alanakbik commited on
Commit
d33b2e0
1 Parent(s): 25200da

initial model commit

Browse files
Files changed (1) hide show
  1. README.md +18 -7
README.md CHANGED
@@ -11,9 +11,9 @@ inference: false
11
 
12
  ## 4-Language NER in Flair (English, German, Dutch and Spanish)
13
 
14
- This is the standard 4-class NER model for 4 CoNLL-03 languages that ships with [Flair](https://github.com/flairNLP/flair/). Also kind of works for related languages like French.
15
 
16
- F1-Score: **92,16** (CoNLL-03 English), **87,33** (CoNLL-03 German revised), **88,96** (CoNLL-03 Dutch), **86,65** (CoNLL-03 Spanish)
17
 
18
 
19
  Predicts 4 tags:
@@ -38,7 +38,7 @@ from flair.data import Sentence
38
  from flair.models import SequenceTagger
39
 
40
  # load tagger
41
- tagger = SequenceTagger.load("flair/ner-multi")
42
 
43
  # make example sentence in any of the four languages
44
  sentence = Sentence("George Washington ging nach Washington")
@@ -101,10 +101,10 @@ embedding_types = [
101
  WordEmbeddings('de'),
102
 
103
  # contextual string embeddings, forward
104
- FlairEmbeddings('multi-forward'),
105
 
106
  # contextual string embeddings, backward
107
- FlairEmbeddings('multi-backward'),
108
  ]
109
 
110
  # embedding stack consists of Flair and GloVe embeddings
@@ -124,7 +124,7 @@ from flair.trainers import ModelTrainer
124
  trainer = ModelTrainer(tagger, corpus)
125
 
126
  # 7. run training
127
- trainer.train('resources/taggers/ner-multi',
128
  train_with_dev=True,
129
  max_epochs=150)
130
  ```
@@ -135,7 +135,18 @@ trainer.train('resources/taggers/ner-multi',
135
 
136
  ### Cite
137
 
138
- Please cite the following paper when using this model.
 
 
 
 
 
 
 
 
 
 
 
139
 
140
  ```
141
  @inproceedings{akbik2018coling,
 
11
 
12
  ## 4-Language NER in Flair (English, German, Dutch and Spanish)
13
 
14
+ This is the fast 4-class NER model for 4 CoNLL-03 languages that ships with [Flair](https://github.com/flairNLP/flair/). Also kind of works for related languages like French.
15
 
16
+ F1-Score: **91,51** (CoNLL-03 English), **85,72** (CoNLL-03 German revised), **86,22** (CoNLL-03 Dutch), **85,78** (CoNLL-03 Spanish)
17
 
18
 
19
  Predicts 4 tags:
 
38
  from flair.models import SequenceTagger
39
 
40
  # load tagger
41
+ tagger = SequenceTagger.load("flair/ner-multi-fast")
42
 
43
  # make example sentence in any of the four languages
44
  sentence = Sentence("George Washington ging nach Washington")
 
101
  WordEmbeddings('de'),
102
 
103
  # contextual string embeddings, forward
104
+ FlairEmbeddings('multi-forward-fast'),
105
 
106
  # contextual string embeddings, backward
107
+ FlairEmbeddings('multi-backward-fast'),
108
  ]
109
 
110
  # embedding stack consists of Flair and GloVe embeddings
 
124
  trainer = ModelTrainer(tagger, corpus)
125
 
126
  # 7. run training
127
+ trainer.train('resources/taggers/ner-multi-fast',
128
  train_with_dev=True,
129
  max_epochs=150)
130
  ```
 
135
 
136
  ### Cite
137
 
138
+ Please cite the following papers when using this model.
139
+
140
+
141
+ ```
142
+ @misc{akbik2019multilingual,
143
+ title={Multilingual sequence labeling with one model},
144
+ author={Akbik, Alan and Bergmann, Tanja and Vollgraf, Roland}
145
+ booktitle = {{NLDL} 2019, Northern Lights Deep Learning Workshop},
146
+ year = {2019}
147
+ }
148
+ ```
149
+
150
 
151
  ```
152
  @inproceedings{akbik2018coling,