Skip to content

Read-only version: For security reasons, the Research Information System (FIS) has been temporarily replaced by this read-only version. All public content and full texts remain available; login, editing, submission and statistics are not available. Questions: fis@uni-bamberg.de

Home › Publications › Contrasting Traditional Models and LLMs : An Evaluation Based on Text Segmentation

Publication

Contrasting Traditional Models and LLMs : An Evaluation Based on Text Segmentation

Jegan, Robin; Henrich, Andreas (2025): Contrasting Traditional Models and LLMs : An Evaluation Based on Text Segmentation, in: Christian Wartena, Ulrich Heid, und Christian Wartena (Hrsg.), Proceedings of the 21st Conference on Natural Language Processing (KONVENS 2025) : Workshops, Hannover: HsH Applied Academics, S. 274–281.

Details

Faculty/Chair
Faculty/Professorship
Author
Jegan, Robin
Title of the compilation
Proceedings of the 21st Conference on Natural Language Processing (KONVENS 2025) : Workshops
Editors
Wartena, Christian
Heid, Ulrich
Conference
21st Conference on Natural Language Processing (KONVENS 2025), Hannover
Publisher Information
HsH Applied Academics : Hannover
Year of publication
2025
Language
English
URL
Abstract

This paper presents a discussion of the relevancy of older natural language processing approaches compared to modern large language models (LLMs), with experimental results for a specific application: the segmentation of video transcripts. An analysis was conducted, if powerful modern LLMs are necessary for tasks such as text segmentation or if traditional and more efficient models – here TextTiling – suffice. In the end, LLMs provide comparable performance to the other models, but the results produced by TextTiling are promising and suggest a discussion about a trade-off regarding efficiency, performance, energy-consumption and other factors.

Keywords
-
Type
Conferenceobject
URI
Activation date
2025-11-12

Permalink: https://fis.uni-bamberg.de/handle/uniba/111250 · Data as JSON · Cite: BibTeX · RIS

Data as of: 2026-11-16

Restricted read-only access: login, editing and submission are temporarily disabled.