Cross-Lingual Conversational Speech Summarization with Large Language Models

TL;DR

This paper addresses cross-lingual conversational speech summarization by augmenting existing datasets with GPT-4 generated summaries, analyzing the impact of transcription and translation errors, and adapting large language models for improved performance.

Contribution

It introduces a new dataset with summaries for Spanish-English speech translation, and demonstrates effective adaptation of LLMs like Mistral-7B for this task.

Findings

Mistral-7B outperforms off-the-shelf models

GPT-4 generated summaries serve as effective ground truth

Transcription and translation errors significantly impact summarization quality

Abstract

Cross-lingual conversational speech summarization is an important problem, but suffers from a dearth of resources. While transcriptions exist for a number of languages, translated conversational speech is rare and datasets containing summaries are non-existent. We build upon the existing Fisher and Callhome Spanish-English Speech Translation corpus by supplementing the translations with summaries. The summaries are generated using GPT-4 from the reference translations and are treated as ground truth. The task is to generate similar summaries in the presence of transcription and translation errors. We build a baseline cascade-based system using open-source speech recognition and machine translation models. We test a range of LLMs for summarization and analyze the impact of transcription and translation errors. Adapting the Mistral-7B model for this task performs significantly better than off-the-shelf models and matches the performance of GPT-4.