| 大会名称 |
| 2019年 総合大会 |
| 大会コ－ド |
| 2019G |
| 開催年 |
|
2019
|
| 発行日 |
|
2019-03-05
|
| セッション番号 |
|
DS-1
|
| セッション名 |
|
COMP 学生シンポジウム
|
| 講演日 |
|
2019/03/19
|
| 講演場所(会議室等) |
|
54号館 102教室
|
| 講演番号 |
|
DS-1-4
|
| タイトル |
|
Multi-Class Sentiment Analysis in Twitter Using Machine Learning and Deep Learning
|
| 著者名 |
|
○Mondher Bouazizi, Tomoaki Ohtsuki,
|
| キーワード |
|
Twitter, Machine Learning, Deep Learning, Sentiment Analysis
|
| 抄録 |
|
Multi-Class sentiment analysis is a particular type of sentiment analysis in which a piece of text is attributed one out of many sentiment classes. This is in contrast to the conventional binary or ternary sentiment analysis where the piece of text is attributed a class out of two or three, respectively. In this report, we introduce an approach that uses both deep learning (DL) and machine learning (ML) techniques to perform multi-class sentiment analysis and improve the classification accuracy compared to the approaches, which rely solely on ML or DL. For 7 different sentiment classes, our approach reaches an accuracy equal to 66.2%, outperforming the ones that rely on ML and DL by around 6% and 1% respectively.
|
| 本文pdf |
|
PDF download
|
