PooryaPiroozfar commited on
Commit
cc99a14
1 Parent(s): 443d30c

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +84 -0
README.md ADDED
@@ -0,0 +1,84 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ tags:
3
+ - flair
4
+ - token-classification
5
+ - sequence-tagger-model
6
+ language: fa
7
+ dataset:
8
+ - NSURL-2019
9
+ widget:
10
+ - text: >-
11
+ کارنامه نشر، وابسته به موسسه خانه کتاب و زیر نظر احمد مسجدی جامعی معاون امور
12
+ فرهنگی وزارت فرهنگ و ارشاد اسلامی است.
13
+ metrics:
14
+ - f1
15
+ ---
16
+
17
+ ## Persian NER Using Flair
18
+
19
+ This is the 7-class Named-entity recognition model for Persian that ships with [Flair](https://github.com/flairNLP/flair/).
20
+
21
+ F1-Score: **90.33** (NSURL-2019)
22
+
23
+ Predicts NER tags:
24
+
25
+ | **tag** | **meaning** |
26
+ |:---------------------------------:|:-----------:|
27
+ | PER | person name |
28
+ | LOC | location name |
29
+ | ORG | organization name |
30
+ | DAT | date |
31
+ | TIM | time |
32
+ | PCT | percent |
33
+ | MON | Money|
34
+
35
+ Based on [Flair embeddings](https://www.aclweb.org/anthology/C18-1139/) and Pars-Bert.
36
+
37
+ ---
38
+
39
+ ### Demo: How to use in Flair
40
+
41
+ Requires: **[Flair](https://github.com/flairNLP/flair/)** (`pip install flair`)
42
+
43
+ ```python
44
+ from flair.data import Sentence
45
+ from flair.models import SequenceTagger
46
+ # load tagger
47
+ tagger = SequenceTagger.load("PooryaPiroozfar/Flair_Persian_NER")
48
+ # make example sentence
49
+ sentence = Sentence("کارنامه نشر، وابسته به موسسه خانه کتاب و زیر نظر احمد مسجدی جامعی معاون امور فرهنگی وزارت فرهنگ و ارشاد اسلامی است.")
50
+ tagger.predict(sentence)
51
+ #print result
52
+ print(sentence.to_tagged_string())
53
+ ```
54
+
55
+ This yields the following output:
56
+ ```
57
+
58
+ ```
59
+
60
+ ---
61
+
62
+ ### Results
63
+ - F-score (micro) 0.9033
64
+ - F-score (macro) 0.8976
65
+ - Accuracy 0.8277
66
+
67
+ ```
68
+ By class:
69
+ precision recall f1-score support
70
+
71
+ ORG 0.9016 0.8667 0.8838 1523
72
+ LOC 0.9113 0.9305 0.9208 1425
73
+ PER 0.9216 0.9322 0.9269 1224
74
+ DAT 0.8623 0.7958 0.8277 480
75
+ MON 0.9665 0.9558 0.9611 181
76
+ PCT 0.9375 0.9740 0.9554 77
77
+ TIM 0.8235 0.7925 0.8077 53
78
+
79
+ micro avg 0.9081 0.8984 0.9033 4963
80
+ macro avg 0.9035 0.8925 0.8976 4963
81
+ weighted avg 0.9076 0.8984 0.9028 4963
82
+ samples avg 0.8277 0.8277 0.8277 4963
83
+
84
+ ```