Efficient Data Generation for Source-grounded Information-seeking Dialogs: A Use Case for Meeting Transcripts

TL;DR

This paper presents a semi-automatic method for generating source-grounded information-seeking dialog data from meeting transcripts using LLMs, resulting in the MISeD dataset and improved model attribution and response quality.

Contribution

It introduces a semi-automatic data generation approach for complex dialogs, creating the MISeD dataset and demonstrating enhanced model performance.

Findings

Models fine-tuned on MISeD outperform off-the-shelf models.

Fine-tuning on MISeD achieves similar response quality to manual data.

Attribution quality improves with the MISeD dataset.

Abstract

Automating data generation with Large Language Models (LLMs) has become increasingly popular. In this work, we investigate the feasibility and effectiveness of LLM-based data generation in the challenging setting of source-grounded information-seeking dialogs, with response attribution, over long documents. Our source texts consist of long and noisy meeting transcripts, adding to the task complexity. Since automating attribution remains difficult, we propose a semi-automatic approach: dialog queries and responses are generated with LLMs, followed by human verification and identification of attribution spans. Using this approach, we created MISeD -- Meeting Information Seeking Dialogs dataset -- a dataset of information-seeking dialogs focused on meeting transcripts. Models finetuned with MISeD demonstrate superior performance compared to off-the-shelf models, even those of larger size. Finetuning on MISeD gives comparable response generation quality to finetuning on fully manual data, while improving attribution quality and reducing time and effort.