DC Field | Value | Language |
---|---|---|
dc.contributor.author | Ko, Bowon | ko |
dc.contributor.author | Choi, Ho-Jin | ko |
dc.date.accessioned | 2020-11-11T07:55:36Z | - |
dc.date.available | 2020-11-11T07:55:36Z | - |
dc.date.created | 2020-11-09 | - |
dc.date.created | 2020-11-09 | - |
dc.date.issued | 2020-02-21 | - |
dc.identifier.citation | 2020 IEEE International Conference on Big Data and Smart Computing, BigComp 2020, pp.217 - 220 | - |
dc.identifier.issn | 2375-933X | - |
dc.identifier.uri | http://hdl.handle.net/10203/277242 | - |
dc.description.abstract | It is very important to analyze the semantic similarity of two sentences for Natural Language Processing (NLP). This paper proposes a Paraphrase-BERT to perform Paraphrase Identification task. We first fine-tune the pre-trained BERT with MRPC data and add a Whole Word Masking, which is pre-training method recently announced by Google, to the BERT. Finally, we perform Multi-Task Learning (MLT) to improve performance. Specifically, the Question Answering task and the Paraphrase Identification (PI) task are learned sequentially to improve performance of PI task. As a result, it has shown that MLT affect a performance improvement of downstream task (11.11% point absolute accuracy improvement, 7.88% point absolute F1 improvement). | - |
dc.language | English | - |
dc.publisher | IEEE,Korean Institute of Information Scientists and Engineers (KIISE) | - |
dc.title | Paraphrase Bidirectional Transformer with Multi-task Learning | - |
dc.type | Conference | - |
dc.identifier.wosid | 000569987500038 | - |
dc.identifier.scopusid | 2-s2.0-85084357870 | - |
dc.type.rims | CONF | - |
dc.citation.beginningpage | 217 | - |
dc.citation.endingpage | 220 | - |
dc.citation.publicationname | 2020 IEEE International Conference on Big Data and Smart Computing, BigComp 2020 | - |
dc.identifier.conferencecountry | KO | - |
dc.identifier.conferencelocation | Busan | - |
dc.identifier.doi | 10.1109/bigcomp48618.2020.00-72 | - |
dc.contributor.localauthor | Choi, Ho-Jin | - |
dc.contributor.nonIdAuthor | Ko, Bowon | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.