/rasa_nlu

Combine Tecent's bert as service model and rasa_nlu for text classification

Primary LanguagePythonApache License 2.0Apache-2.0

Supply Chain Bot for intent classification

This project include servel trainning data from real projects
including supply chain bot and the high talk bot in shanghai xuhui

Trainning with bert model

First

you should clone the Tecent's bert as service project or pip install it directly

pip install bert-serving-server
pip install bert-serving-client

Second

Download the bert's Chinese model from the following link

bert_Chinese_model

Thirdly

Start the Bert serving to load the high talk data sets for classification

bert-serving-start -model_dir D:\chinese_L-12_H-768_A-12 -num_worker=1

Fourth

train rasa nlu with the bert words vectors

python train.py -c sample_configs/config_embedding_bert_intent_classifier.yml --data data/examples/luis/HighTalkSQSWLuisAppStaging-GA-20180824.json --path projects/bert_gongan_v4