Mizuiro-sakura/luke-japanese-base-marcja

このモデルはluke-japanese-baseをファインチューニングして、MARC-ja(positive or negativeの二値分類)に用いれるようにしたものです。

このモデルはluke-japanese-baseを yahoo japan/JGLUEのMARC-ja( https://github.com/yahoojapan/JGLUE ) を用いてファインチューニングしたものです。

positive or negativeの二値分類タスクに用いることができます。

This model is fine-tuned model for MARC-ja which is based on luke-japanese-base

This model is fine-tuned by using yahoo japan JGLUE MARC-ja dataset.

You could use this model for binary classification (positive or negative) tasks.

モデルの精度 accuracy of model

precision : 0.967, accuracy : 0.967, recall : 0.967, f1 : 0.967

既存のモデルの精度　accuracy of existing model

Tohoku BERT large 0.955

Waseda RoBERTa large (seq128) 0.954

Waseda RoBERTa large (seq512) 0.961

XLM RoBERTa large 0.964

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、MARC-jaタスクを解かせることができます。 After install transformers and sentencepiec, please execute this code.

from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

tokenizer = AutoTokenizer.from_pretrained('Mizuiro-sakura/luke-japanese-base-marcja')
model = AutoModelForSequenceClassification.from_pretrained('Mizuiro-sakura/luke-japanese-base-marcja')

text = 'この商品は素晴らしい！とても匂いが良く、満足でした。'

token = tokenizer.encode_plus(text, truncation=True, max_length=128, padding="max_length", return_tenosr='pt')
result = model(torch.tensor(token['input_ids']).unsqueeze(0), torch.tensor(token['attention_mask']).unsqueeze(0))

if torch.argmax(result['logits'])==0:
    print('positive')
if torch.argmax(result['logits'])==1:
    print('negative')

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} }