BERT-Based Sentiment Analysis Using Distillation

dc.contributor.authorLehečka, Jan
dc.contributor.authorŠvec, Jan
dc.contributor.authorIrcing, Pavel
dc.contributor.authorŠmídl, Luboš
dc.date.accessioned2021-03-01T11:00:24Z
dc.date.available2021-03-01T11:00:24Z
dc.date.issued2020
dc.description.abstract-translatedIn this paper, we present our experiments with BERT (Bidirectional Encoder Representations from Transformers) models in the task of sentiment analysis, which aims to predict the sentiment polarity for the given text. We trained an ensemble of BERT models from a large self-collected movie reviews dataset and distilled the knowledge into a single production model. Moreover, we proposed an improved BERT’s pooling layer architecture, which outperforms standard classification layer while enables per-token sentiment predictions. We demonstrate our improvements on a publicly available dataset with Czech movie reviews.en
dc.format13 s.cs
dc.format.mimetypeapplication/pdf
dc.identifier.citationLEHEČKA, J., ŠVEC, J., IRCING, P., ŠMÍDL, L. BERT-Based Sentiment Analysis Using Distillation. In Statistical Language and Speech Processing, SLSP 2020. Cham: Springer, 2020. s. 58-70. ISBN 978-3-030-59429-9, ISSN 0302-9743.cs
dc.identifier.doi10.1007/978-3-030-59430-5_5
dc.identifier.isbn978-3-030-59429-9
dc.identifier.issn0302-9743
dc.identifier.obd43930643
dc.identifier.uri2-s2.0-85092196103
dc.identifier.urihttp://hdl.handle.net/11025/42765
dc.language.isoenen
dc.project.IDTN01000024/Národní centrum kompetence - Kybernetika a umělá inteligencecs
dc.publisherSpringeren
dc.relation.ispartofseriesStatistical Language and Speech Processing, SLSP 2020en
dc.rightsPlný text je přístupný v rámci univerzity přihlášeným uživatelům.cs
dc.rights© Springeren
dc.rights.accessrestrictedAccessen
dc.subject.translatedSentiment analysisen
dc.subject.translatedBERTen
dc.subject.translatedKnowledge distillationen
dc.titleBERT-Based Sentiment Analysis Using Distillationen
dc.typekonferenční příspěvekcs
dc.typeconferenceObjecten
dc.type.statusPeer-revieweden
dc.type.versionpublishedVersionen

Files