|
--- |
|
license: mit |
|
language: ja |
|
tags: |
|
- luke |
|
- pytorch |
|
- transformers |
|
- ner |
|
- 固有表現抽出 |
|
- named entity recognition |
|
- named-entity-recognition |
|
|
|
--- |
|
|
|
# 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。 |
|
このモデルはluke-japanese-largeを |
|
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。 |
|
|
|
固有表現抽出(NER)タスクに用いることができます。 |
|
|
|
# This model is fine-tuned model for Named-Entity-Recognition(NER) which is based on luke-japanese-large |
|
|
|
This model is fine-tuned by using Wikipedia dataset. |
|
|
|
You could use this model for NER tasks. |
|
|
|
# モデルの精度 accuracy of model |
|
全体:0.8453191098032002 |
|
|
|
||precision|recall|f1-score|support| |
|
|-------------|-----|-----|-----|-----| |
|
|その他の組織名|0.78|0.79|0.79|238| |
|
|イベント名|0.83|0.88| 0.85 | 215| |
|
|人名 | 0.88 | 0.89 | 0.89 | 546| |
|
|地名 |0.83 | 0.85 | 0.84 | 440| |
|
|政治的組織名 | 0.80 | 0.84 | 0.82 | 263| |
|
|施設名 | 0.79 | 0.84 | 0.81 | 241| |
|
|法人名 | 0.88 | 0.89 | 0.89 | 487| |
|
|製品名 | 0.79 | 0.80 | 0.79 | 252| |
|
|micro avg | 0.83 | 0.86 | 0.85 | 2682| |
|
|macro avg | 0.82 | 0.85 | 0.83 | 2682| |
|
|weighted avg | 0.83 | 0.86 | 0.85 | 2682| |
|
|
|
|
|
# How to use 使い方 |
|
sentencepieceとtransformersをインストールして (pip install sentencepiece , pip install transformers) |
|
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。 |
|
please execute this code. |
|
```python |
|
from transformers import MLukeTokenizer,pipeline, LukeForTokenClassification |
|
|
|
tokenizer = MLukeTokenizer.from_pretrained('Mizuiro-sakura/luke-japanese-large-finetuned-ner') |
|
model=LukeForTokenClassification.from_pretrained('Mizuiro-sakura/luke-japanese-large-finetuned-ner') # 学習済みモデルの読み込み |
|
|
|
text=('昨日は東京で買い物をした') |
|
|
|
ner=pipeline('ner', model=model, tokenizer=tokenizer) |
|
|
|
result=ner(text) |
|
print(result) |
|
``` |
|
|
|
|
|
# what is Luke? Lukeとは?[1] |
|
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. |
|
|
|
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。 |
|
|
|
# Acknowledgments 謝辞 |
|
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. |
|
|
|
# Citation |
|
[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} } |
|
|
|
|
|
|
|
|
|
|