README.md

Ensemble Classification Demo

Implemented TextCNNs,BERT as well as the weighted vote ensemble method.

• Run entry: src/main/java/EnsembleDemo.java

This repository contains the whole inference process of both BERT and TextCNNs, including preprocessing and prediction.

Preparation

1. Firstly, you must have exported either BERT/TextCNNs models into .pb format. For BERT models, you can find the solution to exporting BERT .pb model by checking https://github.com/cyk1337/BERT-Classification/blob/master/bert/run_classification.py.
2. Then, use the .pb model path to initialize the model class at

   BERT-Text-Classification-J/src/main/java/EnsembleDemo.java

   Line 38 in 18c6444

   String bert_model_path = "pb_model/bert_L4_FC3_Seq128.pb";

   , suppose you have generated .pb model correctly.