A comprehensive solution to retrieval-based chatbot construction

TL;DR

This paper introduces an end-to-end framework for building retrieval-based chatbots using self-supervised contrastive learning, demonstrating its effectiveness in real-world customer support scenarios.

Contribution

It provides a comprehensive pipeline from unlabelled chatlogs to deployed chatbot, including dataset creation, response selection, and architecture design, with a focus on contrastive learning.

Findings

Self-supervised contrastive learning outperforms other models in response selection.

Hierarchical RNN architecture meets deployment speed requirements.

Effective use of chatlogs for dataset creation and training.

Abstract

In this paper we present the results of our experiments in training and deploying a self-supervised retrieval-based chatbot trained with contrastive learning for assisting customer support agents. In contrast to most existing research papers in this area where the focus is on solving just one component of a deployable chatbot, we present an end-to-end set of solutions to take the reader from an unlabelled chatlogs to a deployed chatbot. This set of solutions includes creating a self-supervised dataset and a weakly labelled dataset from chatlogs, as well as a systematic approach to selecting a fixed list of canned responses. We present a hierarchical-based RNN architecture for the response selection model, chosen for its ability to cache intermediate utterance embeddings, which helped to meet deployment inference speed requirements. We compare the performance of this architecture across 3 different learning objectives: self-supervised contrastive learning, binary classification, and multi-class classification. We find that using a self-supervised contrastive learning model outperforms training the binary and multi-class classification models on a weakly labelled dataset. Our results validate that the self-supervised contrastive learning approach can be effectively used for a real-world chatbot scenario.