alanakbik commited on
Commit
eac8f09
1 Parent(s): e7a4695

initial model commit

Browse files

Files changed (4) hide show
  1. README.md +163 -0
  2. loss.tsv +132 -0
  3. pytorch_model.bin +3 -0
  4. training.log +0 -0
README.md ADDED
@@ -0,0 +1,163 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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 Universal Part-of-Speech Tagging in Flair (fast model)
13
+
14
+ This is the fast universal part-of-speech tagging model for English that ships with [Flair](https://github.com/flairNLP/flair/).
15
+
16
+ F1-Score: **98,47** (Ontonotes)
17
+
18
+ Predicts universal POS tags:
19
+
20
+ | **tag** | **meaning** |
21
+ |---------------------------------|-----------|
22
+ |ADJ | adjective |
23
+ | ADP | adposition |
24
+ | ADV | adverb |
25
+ | AUX | auxiliary |
26
+ | CCONJ | coordinating conjunction |
27
+ | DET | determiner |
28
+ | INTJ | interjection |
29
+ | NOUN | noun |
30
+ | NUM | numeral |
31
+ | PART | particle |
32
+ | PRON | pronoun |
33
+ | PROPN | proper noun |
34
+ | PUNCT | punctuation |
35
+ | SCONJ | subordinating conjunction |
36
+ | SYM | symbol |
37
+ | VERB | verb |
38
+ | X | other |
39
+
40
+
41
+
42
+ Based on [Flair embeddings](https://www.aclweb.org/anthology/C18-1139/) and LSTM-CRF.
43
+
44
+ ---
45
+
46
+ ### Demo: How to use in Flair
47
+
48
+ Requires: **[Flair](https://github.com/flairNLP/flair/)** (`pip install flair`)
49
+
50
+ ```python
51
+ from flair.data import Sentence
52
+ from flair.models import SequenceTagger
53
+
54
+ # load tagger
55
+ tagger = SequenceTagger.load("flair/upos-english-fast")
56
+
57
+ # make example sentence
58
+ sentence = Sentence("I love Berlin.")
59
+
60
+ # predict NER tags
61
+ tagger.predict(sentence)
62
+
63
+ # print sentence
64
+ print(sentence)
65
+
66
+ # print predicted NER spans
67
+ print('The following NER tags are found:')
68
+ # iterate over entities and print
69
+ for entity in sentence.get_spans('pos'):
70
+ print(entity)
71
+
72
+ ```
73
+
74
+ This yields the following output:
75
+ ```
76
+ Span [1]: "I" [− Labels: PRON (0.9996)]
77
+ Span [2]: "love" [− Labels: VERB (1.0)]
78
+ Span [3]: "Berlin" [− Labels: PROPN (0.9986)]
79
+ Span [4]: "." [− Labels: PUNCT (1.0)]
80
+ ```
81
+
82
+ So, the word "*I*" is labeled as a **pronoun** (PRON), "*love*" is labeled as a **verb** (VERB) and "*Berlin*" is labeled as a **proper noun** (PROPN) in the sentence "*TheI love Berlin*".
83
+
84
+
85
+ ---
86
+
87
+ ### Training: Script to train this model
88
+
89
+ The following Flair script was used to train this model:
90
+
91
+ ```python
92
+ from flair.data import Corpus
93
+ from flair.datasets import ColumnCorpus
94
+ from flair.embeddings import WordEmbeddings, StackedEmbeddings, FlairEmbeddings
95
+
96
+ # 1. load the corpus (Ontonotes does not ship with Flair, you need to download and reformat into a column format yourself)
97
+ corpus: Corpus = ColumnCorpus(
98
+ "resources/tasks/onto-ner",
99
+ column_format={0: "text", 1: "pos", 2: "upos", 3: "ner"},
100
+ tag_to_bioes="ner",
101
+ )
102
+
103
+ # 2. what tag do we want to predict?
