Several evaluation metrics have been proposed for topic segmentation.
Most of them rely on the paradigm that segmentation is mainly a task
that detects boundaries, and thus are oriented on boundary detection
evaluation. Nevertheless, this paradigm is not appropriate to get homogeneous
chapters, which is one of the major applications of topic segmentation.
For instance on Broadcast News, topic segmentation enables users to
watch a chapter independently of the others.
We propose to consider
segmentation as a task that detects homogeneous segments, and we propose
evaluation metrics oriented on segment retrieval. The proposed metrics
are experimented on various TV shows from different channels. Results
are analysed and discussed, highlighting their relevance.
Cite as: Bouchekif, A., Charlet, D., Damnati, G., Camelin, N., Estève, Y. (2017) Evaluating Automatic Topic Segmentation as a Segment Retrieval Task. Proc. Interspeech 2017, 2924-2928, doi: 10.21437/Interspeech.2017-1231
@inproceedings{bouchekif17_interspeech,
author={Abdessalam Bouchekif and Delphine Charlet and Géraldine Damnati and Nathalie Camelin and Yannick Estève},
title={{Evaluating Automatic Topic Segmentation as a Segment Retrieval Task}},
year=2017,
booktitle={Proc. Interspeech 2017},
pages={2924--2928},
doi={10.21437/Interspeech.2017-1231}
}