Commit
·
7e1ea44
1
Parent(s):
a4c6768
Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,77 @@
|
|
1 |
---
|
2 |
license: mit
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: mit
|
3 |
+
language: ja
|
4 |
+
tags:
|
5 |
+
- luke
|
6 |
+
- pytorch
|
7 |
+
- transformers
|
8 |
+
- ner
|
9 |
+
- 固有表現抽出
|
10 |
+
- named entity recognition
|
11 |
+
- named-entity-recognition
|
12 |
+
|
13 |
---
|
14 |
+
|
15 |
+
# このモデルはluke-japanese-largeをファインチューニングして、固有表現抽出(NER)に用いれるようにしたものです。
|
16 |
+
このモデルはluke-japanese-largeを
|
17 |
+
Wikipediaを用いた日本語の固有表現抽出データセット(ストックマーク社、https://github.com/stockmarkteam/ner-wikipedia-dataset )を用いてファインチューニングしたものです。
|
18 |
+
|
19 |
+
固有表現抽出(NER)タスクに用いることができます。
|
20 |
+
|
21 |
+
# This model is fine-tuned model for Named-Entity-Recognition(NER) which is based on luke-japanese-large
|
22 |
+
|
23 |
+
This model is fine-tuned by using Wikipedia dataset.
|
24 |
+
|
25 |
+
You could use this model for NER tasks.
|
26 |
+
|
27 |
+
# モデルの精度 accuracy of model
|
28 |
+
全体:0.8453191098032002
|
29 |
+
precision recall f1-score support
|
30 |
+
|
31 |
+
その他の組織名 0.78 0.79 0.79 238
|
32 |
+
イベント名 0.83 0.88 0.85 215
|
33 |
+
人名 0.88 0.89 0.89 546
|
34 |
+
地名 0.83 0.85 0.84 440
|
35 |
+
政治的組織名 0.80 0.84 0.82 263
|
36 |
+
施設名 0.79 0.84 0.81 241
|
37 |
+
法人名 0.88 0.89 0.89 487
|
38 |
+
製品名 0.79 0.80 0.79 252
|
39 |
+
|
40 |
+
micro avg 0.83 0.86 0.85 2682
|
41 |
+
macro avg 0.82 0.85 0.83 2682
|
42 |
+
weighted avg 0.83 0.86 0.85 2682
|
43 |
+
|
44 |
+
|
45 |
+
# How to use 使い方
|
46 |
+
sentencepieceとtransformersをインストールして (pip install sentencepiece , pip install transformers)
|
47 |
+
以下のコードを実行することで、NERタスクを解かせることができます。
|
48 |
+
please execute this code.
|
49 |
+
```python
|
50 |
+
from transformers import MLukeTokenizer,pipeline, LukeForTokenClassification
|
51 |
+
|
52 |
+
tokenizer = MLukeTokenizer.from_pretrained('Mizuiro-sakura/luke-japanese-large-finetuned-ner')
|
53 |
+
model=LukeForTokenClassification.from_pretrained('Mizuiro-sakura/luke-japanese-large-finetuned-ner') # 学習済みモデルの読み込み
|
54 |
+
|
55 |
+
text=('昨日は東京で買い物をした')
|
56 |
+
|
57 |
+
ner=pipeline('ner', model=model, tokenizer=tokenizer)
|
58 |
+
|
59 |
+
result=ner(text)
|
60 |
+
print(result)
|
61 |
+
```
|
62 |
+
|
63 |
+
|
64 |
+
# what is Luke? Lukeとは?[1]
|
65 |
+
LUKE (Language Understanding with Knowledge-based Embeddings) is a new pre-trained contextualized representation of words and entities based on transformer. LUKE treats words and entities in a given text as independent tokens, and outputs contextualized representations of them. LUKE adopts an entity-aware self-attention mechanism that is an extension of the self-attention mechanism of the transformer, and considers the types of tokens (words or entities) when computing attention scores.
|
66 |
+
|
67 |
+
LUKE achieves state-of-the-art results on five popular NLP benchmarks including SQuAD v1.1 (extractive question answering), CoNLL-2003 (named entity recognition), ReCoRD (cloze-style question answering), TACRED (relation classification), and Open Entity (entity typing). luke-japaneseは、単語とエンティティの知識拡張型訓練済み Transformer モデルLUKEの日本語版です。LUKE は単語とエンティティを独立したトークンとして扱い、これらの文脈を考慮した表現を出力します。
|
68 |
+
|
69 |
+
# Acknowledgments 謝辞
|
70 |
+
Lukeの開発者である山田先生とStudio ousiaさんには感謝いたします。 I would like to thank Mr.Yamada @ikuyamada and Studio ousia @StudioOusia.
|
71 |
+
|
72 |
+
# Citation
|
73 |
+
[1]@inproceedings{yamada2020luke, title={LUKE: Deep Contextualized Entity Representations with Entity-aware Self-attention}, author={Ikuya Yamada and Akari Asai and Hiroyuki Shindo and Hideaki Takeda and Yuji Matsumoto}, booktitle={EMNLP}, year={2020} }
|
74 |
+
|
75 |
+
|
76 |
+
|
77 |
+
|