@inproceedings{lubis-etal-2022-dialogue,
title = "Dialogue Evaluation with Offline Reinforcement Learning",
author = "Lubis, Nurul and
Geishauser, Christian and
Lin, Hsien-chin and
van Niekerk, Carel and
Heck, Michael and
Feng, Shutong and
Gasic, Milica",
editor = "Lemon, Oliver and
Hakkani-Tur, Dilek and
Li, Junyi Jessy and
Ashrafzadeh, Arash and
Garcia, Daniel Hern{\'a}ndez and
Alikhani, Malihe and
Vandyke, David and
Du{\v{s}}ek, Ond{\v{r}}ej",
booktitle = "Proceedings of the 23rd Annual Meeting of the Special Interest Group on Discourse and Dialogue",
month = sep,
year = "2022",
address = "Edinburgh, UK",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2022.sigdial-1.46",
doi = "10.18653/v1/2022.sigdial-1.46",
pages = "478--489",
abstract = "Task-oriented dialogue systems aim to fulfill user goals through natural language interactions. They are ideally evaluated with human users, which however is unattainable to do at every iteration of the development phase. Simulated users could be an alternative, however their development is nontrivial. Therefore, researchers resort to offline metrics on existing human-human corpora, which are more practical and easily reproducible. They are unfortunately limited in reflecting real performance of dialogue systems. BLEU for instance is poorly correlated with human judgment, and existing corpus-based metrics such as success rate overlook dialogue context mismatches. There is still a need for a reliable metric for task-oriented systems with good generalization and strong correlation with human judgements. In this paper, we propose the use of offline reinforcement learning for dialogue evaluation based on static data. Such an evaluator is typically called a critic and utilized for policy optimization. We go one step further and show that offline RL critics can be trained for any dialogue system as external evaluators, allowing dialogue performance comparisons across various types of systems. This approach has the benefit of being corpus- and model-independent, while attaining strong correlation with human judgements, which we confirm via an interactive user trial.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="lubis-etal-2022-dialogue">
<titleInfo>
<title>Dialogue Evaluation with Offline Reinforcement Learning</title>
</titleInfo>
<name type="personal">
<namePart type="given">Nurul</namePart>
<namePart type="family">Lubis</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Christian</namePart>
<namePart type="family">Geishauser</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hsien-chin</namePart>
<namePart type="family">Lin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Carel</namePart>
<namePart type="family">van Niekerk</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Michael</namePart>
<namePart type="family">Heck</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shutong</namePart>
<namePart type="family">Feng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Milica</namePart>
<namePart type="family">Gasic</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2022-09</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 23rd Annual Meeting of the Special Interest Group on Discourse and Dialogue</title>
</titleInfo>
<name type="personal">
<namePart type="given">Oliver</namePart>
<namePart type="family">Lemon</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Dilek</namePart>
<namePart type="family">Hakkani-Tur</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Junyi</namePart>
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Arash</namePart>
<namePart type="family">Ashrafzadeh</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Daniel</namePart>
<namePart type="given">Hernández</namePart>
<namePart type="family">Garcia</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Malihe</namePart>
<namePart type="family">Alikhani</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">David</namePart>
<namePart type="family">Vandyke</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ondřej</namePart>
<namePart type="family">Dušek</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Edinburgh, UK</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Task-oriented dialogue systems aim to fulfill user goals through natural language interactions. They are ideally evaluated with human users, which however is unattainable to do at every iteration of the development phase. Simulated users could be an alternative, however their development is nontrivial. Therefore, researchers resort to offline metrics on existing human-human corpora, which are more practical and easily reproducible. They are unfortunately limited in reflecting real performance of dialogue systems. BLEU for instance is poorly correlated with human judgment, and existing corpus-based metrics such as success rate overlook dialogue context mismatches. There is still a need for a reliable metric for task-oriented systems with good generalization and strong correlation with human judgements. In this paper, we propose the use of offline reinforcement learning for dialogue evaluation based on static data. Such an evaluator is typically called a critic and utilized for policy optimization. We go one step further and show that offline RL critics can be trained for any dialogue system as external evaluators, allowing dialogue performance comparisons across various types of systems. This approach has the benefit of being corpus- and model-independent, while attaining strong correlation with human judgements, which we confirm via an interactive user trial.</abstract>
<identifier type="citekey">lubis-etal-2022-dialogue</identifier>
<identifier type="doi">10.18653/v1/2022.sigdial-1.46</identifier>
<location>
<url>https://aclanthology.org/2022.sigdial-1.46</url>
</location>
<part>
<date>2022-09</date>
<extent unit="page">
<start>478</start>
<end>489</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Dialogue Evaluation with Offline Reinforcement Learning
%A Lubis, Nurul
%A Geishauser, Christian
%A Lin, Hsien-chin
%A van Niekerk, Carel
%A Heck, Michael
%A Feng, Shutong
%A Gasic, Milica
%Y Lemon, Oliver
%Y Hakkani-Tur, Dilek
%Y Li, Junyi Jessy
%Y Ashrafzadeh, Arash
%Y Garcia, Daniel Hernández
%Y Alikhani, Malihe
%Y Vandyke, David
%Y Dušek, Ondřej
%S Proceedings of the 23rd Annual Meeting of the Special Interest Group on Discourse and Dialogue
%D 2022
%8 September
%I Association for Computational Linguistics
%C Edinburgh, UK
%F lubis-etal-2022-dialogue
%X Task-oriented dialogue systems aim to fulfill user goals through natural language interactions. They are ideally evaluated with human users, which however is unattainable to do at every iteration of the development phase. Simulated users could be an alternative, however their development is nontrivial. Therefore, researchers resort to offline metrics on existing human-human corpora, which are more practical and easily reproducible. They are unfortunately limited in reflecting real performance of dialogue systems. BLEU for instance is poorly correlated with human judgment, and existing corpus-based metrics such as success rate overlook dialogue context mismatches. There is still a need for a reliable metric for task-oriented systems with good generalization and strong correlation with human judgements. In this paper, we propose the use of offline reinforcement learning for dialogue evaluation based on static data. Such an evaluator is typically called a critic and utilized for policy optimization. We go one step further and show that offline RL critics can be trained for any dialogue system as external evaluators, allowing dialogue performance comparisons across various types of systems. This approach has the benefit of being corpus- and model-independent, while attaining strong correlation with human judgements, which we confirm via an interactive user trial.
%R 10.18653/v1/2022.sigdial-1.46
%U https://aclanthology.org/2022.sigdial-1.46
%U https://doi.org/10.18653/v1/2022.sigdial-1.46
%P 478-489
Markdown (Informal)
[Dialogue Evaluation with Offline Reinforcement Learning](https://aclanthology.org/2022.sigdial-1.46) (Lubis et al., SIGDIAL 2022)
ACL
- Nurul Lubis, Christian Geishauser, Hsien-chin Lin, Carel van Niekerk, Michael Heck, Shutong Feng, and Milica Gasic. 2022. Dialogue Evaluation with Offline Reinforcement Learning. In Proceedings of the 23rd Annual Meeting of the Special Interest Group on Discourse and Dialogue, pages 478–489, Edinburgh, UK. Association for Computational Linguistics.