104
+ tag_type = 'upos'
105
+
106
+ # 3. make the tag dictionary from the corpus
107
+ tag_dictionary = corpus.make_tag_dictionary(tag_type=tag_type)
108
+
109
+ # 4. initialize each embedding we use
110
+ embedding_types = [
111
+
112
+ # contextual string embeddings, forward
113
+ FlairEmbeddings('news-forward-fast'),
114
+
115
+ # contextual string embeddings, backward
116
+ FlairEmbeddings('news-backward-fast'),
117
+ ]
118
+
119
+ # embedding stack consists of Flair and GloVe embeddings
120
+ embeddings = StackedEmbeddings(embeddings=embedding_types)
121
+
122
+ # 5. initialize sequence tagger
123
+ from flair.models import SequenceTagger
124
+
125
+ tagger = SequenceTagger(hidden_size=256,
126
+ embeddings=embeddings,
127
+ tag_dictionary=tag_dictionary,
128
+ tag_type=tag_type)
129
+
130
+ # 6. initialize trainer
131
+ from flair.trainers import ModelTrainer
132
+
133
+ trainer = ModelTrainer(tagger, corpus)
134
+
135
+ # 7. run training
136
+ trainer.train('resources/taggers/upos-english-fast',
137
+ train_with_dev=True,
138
+ max_epochs=150)
139
+ ```
140
+
141
+
142
+
143
+ ---
144
+
145
+ ### Cite
146
+
147
+ Please cite the following paper when using this model.
148
+
149
+ ```
150
+ @inproceedings{akbik2018coling,
151
+ title={Contextual String Embeddings for Sequence Labeling},
152
+ author={Akbik, Alan and Blythe, Duncan and Vollgraf, Roland},
153
+ booktitle = {{COLING} 2018, 27th International Conference on Computational Linguistics},
154
+ pages = {1638--1649},
155
+ year = {2018}
156
+ }
157
+ ```
158
+
159
+ ---
160
+
161
+ ### Issues?
162
+
163
+ The Flair issue tracker is available [here](https://github.com/flairNLP/flair/issues/).
loss.tsv ADDED
@@ -0,0 +1,132 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ EPOCH TIMESTAMP BAD_EPOCHS LEARNING_RATE TRAIN_LOSS
2
+ 0 17:03:39 0 0.1000 3.8983645198929984
3
+ 1 17:11:32 0 0.1000 2.396464760978267
4
+ 2 17:19:17 0 0.1000 2.1225894081367636
5
+ 3 17:27:11 0 0.1000 1.9884647013443821
6
+ 4 17:35:04 0 0.1000 1.885558801484558
7
+ 5 17:42:58 0 0.1000 1.8396147863370067
8
+ 6 17:50:52 0 0.1000 1.7672407321210177
9
+ 7 17:58:33 0 0.1000 1.7209206173217522
10
+ 8 18:06:14 0 0.1000 1.684758366188913
11
+ 9 18:13:54 0 0.1000 1.675738244360348
12
+ 10 18:21:33 0 0.1000 1.6346709968683855
13
+ 11 18:29:15 0 0.1000 1.5976093570016465
14
+ 12 18:37:02 0 0.1000 1.572903771839052
15
+ 13 18:44:47 0 0.1000 1.5621724839255495
16
+ 14 18:52:31 0 0.1000 1.5344491067360033
17
+ 15 19:00:09 0 0.1000 1.5280954311141428
18
+ 16 19:07:50 0 0.1000 1.5018526020027556
19
+ 17 19:15:29 1 0.1000 1.505718067726999
20
+ 18 19:23:08 0 0.1000 1.4736719750233416
