The Relationship between Embedding and Deductive Reasoning Ability on BERT
DC Field | Value | Language |
---|---|---|
dc.contributor.author | Ryu, Jongwon | - |
dc.contributor.author | Kim, Junyeong | - |
dc.date.accessioned | 2024-03-28T04:30:32Z | - |
dc.date.available | 2024-03-28T04:30:32Z | - |
dc.date.issued | 2024-01 | - |
dc.identifier.issn | 0747-668X | - |
dc.identifier.uri | https://scholarworks.bwise.kr/cau/handle/2019.sw.cau/73049 | - |
dc.description.abstract | BERT is a pre-Trained language model that has been shown to be effective for a variety of tasks, including natural language inference (NLI) and textual entailment. However, it is not clear how well BERT can identify logical correctness or inconsistency. In this paper, we investigate the ability of BERT to distinguish between logically correct sentences and those with logical contradictions. We perform this by comparing the similarity scores of these two types of sentences using a variety of BERT models. Our results were conducted to understand the relationship between BERT embeddings and ability of logical reasoning. To this end, we identified that BERT is able to capture some aspects of logical relationships between words or sentences. However, our results also show that the logical reasoning ability of BERT varies depending on the complexity of the logic involved. Our findings contribute to the understanding of how language models can be used for deductive reasoning tasks. In addition, we are the first to introduce a new zero-shot evaluation dataset for deductive reasoning. We believe that our works could be used to improve the performance of language models for these tasks. © 2024 IEEE. | - |
dc.language | 영어 | - |
dc.language.iso | ENG | - |
dc.publisher | Institute of Electrical and Electronics Engineers Inc. | - |
dc.title | The Relationship between Embedding and Deductive Reasoning Ability on BERT | - |
dc.type | Article | - |
dc.identifier.doi | 10.1109/ICCE59016.2024.10444307 | - |
dc.identifier.bibliographicCitation | Digest of Technical Papers - IEEE International Conference on Consumer Electronics, v.2024 IEEE | - |
dc.description.isOpenAccess | N | - |
dc.identifier.scopusid | 2-s2.0-85186974084 | - |
dc.citation.title | Digest of Technical Papers - IEEE International Conference on Consumer Electronics | - |
dc.citation.volume | 2024 IEEE | - |
dc.type.docType | Conference paper | - |
dc.description.journalRegisteredClass | scopus | - |
Items in ScholarWorks are protected by copyright, with all rights reserved, unless otherwise indicated.
84, Heukseok-ro, Dongjak-gu, Seoul, Republic of Korea (06974)02-820-6194
COPYRIGHT 2019 Chung-Ang University All Rights Reserved.
Certain data included herein are derived from the © Web of Science of Clarivate Analytics. All rights reserved.
You may not copy or re-distribute this material in whole or in part without the prior written consent of Clarivate Analytics.