NORMA eResearch @NCI Library

The Semi-supervised Approach Using GAN and BERT for News Text Classification

Bhor, Sourav Prabhakar (2022) The Semi-supervised Approach Using GAN and BERT for News Text Classification. Masters thesis, Dublin, National College of Ireland.

[thumbnail of Master of Science]
PDF (Master of Science)
Download (1MB) | Preview
[thumbnail of Configuration manual]
PDF (Configuration manual)
Download (800kB) | Preview


The transformers for Natural Language Processing (NLP), such as Bidirectional Encoder Representations from Transformers (BERT), are impressively effective with these tasks. To test the effectiveness of this model, benchmarks set for the evaluations were made based on the training data, which is humongous in size. In reality, finding the training data appropriately labelled and neatly organised in separate categories is rare at first but also very expensive because of its higher usability. Because training the model to give better suggestions and near-perfect context understanding would require such data in considerable quantity. The semisupervised GAN models used in image data processing are very effective. This research focuses on fine-tuning the base of the BERT model with the Generative Adversarial Network(GAN) extension. The research aims to provide better performance where the labelled and unlabelled data are meagre. Thus also enabling the use of unlabelled data for the training purpose. I found that the model outperformed the baseline model, especially when the annotated data is below 20 per cent with the dataset I used for this research work.

Item Type: Thesis (Masters)
Subjects: Q Science > QA Mathematics > Electronic computers. Computer science
T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science
P Language and Literature > P Philology. Linguistics > Computational linguistics. Natural language processing
Divisions: School of Computing > Master of Science in Data Analytics
Depositing User: Tamara Malone
Date Deposited: 19 Jan 2023 12:12
Last Modified: 06 Mar 2023 16:26

Actions (login required)

View Item View Item