21
+ 19 19:30:50 1 0.1000 1.4839579892720816
22
+ 20 19:38:35 0 0.1000 1.4585183224250686
23
+ 21 19:46:19 1 0.1000 1.4610416818902177
24
+ 22 19:54:12 0 0.1000 1.437192275220493
25
+ 23 20:01:52 0 0.1000 1.4222804964825793
26
+ 24 20:09:33 0 0.1000 1.4003380132058882
27
+ 25 20:17:17 1 0.1000 1.4161622376374479
28
+ 26 20:24:54 2 0.1000 1.4020970809572149
29
+ 27 20:32:32 0 0.1000 1.3987249116852598
30
+ 28 20:40:11 0 0.1000 1.3625548289969283
31
+ 29 20:47:51 1 0.1000 1.381220668767983
32
+ 30 20:55:30 2 0.1000 1.3701620033552062
33
+ 31 21:03:17 3 0.1000 1.3630763605293237
34
+ 32 21:11:05 0 0.1000 1.3467498509051665
35
+ 33 21:18:54 1 0.1000 1.3495412202095085
36
+ 34 21:26:44 0 0.1000 1.340426192193661
37
+ 35 21:34:35 0 0.1000 1.3255774740228112
38
+ 36 21:42:28 1 0.1000 1.341141459649464
39
+ 37 21:50:16 2 0.1000 1.3301069232652771
40
+ 38 21:58:05 0 0.1000 1.3155438300011293
41
+ 39 22:05:53 1 0.1000 1.3180485034101415
42
+ 40 22:13:33 0 0.1000 1.3101363613583008
43
+ 41 22:21:15 1 0.1000 1.3239353564212908
44
+ 42 22:28:55 0 0.1000 1.2985683835677382
45
+ 43 22:36:36 1 0.1000 1.2987655120300796
46
+ 44 22:44:16 0 0.1000 1.293294859140549
47
+ 45 22:51:55 1 0.1000 1.2934898600825724
48
+ 46 22:59:35 0 0.1000 1.2742974282997959
49
+ 47 23:07:16 0 0.1000 1.257929092420722
50
+ 48 23:14:57 1 0.1000 1.2636124875410548
51
+ 49 23:22:35 2 0.1000 1.2605103574271472
52
+ 50 23:30:15 3 0.1000 1.2628181801202163
53
+ 51 23:37:55 4 0.1000 1.2682071375397017
54
+ 52 23:45:35 0 0.0500 1.2192658351502328
55
+ 53 23:53:15 0 0.0500 1.189723878941446
56
+ 54 00:00:56 0 0.0500 1.181310292977207
57
+ 55 00:08:36 1 0.0500 1.1813142526599596
58
+ 56 00:16:15 0 0.0500 1.1490525012646082
59
+ 57 00:24:04 1 0.0500 1.150567943037681
60
+ 58 00:31:59 2 0.0500 1.153844450498527
61
+ 59 00:39:53 3 0.0500 1.1547138257521503
62
+ 60 00:47:36 0 0.0500 1.138099388097817
63
+ 61 00:55:17 1 0.0500 1.1522783655265592
64
+ 62 01:03:07 0 0.0500 1.1201619118114687
65
+ 63 01:11:03 1 0.0500 1.140103389699504
66
+ 64 01:18:57 2 0.0500 1.1306282293909
67
+ 65 01:26:43 3 0.0500 1.1392165621946442
68
+ 66 01:34:30 4 0.0500 1.1320033756404553
69
+ 67 01:42:13 0 0.0250 1.0931724692290683
70
+ 68 01:50:05 1 0.0250 1.093446401728774
71
+ 69 01:57:59 0 0.0250 1.0766996851900839
72
+ 70 02:05:47 1 0.0250 1.085443768231374
73
+ 71 02:13:32 2 0.0250 1.0840452198824793
74
+ 72 02:21:24 3 0.0250 1.0943272675770634
75
+ 73 02:29:13 0 0.0250 1.0741095490050765
76
+ 74 02:36:59 1 0.0250 1.0775160627657512
77
+ 75 02:44:45 0 0.0250 1.0723835660601562
78
+ 76 02:52:28 0 0.0250 1.0675190647593085
79
+ 77 03:00:14 0 0.0250 1.062752323026927
80
