Skip to main navigation Skip to search Skip to main content

An Experimental Study on the Interpretability of Transformer Models for Dialog Understanding

  • Stony Brook University

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Identifying the intention of an utterance (e.g., spoken sentence) is part of semantic understanding. Transformer models offer promising performance for intention identification, however model interpretability remains low. This paper presents a comprehensive experimental study on the interpretability of BERT-like models, such as DistilBERT, for spoken dialog intention understanding. A detailed discussion explains the main features used in intention classification. This insight can be a starting point to devise more interpretable transformer models.

Original languageEnglish
Title of host publicationProceedings - 2025 IEEE Conference on Artificial Intelligence, CAI 2025
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages229-235
Number of pages7
ISBN (Electronic)9798331524005
DOIs
StatePublished - 2025
Event3rd IEEE Conference on Artificial Intelligence, CAI 2025 - Santa Clara, United States
Duration: May 5 2025May 7 2025

Publication series

NameProceedings - 2025 IEEE Conference on Artificial Intelligence, CAI 2025

Conference

Conference3rd IEEE Conference on Artificial Intelligence, CAI 2025
Country/TerritoryUnited States
CitySanta Clara
Period05/5/2505/7/25

Keywords

  • experimental study
  • interpretability
  • problem solving
  • spoken dialog understanding
  • transformer model

Fingerprint

Dive into the research topics of 'An Experimental Study on the Interpretability of Transformer Models for Dialog Understanding'. Together they form a unique fingerprint.

Cite this