@inproceedings{kachuee-lee-2023-constrained,
title = "Constrained Policy Optimization for Controlled Self-Learning in Conversational {AI} Systems",
author = "Kachuee, Mohammad and
Lee, Sungjin",
editor = "Sitaram, Sunayana and
Beigman Klebanov, Beata and
Williams, Jason D",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 5: Industry Track)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-industry.5",
doi = "10.18653/v1/2023.acl-industry.5",
pages = "43--52",
abstract = "Recently, self-learning methods based on user satisfaction metrics and contextual bandits have shown promising results to enable consistent improvements in conversational AI systems. However, directly targeting such metrics by off-policy bandit learning objectives often increases the risk of making abrupt policy changes that break the current user experience. In this study, we introduce a scalable framework for supporting fine-grained exploration targets for individual domains via user-defined constraints. For example, we may want to ensure fewer policy deviations in business-critical domains such as shopping, while allocating more exploration budget to domains such as music. We present a novel meta-gradient learning approach that is scalable and practical to address this problem. The proposed method adjusts constraint violation penalty terms adaptively through a meta objective that encourages balanced constraint satisfaction across domains. We conducted extensive experiments on a real-world conversational AI and using a set of realistic constraint benchmarks. The proposed approach has been deployed in production for a large-scale commercial assistant, enabling the best balance between the policy value and constraint satisfaction rate.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="kachuee-lee-2023-constrained">
<titleInfo>
<title>Constrained Policy Optimization for Controlled Self-Learning in Conversational AI Systems</title>
</titleInfo>
<name type="personal">
<namePart type="given">Mohammad</namePart>
<namePart type="family">Kachuee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sungjin</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2023-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 5: Industry Track)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Sunayana</namePart>
<namePart type="family">Sitaram</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Beata</namePart>
<namePart type="family">Beigman Klebanov</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jason</namePart>
<namePart type="given">D</namePart>
<namePart type="family">Williams</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Toronto, Canada</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Recently, self-learning methods based on user satisfaction metrics and contextual bandits have shown promising results to enable consistent improvements in conversational AI systems. However, directly targeting such metrics by off-policy bandit learning objectives often increases the risk of making abrupt policy changes that break the current user experience. In this study, we introduce a scalable framework for supporting fine-grained exploration targets for individual domains via user-defined constraints. For example, we may want to ensure fewer policy deviations in business-critical domains such as shopping, while allocating more exploration budget to domains such as music. We present a novel meta-gradient learning approach that is scalable and practical to address this problem. The proposed method adjusts constraint violation penalty terms adaptively through a meta objective that encourages balanced constraint satisfaction across domains. We conducted extensive experiments on a real-world conversational AI and using a set of realistic constraint benchmarks. The proposed approach has been deployed in production for a large-scale commercial assistant, enabling the best balance between the policy value and constraint satisfaction rate.</abstract>
<identifier type="citekey">kachuee-lee-2023-constrained</identifier>
<identifier type="doi">10.18653/v1/2023.acl-industry.5</identifier>
<location>
<url>https://aclanthology.org/2023.acl-industry.5</url>
</location>
<part>
<date>2023-07</date>
<extent unit="page">
<start>43</start>
<end>52</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Constrained Policy Optimization for Controlled Self-Learning in Conversational AI Systems
%A Kachuee, Mohammad
%A Lee, Sungjin
%Y Sitaram, Sunayana
%Y Beigman Klebanov, Beata
%Y Williams, Jason D.
%S Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 5: Industry Track)
%D 2023
%8 July
%I Association for Computational Linguistics
%C Toronto, Canada
%F kachuee-lee-2023-constrained
%X Recently, self-learning methods based on user satisfaction metrics and contextual bandits have shown promising results to enable consistent improvements in conversational AI systems. However, directly targeting such metrics by off-policy bandit learning objectives often increases the risk of making abrupt policy changes that break the current user experience. In this study, we introduce a scalable framework for supporting fine-grained exploration targets for individual domains via user-defined constraints. For example, we may want to ensure fewer policy deviations in business-critical domains such as shopping, while allocating more exploration budget to domains such as music. We present a novel meta-gradient learning approach that is scalable and practical to address this problem. The proposed method adjusts constraint violation penalty terms adaptively through a meta objective that encourages balanced constraint satisfaction across domains. We conducted extensive experiments on a real-world conversational AI and using a set of realistic constraint benchmarks. The proposed approach has been deployed in production for a large-scale commercial assistant, enabling the best balance between the policy value and constraint satisfaction rate.
%R 10.18653/v1/2023.acl-industry.5
%U https://aclanthology.org/2023.acl-industry.5
%U https://doi.org/10.18653/v1/2023.acl-industry.5
%P 43-52
Markdown (Informal)
[Constrained Policy Optimization for Controlled Self-Learning in Conversational AI Systems](https://aclanthology.org/2023.acl-industry.5) (Kachuee & Lee, ACL 2023)
ACL