+ 78 03:07:55 1 0.0250 1.0638396440924338
81
+ 79 03:15:34 0 0.0250 1.0551368798849718
82
+ 80 03:23:13 0 0.0250 1.0540687316993498
83
+ 81 03:30:54 0 0.0250 1.0486293900575279
84
+ 82 03:38:39 1 0.0250 1.0578650972190893
85
+ 83 03:46:29 2 0.0250 1.050876642150699
86
+ 84 03:54:21 0 0.0250 1.0444189010476166
87
+ 85 04:02:17 0 0.0250 1.036741197986423
88
+ 86 04:10:04 1 0.0250 1.0422700380716683
89
+ 87 04:17:57 2 0.0250 1.053200504015077
90
+ 88 04:25:50 3 0.0250 1.0567198398428144
91
+ 89 04:33:41 4 0.0250 1.038592992784842
92
+ 90 04:41:31 1 0.0125 1.0402668333278513
93
+ 91 04:49:15 0 0.0125 1.0200082490354214
94
+ 92 04:56:57 1 0.0125 1.0332945613703637
95
+ 93 05:04:44 2 0.0125 1.0235844095023172
96
+ 94 05:12:38 3 0.0125 1.030887721619516
97
+ 95 05:20:33 4 0.0125 1.03034149728856
98
+ 96 05:28:29 1 0.0063 1.032665410379194
99
+ 97 05:36:24 0 0.0063 1.0145184545584445
100
+ 98 05:44:18 0 0.0063 1.004028284752144
101
+ 99 05:52:14 1 0.0063 1.0066242653356408
102
+ 100 06:00:01 2 0.0063 1.0042478304876472
103
+ 101 06:07:45 3 0.0063 1.0221682896141735
104
+ 102 06:15:28 4 0.0063 1.0171712939975397
105
+ 103 06:23:10 1 0.0031 1.0051458630696783
106
+ 104 06:30:52 0 0.0031 0.9894583106828185
107
+ 105 06:38:35 1 0.0031 0.9949013568826441
108
+ 106 06:46:21 2 0.0031 1.0099847611166397
109
+ 107 06:54:17 3 0.0031 1.0110677263646755
110
+ 108 07:02:09 4 0.0031 0.9900631292529826
111
+ 109 07:09:53 1 0.0016 0.9965992866034777
112
+ 110 07:17:39 0 0.0016 0.988067799041856
113
+ 111 07:25:26 1 0.0016 1.002729972747137
114
+ 112 07:33:17 2 0.0016 1.0075599195597307
115
+ 113 07:40:59 3 0.0016 0.9934051743318449
116
+ 114 07:48:47 4 0.0016 0.9908848639141838
117
+ 115 07:56:32 1 0.0008 0.9981947860515342
118
+ 116 08:04:19 2 0.0008 0.9895183097191577
119
+ 117 08:11:59 3 0.0008 0.9920743883836944
120
+ 118 08:19:38 4 0.0008 0.9939175610103698
121
+ 119 08:27:18 1 0.0004 1.0007407332141445
122
+ 120 08:34:58 2 0.0004 1.0047922001807195
123
+ 121 08:42:37 0 0.0004 0.9872947578835037
124
+ 122 08:50:15 0 0.0004 0.9852443703671672
125
+ 123 08:57:54 1 0.0004 0.9936418686610348
126
+ 124 09:05:37 2 0.0004 0.9901605238104766
127
+ 125 09:13:23 3 0.0004 0.9907275987008832
128
+ 126 09:21:09 4 0.0004 0.99081547033112
129
+ 127 09:28:51 1 0.0002 0.9894531191295048
130
+ 128 09:36:34 2 0.0002 0.9955960737309366
131
+ 129 09:44:20 3 0.0002 0.9950949703578679
132
+ 130 09:52:13 4 0.0002 1.0062029107730344
pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4395e8af6f93dab948fb498ab07906a987c2cb63c2fef94472a0ee26092a2023
3
+ size 75175004
training.log ADDED
The diff for this file is too large to render. See raw diff