Counseling Chatbot with Wellness Dataset

Award

1st prize in AI for Future Society 2022

How to Use

Chatbot Model

1. Run preprocess.py

2. Run build_graph.py

3. Run train_bert.py

4. Run roberta_wellness_chatbot.py

APP

1. Build ./counselingApp (have to check your server ip)

2. After entering the message, press the send button and the chatbot answers.

Skill Stack

• Firebase
• HTTP connect
• Asychronous
• Recyclerview

Requirements

• dgl-cu113 == 0.9.1.post1
• ignite == 1.1.0
• python == 3.6.9
• torch == 1.10.0+cu113
• scikit-learn== 0.24.2
• transformers == 4.18.0
• numpy =< 1.19.5
• networkx == 2.5.1

Future work

• Make graph for hierarchical class
• Auto build graph for unseem data ( Now, Just Replace Embedding of last node. So, prediction task is abnormal )

Reference

Dataset

• 웰니스 대화 스크립트 데이터셋

Backbone Model

• klue/roberta-base

Citation

• BertGCN: Transductive Text Classification by Combining GCN and BERT

Workers

노영준

• Flask API
• Change build graph method
• Preprocessing
• Finetune KLUE/roberta-base

이가은

• Develop Application

박정원

• Finetune RoBERTaGCN