Saved in:
Bibliographic Details
Main Authors: Tsuta, Yuma, Yoshinaga, Naoki, Sato, Shoetsu, Toyoda, Masashi
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2401.02256
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914630024036352
author Tsuta, Yuma
Yoshinaga, Naoki
Sato, Shoetsu
Toyoda, Masashi
author_facet Tsuta, Yuma
Yoshinaga, Naoki
Sato, Shoetsu
Toyoda, Masashi
contents Open-domain dialogue systems have started to engage in continuous conversations with humans. Those dialogue systems are required to be adjusted to the human interlocutor and evaluated in terms of their perspective. However, it is questionable whether the current automatic evaluation methods can approximate the interlocutor's judgments. In this study, we analyzed and examined what features are needed in an automatic response evaluator from the interlocutor's perspective. The first experiment on the Hazumi dataset revealed that interlocutor awareness plays a critical role in making automatic response evaluation correlate with the interlocutor's judgments. The second experiment using massive conversations on X (formerly Twitter) confirmed that dialogue continuity prediction can train an interlocutor-aware response evaluator without human feedback while revealing the difficulty in evaluating generated responses compared to human responses.
format Preprint
id arxiv_https___arxiv_org_abs_2401_02256
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Rethinking Response Evaluation from Interlocutor's Eye for Open-Domain Dialogue Systems
Tsuta, Yuma
Yoshinaga, Naoki
Sato, Shoetsu
Toyoda, Masashi
Computation and Language
Open-domain dialogue systems have started to engage in continuous conversations with humans. Those dialogue systems are required to be adjusted to the human interlocutor and evaluated in terms of their perspective. However, it is questionable whether the current automatic evaluation methods can approximate the interlocutor's judgments. In this study, we analyzed and examined what features are needed in an automatic response evaluator from the interlocutor's perspective. The first experiment on the Hazumi dataset revealed that interlocutor awareness plays a critical role in making automatic response evaluation correlate with the interlocutor's judgments. The second experiment using massive conversations on X (formerly Twitter) confirmed that dialogue continuity prediction can train an interlocutor-aware response evaluator without human feedback while revealing the difficulty in evaluating generated responses compared to human responses.
title Rethinking Response Evaluation from Interlocutor's Eye for Open-Domain Dialogue Systems
topic Computation and Language
url https://arxiv.org/abs/2401.02256