An empirical bayes framework for open-domain dialogue generation
To engage human users in meaningful conversation, open-domain dialogue agents are required to generate diverse and contextually coherent dialogue. Despite recent advancements, which can be attributed to the usage of pretrained language models, the generation of diverse and coherent dialogue remai...
| Main Authors: | , , |
|---|---|
| Other Authors: | |
| Format: | Conference Paper |
| Language: | English |
| Published: |
2023
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/172415 https://gem-benchmark.com/workshop https://2023.emnlp.org/program/workshops/ |
| _version_ | 1826120022663102464 |
|---|---|
| author | Lee, Jing Yang Lee, Kong Aik Gan, Woon-Seng |
| author2 | School of Electrical and Electronic Engineering |
| author_facet | School of Electrical and Electronic Engineering Lee, Jing Yang Lee, Kong Aik Gan, Woon-Seng |
| author_sort | Lee, Jing Yang |
| collection | NTU |
| description | To engage human users in meaningful conversation, open-domain dialogue agents
are required to generate diverse and contextually coherent dialogue. Despite
recent advancements, which can be attributed to the usage of pretrained
language models, the generation of diverse and coherent dialogue remains an
open research problem. A popular approach to address this issue involves the
adaptation of variational frameworks. However, while these approaches
successfully improve diversity, they tend to compromise on contextual
coherence. Hence, we propose the Bayesian Open-domain Dialogue with Empirical
Bayes (BODEB) framework, an empirical bayes framework for constructing an
Bayesian open-domain dialogue agent by leveraging pretrained parameters to
inform the prior and posterior parameter distributions. Empirical results show
that BODEB achieves better results in terms of both diversity and coherence
compared to variational frameworks. |
| first_indexed | 2024-10-01T05:09:45Z |
| format | Conference Paper |
| id | ntu-10356/172415 |
| institution | Nanyang Technological University |
| language | English |
| last_indexed | 2024-10-01T05:09:45Z |
| publishDate | 2023 |
| record_format | dspace |
| spelling | ntu-10356/1724152023-12-29T15:40:03Z An empirical bayes framework for open-domain dialogue generation Lee, Jing Yang Lee, Kong Aik Gan, Woon-Seng School of Electrical and Electronic Engineering 3rd Generation, Evaluation and Metrics (GEM) Workshop at EMNLP 2023 Engineering::Electrical and electronic engineering Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence Open-Domain Dialogue Chatbot To engage human users in meaningful conversation, open-domain dialogue agents are required to generate diverse and contextually coherent dialogue. Despite recent advancements, which can be attributed to the usage of pretrained language models, the generation of diverse and coherent dialogue remains an open research problem. A popular approach to address this issue involves the adaptation of variational frameworks. However, while these approaches successfully improve diversity, they tend to compromise on contextual coherence. Hence, we propose the Bayesian Open-domain Dialogue with Empirical Bayes (BODEB) framework, an empirical bayes framework for constructing an Bayesian open-domain dialogue agent by leveraging pretrained parameters to inform the prior and posterior parameter distributions. Empirical results show that BODEB achieves better results in terms of both diversity and coherence compared to variational frameworks. Submitted/Accepted version 2023-12-29T02:10:10Z 2023-12-29T02:10:10Z 2023 Conference Paper Lee, J. Y., Lee, K. A. & Gan, W. (2023). An empirical bayes framework for open-domain dialogue generation. 3rd Generation, Evaluation and Metrics (GEM) Workshop at EMNLP 2023. https://hdl.handle.net/10356/172415 https://gem-benchmark.com/workshop https://2023.emnlp.org/program/workshops/ en © 2023 Association for Computational Linguistics. All rights reserved. This article may be downloaded for personal use only. Any other use requires prior permission of the copyright holder. application/pdf |
| spellingShingle | Engineering::Electrical and electronic engineering Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence Open-Domain Dialogue Chatbot Lee, Jing Yang Lee, Kong Aik Gan, Woon-Seng An empirical bayes framework for open-domain dialogue generation |
| title | An empirical bayes framework for open-domain dialogue generation |
| title_full | An empirical bayes framework for open-domain dialogue generation |
| title_fullStr | An empirical bayes framework for open-domain dialogue generation |
| title_full_unstemmed | An empirical bayes framework for open-domain dialogue generation |
| title_short | An empirical bayes framework for open-domain dialogue generation |
| title_sort | empirical bayes framework for open domain dialogue generation |
| topic | Engineering::Electrical and electronic engineering Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence Open-Domain Dialogue Chatbot |
| url | https://hdl.handle.net/10356/172415 https://gem-benchmark.com/workshop https://2023.emnlp.org/program/workshops/ |
| work_keys_str_mv | AT leejingyang anempiricalbayesframeworkforopendomaindialoguegeneration AT leekongaik anempiricalbayesframeworkforopendomaindialoguegeneration AT ganwoonseng anempiricalbayesframeworkforopendomaindialoguegeneration AT leejingyang empiricalbayesframeworkforopendomaindialoguegeneration AT leekongaik empiricalbayesframeworkforopendomaindialoguegeneration AT ganwoonseng empiricalbayesframeworkforopendomaindialoguegeneration |