An End-to-End Dialogue Summarization System for Sales Calls

TL;DR

This paper introduces a production system for sales call summarization that combines fine-tuned generative models with human-in-the-loop curation, addressing real-world challenges like long dialogues and data scarcity.

Contribution

It presents a novel end-to-end dialogue summarization system that leverages GPT-3 for data labeling and integrates human feedback for improved quality.

Findings

Significant improvements in summarization accuracy on public datasets

Effective handling of long dialogues and content validation

Successful use of GPT-3 for data labeling in an industrial setting

Abstract

Summarizing sales calls is a routine task performed manually by salespeople. We present a production system which combines generative models fine-tuned for customer-agent setting, with a human-in-the-loop user experience for an interactive summary curation process. We address challenging aspects of dialogue summarization task in a real-world setting including long input dialogues, content validation, lack of labeled data and quality evaluation. We show how GPT-3 can be leveraged as an offline data labeler to handle training data scarcity and accommodate privacy constraints in an industrial setting. Experiments show significant improvements by our models in tackling the summarization and content validation tasks on public datasets.