Home // DATA ANALYTICS 2023, The Twelfth International Conference on Data Analytics // View article

Text Classification Using a Word-Reduced Graph

Authors:
Hiromu Nakajima
Minoru Sasaki

Keywords: text classification, graph convolutional neural network, semi-supervised learning.

Abstract:
Text classification, which determines the label of a document based on cues such as the co-occurrence of words and their frequency of occurrence, has been studied in various approaches to date. Conventional text classification methods using graph structure data express the relationship between words, the relationship between words and documents, and the relationship between documents in terms of the weights of edges between each node. They are then trained by inputting to a graph neural network. However, text classification methods using those graph-structured data require a very large amount of memory, and therefore, in some environments, they do not work properly or cannot handle large data. In this study, we propose a graph structure that is more compact than conventional methods by removing words that appear in only one document and are considered unnecessary for text classification. In addition to save memory, the proposed method can use a larger trained model by utilizing the saved memory. The results showed that the method succeeded in saving memory while maintaining the accuracy of the conventional method. By utilizing the saved memory, the proposed method succeeded in using larger trained models, and the classification accuracy of the proposed method was dramatically improved compared to the conventional method.

Pages: 25 to 30

Copyright: Copyright (c) IARIA, 2023

Publication date: September 25, 2023

Published in: conference

ISSN: 2308-4464

ISBN: 978-1-68558-111-4

Location: Porto, Portugal

Dates: from September 25, 2023 to September 29, 2023