Towards Multi-Level Transcript Segmentation: LoRA Fine-Tuning for Table-of-Contents Generation

TL;DR

This paper presents a hierarchical topic segmentation method for speech transcripts, using LoRA fine-tuning and speech pause features to generate multi-level tables of contents, improving over existing baselines.

Contribution

It introduces a novel hierarchical segmentation approach with LoRA fine-tuning and multi-level evaluation, advancing transcript organization techniques.

Findings

Significant improvements over baseline segmentation methods.

Effective multi-language and multi-level transcript segmentation.

Enhanced evaluation metric for hierarchical segmentation.

Abstract

Segmenting speech transcripts into thematic sections benefits both downstream processing and users who depend on written text for accessibility. We introduce a novel approach to hierarchical topic segmentation in transcripts, generating multi-level tables of contents that capture both topic and subtopic boundaries. We compare zero-shot prompting and LoRA fine-tuning on large language models, while also exploring the integration of high-level speech pause features. Evaluations on English meeting recordings and multilingual lecture transcripts (Portuguese, German) show significant improvements over established topic segmentation baselines. Additionally, we adapt a common evaluation measure for multi-level segmentation, taking into account all hierarchical levels within one metric.