Augmenting Training Data for a Virtual Character Using GPT-3.5

This paper compares different methods of using a large language model (GPT-3.5) for creating synthetic training data for a retrieval-based conversational character. The training data are in the form of linked questions and answers, which allow a classifier to retrieve a pre-recorded answer to an uns...

Full description

Saved in:
Bibliographic Details
Main Authors: Elizabeth Chen, Ron Artstein
Format: Article
Language:English
Published: LibraryPress@UF 2024-05-01
Series:Proceedings of the International Florida Artificial Intelligence Research Society Conference
Subjects:
Online Access:https://journals.flvc.org/FLAIRS/article/view/135552
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849736777731932160
author Elizabeth Chen
Ron Artstein
author_facet Elizabeth Chen
Ron Artstein
author_sort Elizabeth Chen
collection DOAJ
description This paper compares different methods of using a large language model (GPT-3.5) for creating synthetic training data for a retrieval-based conversational character. The training data are in the form of linked questions and answers, which allow a classifier to retrieve a pre-recorded answer to an unseen question; the intuition is that a large language model could predict what human users might ask, thus saving the effort of collecting real user questions as training data. Results show small improvements in test performance for all synthetic datasets. However, a classifier trained on only small amounts of collected user data resulted in a higher F-score than the classifiers trained on much larger amounts of synthetic data generated using GPT-3.5. Based on these results, we see a potential in using large language models for generating training data, but at this point it is not as valuable as collecting actual user data for training.
format Article
id doaj-art-c2c758a7e70c4fb291e674494a9cb971
institution DOAJ
issn 2334-0754
2334-0762
language English
publishDate 2024-05-01
publisher LibraryPress@UF
record_format Article
series Proceedings of the International Florida Artificial Intelligence Research Society Conference
spelling doaj-art-c2c758a7e70c4fb291e674494a9cb9712025-08-20T03:07:10ZengLibraryPress@UFProceedings of the International Florida Artificial Intelligence Research Society Conference2334-07542334-07622024-05-013710.32473/flairs.37.1.13555271931Augmenting Training Data for a Virtual Character Using GPT-3.5Elizabeth Chen0https://orcid.org/0009-0006-7188-6859Ron Artstein1https://orcid.org/0009-0005-5187-6381Wellesley CollegeInstitute for Creative Technologies, University of Southern CaliforniaThis paper compares different methods of using a large language model (GPT-3.5) for creating synthetic training data for a retrieval-based conversational character. The training data are in the form of linked questions and answers, which allow a classifier to retrieve a pre-recorded answer to an unseen question; the intuition is that a large language model could predict what human users might ask, thus saving the effort of collecting real user questions as training data. Results show small improvements in test performance for all synthetic datasets. However, a classifier trained on only small amounts of collected user data resulted in a higher F-score than the classifiers trained on much larger amounts of synthetic data generated using GPT-3.5. Based on these results, we see a potential in using large language models for generating training data, but at this point it is not as valuable as collecting actual user data for training.https://journals.flvc.org/FLAIRS/article/view/135552large language modelssynthetic datadialog based systems
spellingShingle Elizabeth Chen
Ron Artstein
Augmenting Training Data for a Virtual Character Using GPT-3.5
Proceedings of the International Florida Artificial Intelligence Research Society Conference
large language models
synthetic data
dialog based systems
title Augmenting Training Data for a Virtual Character Using GPT-3.5
title_full Augmenting Training Data for a Virtual Character Using GPT-3.5
title_fullStr Augmenting Training Data for a Virtual Character Using GPT-3.5
title_full_unstemmed Augmenting Training Data for a Virtual Character Using GPT-3.5
title_short Augmenting Training Data for a Virtual Character Using GPT-3.5
title_sort augmenting training data for a virtual character using gpt 3 5
topic large language models
synthetic data
dialog based systems
url https://journals.flvc.org/FLAIRS/article/view/135552
work_keys_str_mv AT elizabethchen augmentingtrainingdataforavirtualcharacterusinggpt35
AT ronartstein augmentingtrainingdataforavirtualcharacterusinggpt35