Prediction of Anime Series’ Success using Aspect-Based Sentiment Analysis and Deep Learning

No Thumbnail Available

Date

2021

Journal Title

Journal ISSN

Volume Title

Publisher

Saudi Digital Library

Abstract

Natural language processing has a wide range of applications, such as aspect detection and sentiment analysis. These fields aim to understand the opinions and emotions of people by analyzing their textual reviews. Recently, people tend to write reviews about many things, such as the product they bought or the movie they watched. These raw reviews can be valuable for many people if we use them in the right way. This thesis in troduces Deep learning techniques to develop an Aspect-Based Sentiment Analysis model to predict the Anime series’ success of the MyAnimeList dataset. The proposed method is divided into two main tasks: identifying the main aspects in the Anime series’ reviews and determining the corresponding sentiment for each aspect. The results of these two tasks will give an accurate prediction of the Anime series’ success. The main goal is to help the decision-maker and the Anime producers by providing them knowledge mining from reviews. This knowledge allows them to understand the viewers’ emotions and sat isfactions. Four models were trained to perform aspect and sentiment classification tasks. The best model performance is determined based on the four metrics: Accuracy, recall, precision, and F1-score. The Long Short-Term Memory (LSTM) aspect model has the best value in all metrics with 67%. The Convolutional Neural Network (CNN) sentiment model has the best testing accuracy value with 95%, precision value with 96%, recall value with 98% and F1-score value of 97%. The LSTM sentiment model obtained the best Area Under the Curve (AUC) value of 77%.

Description

Keywords

Citation

Endorsement

Review

Supplemented By

Referenced By

Copyright owned by the Saudi Digital Library (SDL) © 2025