623265981675
8
+ 6 00:26:17 0 0.1000 0.9606320605637892
9
+ 7 00:39:04 0 0.1000 0.9167048768839746
10
+ 8 00:51:51 0 0.1000 0.8835731125327776
11
+ 9 01:04:36 0 0.1000 0.8626177742233816
12
+ 10 01:17:23 0 0.1000 0.8316960627290437
13
+ 11 01:30:11 0 0.1000 0.8111448318440959
14
+ 12 01:42:58 0 0.1000 0.7882771724912355
15
+ 13 01:55:47 0 0.1000 0.7772803506704996
16
+ 14 02:08:34 0 0.1000 0.7563603024718897
17
+ 15 02:21:27 0 0.1000 0.7446274626255035
18
+ 16 02:34:27 0 0.1000 0.7325120459637552
19
+ 17 02:47:28 0 0.1000 0.7164706003328539
20
+ 18 03:00:18 0 0.1000 0.6953795908475822
21
+ 19 03:13:06 1 0.1000 0.6954403392886216
22
+ 20 03:26:06 0 0.1000 0.6850152359368666
23
+ 21 03:38:58 0 0.1000 0.6716467778918878
24
+ 22 03:52:00 0 0.1000 0.6623472330142867
25
+ 23 04:04:49 0 0.1000 0.6543980406590227
26
+ 24 04:17:42 0 0.1000 0.6484055938135903
27
+ 25 04:30:37 0 0.1000 0.6380644176703579
28
+ 26 04:43:42 0 0.1000 0.637061537284896
29
+ 27 04:56:49 0 0.1000 0.6342280050493636
30
+ 28 05:09:43 0 0.1000 0.6191383106472357
31
+ 29 05:22:41 0 0.1000 0.613722907181056
32
+ 30 05:35:39 0 0.1000 0.6094017116303714
33
+ 31 05:48:35 0 0.1000 0.600858421494376
34
+ 32 06:01:36 1 0.1000 0.6034416157400833
35
+ 33 06:14:33 0 0.1000 0.5933149380335268
36
+ 34 06:27:31 0 0.1000 0.5902228662202943
37
+ 35 06:40:30 0 0.1000 0.5814154819609983
38
+ 36 06:53:30 1 0.1000 0.5834101356706529
39
+ 37 07:06:23 2 0.1000 0.581889728307724
40
+ 38 07:19:15 0 0.1000 0.5660430806537844
41
+ 39 07:32:06 1 0.1000 0.5683663231921646
42
+ 40 07:45:02 0 0.1000 0.558734485265219
43
+ 41 07:57:57 0 0.1000 0.5557521927581643
44
+ 42 08:10:54 0 0.1000 0.5528975577568108
45
+ 43 08:24:03 0 0.1000 0.5484419691956268
46
+ 44 08:36:59 1 0.1000 0.5568543178405402
47
+ 45 08:49:58 0 0.1000 0.5446653557610962
48
+ 46 09:03:11 0 0.1000 0.5384193196161738
49
+ 47 09:16:10 0 0.1000 0.5350716501699304
50
+ 48 09:29:27 0 0.1000 0.5284495891883688
51
+ 49 09:42:25 0 0.1000 0.5265147034627087
52
+ 50 09:55:24 0 0.1000 0.5207880691256164
53
+ 51 10:08:23 1 0.1000 0.5229102363901318
54
+ 52 10:21:34 2 0.1000 0.5247485248997527
55
+ 53 10:34:47 0 0.1000 0.5197978817910518
56
+ 54 10:47:47 0 0.1000 0.5088265573809732
57
+ 55 11:00:42 1 0.1000 0.5092196283081792
58
+ 56 11:13:34 2 0.1000 0.5101087852019184
59
+ 57 11:26:17 3 0.1000 0.5114516223376652
60
+ 58 11:39:13 0 0.1000 0.5055079925453888
61
+ 59 11:52:01 0 0.1000 0.5028705823815094
62
+ 60 12:04:48 1 0.1000 0.5077681644907538
63
+ 61 12:17:31 0 0.1000 0.493943511888666
64
+ 62 12:30:40 1 0.1000 0.49909354941884304
65
+ 63 12:43:38 0 0.1000 0.4935010322253659
66
+ 64 12:56:41 0 0.1000 0.49263371167317876
67
+ 65 13:09:30 1 0.1000 0.49477802515029906
68
+ 66 13:22:17 0 0.1000 0.4894873375937624
69
+ 67 13:35:02 1 0.1000 0.4900508265337854
70
+ 68 13:47:54 0 0.1000 0.48103408392307895
71
+ 69 14:00:52 1 0.1000 0.482991996205078
72
+ 70 14:14:00 2 0.1000 0.48390048535364977
73
+ 71 14:27:07 3 0.1000 0.4846018293106331
74
+ 72 14:40:01 0 0.1000 0.480364065518919
75
+ 73 14:53:03 0 0.1000 0.47812797212375785
76
+ 74 15:05:56 0 0.1000 0.4719673815369606
77
+ 75 15:18:49 1 0.1000 0.4757926766265113
78
+ 76 15:31:41 0 0.1000 0.46993971415285796
79
+ 77 15:44:32 1 0.1000 0.4722084892920728
80
+ 78 15:57:20 2 0.1000 0.47019626197106434
81
+ 79 16:10:14 0 0.1000 0.4698862019406175
82
+ 80 16:23:07 0 0.1000 0.46922945463994764
