A dual latent variable personalized dialogue agent
Personalized dialogue agents are capable of generating responses consistent with a specific persona. Typically, personalized dialogue agents generate responses based on both the dialogue history and a representation of the agent’s desired persona. As it is impractical to obtain the persona represent...
| Main Authors: | , , |
|---|---|
| Other Authors: | |
| Format: | Journal Article |
| Language: | English |
| Published: |
2023
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/169209 |
| _version_ | 1811691541139292160 |
|---|---|
| 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 | Personalized dialogue agents are capable of generating responses consistent with a specific persona. Typically, personalized dialogue agents generate responses based on both the dialogue history and a representation of the agent’s desired persona. As it is impractical to obtain the persona representations for every interlocutor in real-world implementations, recent works have explored the possibility of generating personalized dialogue by finetuning the agent with dialogue examples corresponding to a given persona instead. However, in real-world implementations, a sufficient number of corresponding dialogue examples are also rarely available. Hence, in this paper, we introduce the Dual Latent Variable Generator (DLVGen), a variational personalized dialogue agent capable of generating personalized dialogue without any persona information or any corresponding dialogue examples. Unlike previous works, DLVGen models the latent distribution over potential dialogue response intents as well as the latent distribution over the agent’s potential persona. During inference, latent variables are sampled from both distributions and fed to the decoder. Extensive experiments on the popular ConvAI2 personalized dialogue corpus show that DLVGen is capable of generating natural, persona consistent responses. Additionally, we also introduce a variance regularization and response selection approach which further improved overall response quality. |
| first_indexed | 2024-10-01T06:21:32Z |
| format | Journal Article |
| id | ntu-10356/169209 |
| institution | Nanyang Technological University |
| language | English |
| last_indexed | 2024-10-01T06:21:32Z |
| publishDate | 2023 |
| record_format | dspace |
| spelling | ntu-10356/1692092023-07-07T02:27:36Z A dual latent variable personalized dialogue agent Lee, Jing Yang Lee, Kong Aik Gan, Woon Seng School of Electrical and Electronic Engineering Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence Natural Language Generation Conversational AI Personalized dialogue agents are capable of generating responses consistent with a specific persona. Typically, personalized dialogue agents generate responses based on both the dialogue history and a representation of the agent’s desired persona. As it is impractical to obtain the persona representations for every interlocutor in real-world implementations, recent works have explored the possibility of generating personalized dialogue by finetuning the agent with dialogue examples corresponding to a given persona instead. However, in real-world implementations, a sufficient number of corresponding dialogue examples are also rarely available. Hence, in this paper, we introduce the Dual Latent Variable Generator (DLVGen), a variational personalized dialogue agent capable of generating personalized dialogue without any persona information or any corresponding dialogue examples. Unlike previous works, DLVGen models the latent distribution over potential dialogue response intents as well as the latent distribution over the agent’s potential persona. During inference, latent variables are sampled from both distributions and fed to the decoder. Extensive experiments on the popular ConvAI2 personalized dialogue corpus show that DLVGen is capable of generating natural, persona consistent responses. Additionally, we also introduce a variance regularization and response selection approach which further improved overall response quality. Ministry of Education (MOE) This research project was funded by the Ministry of Education, Singapore, under Grant AcRF TIER 2-2017-T2-2-060. 2023-07-07T02:27:36Z 2023-07-07T02:27:36Z 2023 Journal Article Lee, J. Y., Lee, K. A. & Gan, W. S. (2023). A dual latent variable personalized dialogue agent. SN Computer Science, 4(2), 159-. https://dx.doi.org/10.1007/s42979-022-01548-5 2661-8907 https://hdl.handle.net/10356/169209 10.1007/s42979-022-01548-5 2-s2.0-85146290558 2 4 159 en TIER 2-2017-T2-2-060 SN Computer Science © The Author(s), under exclusive licence to Springer Nature Singapore Pte Ltd 2023. All rights reserved. |
| spellingShingle | Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence Natural Language Generation Conversational AI Lee, Jing Yang Lee, Kong Aik Gan, Woon Seng A dual latent variable personalized dialogue agent |
| title | A dual latent variable personalized dialogue agent |
| title_full | A dual latent variable personalized dialogue agent |
| title_fullStr | A dual latent variable personalized dialogue agent |
| title_full_unstemmed | A dual latent variable personalized dialogue agent |
| title_short | A dual latent variable personalized dialogue agent |
| title_sort | dual latent variable personalized dialogue agent |
| topic | Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence Natural Language Generation Conversational AI |
| url | https://hdl.handle.net/10356/169209 |
| work_keys_str_mv | AT leejingyang aduallatentvariablepersonalizeddialogueagent AT leekongaik aduallatentvariablepersonalizeddialogueagent AT ganwoonseng aduallatentvariablepersonalizeddialogueagent AT leejingyang duallatentvariablepersonalizeddialogueagent AT leekongaik duallatentvariablepersonalizeddialogueagent AT ganwoonseng duallatentvariablepersonalizeddialogueagent |