shrekrice / MUFF-NER

Multi-feature-fuson-fusion-NER（MUFF）

This project proposes a fused multi-feature NER model for Chinese named entity recognition.

1. Code Description

1.1 Operational environment

Python3.6~3.8
transformers=3.0.0

1.2 Pre-trained models that need to be downloaded in advance

Pre-trained embeddings (word embeddings, character embeddings and double character embeddings) are the same as Lattice LSTM (only Word(Lattice) embeddings are used in this project) 下载地址

Introduction to Word(Lattice) embeddings:

Word(Lattice) embeddings(Just download it directly and place it in the data directory)：ctb.50d.vec

1.3 代码运行

Train the .all file in the Weibo dataset and keep the results in the result file:

!python main.py --resultfile="result/demo.txt"

（The code for training other datasets is similar）