C-BiLSTM

C-BiLSTM

This code is the implementation for C-BiLSTM model.

Requirements

--python==3.8.10

--torch==1.7.0

Usage

Download character embedding

The Chinese character embeddings token_vec_300.bin can be obtain from here. Download and put token_vec_300.bin in directory pretrained.

Run pre_embedding.py to obtain char2vec_file.mat.npy and word2id.npy and put them in directory pretrained for the following training.

├── pretrained
│     └── token_vec_300.bin
│     └── char2vec_file.mat.npy
│     └── word2id.npy

Data preprocess

Put the processed sentence-level results in data directory.

├── data
│     └── train_sentence.json
│     └── dev_sentence.json
│     └── test_sentence.json

Train and test

Run run.py and set paramater --train is True when train and False when test.