Proposal of Abstractive Dialogue Format Summary and Analysis of its Characteristics and Workload on Dialogue Handover
Journal of Natural Language Processing · 2026 · 33 (1) · 336-369
DOI: 10.5715/jnlp.33.336 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Natural Language Processing, 2026. Verfasst von Sanae Yamashita und 1 weiteren Autor:innen. Die Arbeit nennt 80 Literaturangaben.
Themen: Übergabe
Abstract (Originalsprache)
Dialogue summarization is considered useful not only for understanding the dialogue content but also for taking over an ongoing dialogue, that is, for joining the dialogue to accomplish its original objective. Because the flow and nuances of previous exchanges are important for dialogue handover, we propose a dialogue format summary (DFS) that summarizes a dialogue into a shorter dialogue. In this study, we first constructed evaluation axes for dialogue summarization and clarified the characteristics of DFS by comparing multiple types of dialogue summaries on these evaluation axes. We also conducted dialogue handover experiments to simulate dialogue handover for multiple types of dialogue summaries. The results revealed that DFS has a low handover workload and is useful for creating the next utterance. Thus, we confirmed that DFS is suitable for dialogue handover.
Zitieren
APA
Yamashita, S., & Higashinaka, R. (2026). Proposal of Abstractive Dialogue Format Summary and Analysis of its Characteristics and Workload on Dialogue Handover. Journal of Natural Language Processing, 33(1), 336-369. https://doi.org/10.5715/jnlp.33.336
Vancouver
Yamashita S, Higashinaka R. Proposal of Abstractive Dialogue Format Summary and Analysis of its Characteristics and Workload on Dialogue Handover. Journal of Natural Language Processing. 2026;33(1):336-369. doi:10.5715/jnlp.33.336
BibTeX
@article{Yamashita2026,
author = {Yamashita, Sanae and Higashinaka, Ryuichiro},
title = {Proposal of Abstractive Dialogue Format Summary and Analysis of its Characteristics and Workload on Dialogue Handover},
journal = {Journal of Natural Language Processing},
year = {2026},
volume = {33},
number = {1},
pages = {336--369},
doi = {10.5715/jnlp.33.336}
}
EUVIMED