Skip to content
达观信息提取比赛第九名代码
Python
Branch: master
Clone or download
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
png
pydatagrand add code Sep 26, 2019
README.md add code Sep 26, 2019
prepare_fold_data.py
prepare_lm_data_mask.py
prepare_lm_data_ngram.py Update prepare_lm_data_ngram.py Oct 6, 2019
run_bert_crf.py
run_bert_span.py Update run_bert_span.py Sep 28, 2019
run_submit.py
train_bert_model.py add code Sep 26, 2019

README.md

datagrand_2019_rank9

2019年达观信息提取比赛第九名代码和答辩ppt

比赛地址:官网

代码目录结构

├── pydatagrand
|  └── callback
|  |  └── lrscheduler.py  
|  |  └── trainingmonitor.py 
|  |  └── ...
|  └── config
|  |  └── basic_config.py #a configuration file for storing model parameters
|  └── dataset   
|  └── io    
|  |  └── dataset.py  
|  |  └── data_transformer.py  
|  └── model
|  |  └── nn 
|  |  └── pretrain 
|  └── output #save the ouput of model
|  └── preprocessing #text preprocessing 
|  └── train #used for training a model
|  |  └── trainer.py 
|  |  └── ...
|  └── common # a set of utility functions
├── prepare_fold_data.py  # 数据切分
├── prepare_lm_data_mask.py  # 随机mask
├── prepare_lm_data_ngram.py  #ngram mask
├── run_bert_crf.py        # crf结构
├── run_bert_span.py   # span结构
├── train_bert_model.py  #训练bert模型

预训练模型

主要训练了8层跟12层BERT模型,采用随机mask + ngram mask两种混合动态masking模式

方案1

方案1主要采用BERT+LSTM+CRF结构

方案2

方案2在方案1的基础上增加了MDP结构

方案3

方案3主要采用BERT+LSTM+SPAN结构

结果

最终结果如下所示:

文档

十强答辩ppt下载地址: https://pan.baidu.com/s/1yvXFf5GzyvDksdBKNp9FKQ 提取码: svr2

You can’t perform that action at this time.