provide a baseline for ccks-2021-task2(address parsing)

1. prepare bert_pretrained model and revised '--model_name_or_path'

2. prepare bigram and char embedding and revised pretrain_unigram_path,pretrain_bigram_path

3. prepare dataset and revised '--data_dir'

4. run on colab (main.ipynb). turn debug on False before run main.

5. run on local device(main.py). turn debug on False before run main.

The main module contains the follow files:

• The load_data.py Text process -> read a file and convert it to a format for model (fastNLP package).

• model.py Build Model -> char,bigram and bert embedding + Bi-LSTM + CRF model and other model can be added.

• pipeline.py contain two classes. Trainer is for training process. Tester is for testing process which contains model predict and evaluation.

• main.py main file to run on local device.

• config.py

• data folder contains files(train.conll,dev.conll,test.conll).

• 1.one can add some trick to import prediction performance. For example model average,Pseudo label,model stacking. Details can be seenBDCI top1 scheme.
• 2.model_name_or_path contains pretrained bert model files (.bin,.json,.txt) which can be downloaded Chinese-BERT-wwm for chinese text (also support other language).
• 3.char and bigram embedding can be downloaded from Flat
• 4.Flat model achieves 87.93, cross validate:88.73, pseudo-labelling: 89.73
• 5.char-bigram-bilstm 86.88
• 6.biaffine-ner, 81.63
• 7.our provided model(char-bigram-bert-bilstm-crf) 90+