83
+ 81 16:35:56 0 0.1000 0.46842513900320487
84
+ 82 16:48:45 0 0.1000 0.4596653935369456
85
+ 83 17:01:35 1 0.1000 0.46220648641293904
86
+ 84 17:14:30 2 0.1000 0.4606187267460913
87
+ 85 17:27:26 0 0.1000 0.45330136719177355
88
+ 86 17:40:23 1 0.1000 0.4552749111404959
89
+ 87 17:53:27 2 0.1000 0.4595688052559799
90
+ 88 18:06:15 0 0.1000 0.45305425408876165
91
+ 89 18:19:11 1 0.1000 0.4585241228904364
92
+ 90 18:32:06 2 0.1000 0.4604555804212138
93
+ 91 18:45:08 3 0.1000 0.4554677476860442
94
+ 92 18:58:04 0 0.1000 0.4489068861839906
95
+ 93 19:10:59 1 0.1000 0.45116823060332606
96
+ 94 19:23:45 2 0.1000 0.4489288940407195
97
+ 95 19:36:36 0 0.1000 0.44275297022653076
98
+ 96 19:49:23 1 0.1000 0.4452887841103212
99
+ 97 20:02:12 2 0.1000 0.4453210852955872
100
+ 98 20:14:58 3 0.1000 0.4464509549905669
101
+ 99 20:28:04 4 0.1000 0.44596645003782126
102
+ 100 20:41:13 0 0.0500 0.41892101504330365
103
+ 101 20:54:19 0 0.0500 0.3984660865253997
104
+ 102 21:07:12 0 0.0500 0.3909759231783309
105
+ 103 21:20:22 0 0.0500 0.38897691094088105
106
+ 104 21:33:12 0 0.0500 0.38891661282980217
107
+ 105 21:46:02 0 0.0500 0.3788945140141361
108
+ 106 21:58:47 0 0.0500 0.37884936595300456
109
+ 107 22:11:42 0 0.0500 0.37052636316924725
110
+ 108 22:24:46 1 0.0500 0.3740457253186208
111
+ 109 22:37:55 2 0.0500 0.3722470565224594
112
+ 110 22:51:02 0 0.0500 0.3700024942125914
113
+ 111 23:04:09 0 0.0500 0.36512322439337674
114
+ 112 23:16:50 0 0.0500 0.360866011077503
115
+ 113 23:29:36 0 0.0500 0.3606146826777818
116
+ 114 23:42:29 1 0.0500 0.36123125600364975
117
+ 115 23:55:19 0 0.0500 0.3542564442135253
118
+ 116 00:08:10 1 0.0500 0.3587884951396933
119
+ 117 00:20:55 2 0.0500 0.3573749113926348
120
+ 118 00:33:41 0 0.0500 0.3419303108273812
121
+ 119 00:46:29 1 0.0500 0.3496313952162581
122
+ 120 00:59:24 2 0.0500 0.3525260106769373
123
+ 121 01:12:26 3 0.0500 0.3468611272579094
124
+ 122 01:25:16 4 0.0500 0.3510722661468218
125
+ 123 01:38:13 0 0.0250 0.3362890320566465
126
+ 124 01:51:10 0 0.0250 0.3326504863151964
127
+ 125 02:03:59 0 0.0250 0.3284786853756545
128
+ 126 02:16:50 0 0.0250 0.32160441274912854
129
+ 127 02:29:40 1 0.0250 0.325127320936266
130
+ 128 02:42:33 0 0.0250 0.31972479293931205
131
+ 129 02:55:40 0 0.0250 0.3179566932230625
132
+ 130 03:08:39 1 0.0250 0.3196000659690713
133
+ 131 03:21:27 2 0.0250 0.32247103282162604
134
+ 132 03:34:27 0 0.0250 0.31690518119425143
135
+ 133 03:47:38 0 0.0250 0.3140822275742045
136
+ 134 04:00:47 1 0.0250 0.31784330996702304
137
+ 135 04:13:51 2 0.0250 0.3173914384504534
138
+ 136 04:26:53 0 0.0250 0.3117131764708825
139
+ 137 04:39:44 0 0.0250 0.30754536822157086
140
+ 138 04:52:35 0 0.0250 0.3054389862997352
141
+ 139 05:05:34 1 0.0250 0.3097640820817565
142
+ 140 05:18:40 2 0.0250 0.3103047859443808
143
+ 141 05:31:34 3 0.0250 0.3074088339310772
144
+ 142 05:44:36 0 0.0250 0.30093574178668686
145
+ 143 05:57:27 1 0.0250 0.30524236956997863
146
+ 144 06:10:33 0 0.0250 0.29936067997846966
147
+ 145 06:23:28 1 0.0250 0.3067479654752983
148
+ 146 06:36:27 2 0.0250 0.30121562754208187
149
+ 147 06:49:30 0 0.0250 0.2983897975537012
150
+ 148 07:02:25 1 0.0250 0.30219820415355125
151
+ 149 07:15:20 2 0.0250 0.30205940004227294
pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3a08d6bcbe6be469b9be1e0bdedb6b740469ac2f2915418af77ab949b965e4a6
3
+ size 1331379415