Categorization of Text using Long Short-Term Memory with Glove

Abstract

Abstract Text files from the web, articles from magazines, and medical research can all be organized, arranged, and categorized using text classifiers. For illustration, categories like entertainment, business, sports, science, and technology might be used to categorize new content; When classifying text, important feature selection and data sparsity problems repeatedly arise in standard techniques. Text classification using conventional machine learning techniques is highly effective and has qualities of stability. Regarding large-scale dataset training, it does have some drawbacks. In this instance, grouping news articles into some labels in the dataset requires the use of a multi-label text grouping method. A potential approach to fixing issues with text categorization systems is deep learning. The LSTM was utilized in this research to utilize one method using recurrent neural networks for deep learning. The technique described in this article utilizes 300-dimensional word embeddings generated by Global Vectors (GloVe). The parameters were carefully selected to demonstrate the effectiveness of utilizing LSTM with GloVe features in text categorization. A comprehensive tuning process was carried out, assessing the performance of the four suggested LSTM models by evaluating them against a sizable corpus for comparative analysis. According to the results, the third model's accuracy for text categorization using LSTM and GloVe is 96.36, while The average F1-score, precision, and recall are 96. Furthermore, Utilising the GloVe feature, LSTM typically generates visual outcomes that are nearly well-fit.