An Empirical Analysis of Topic Categorization Using PaLM, GPT and BERT Models
Dhanvanth Reddy Yerramreddy, Jayasurya Marasani, Ponnuru Sathwik Venkata Gowtham, S Abhishek, Anjali Anjali · 2023
Topic Categorization is a crucial task for organizing and managing information. It is very useful for making it easier to find and retrieve relevant documents. In a mode of digital world filled with information, topic categorization plays pivotal role in many aspects. Numerous number of models were introduced in order to categorize the topics using various datasets whereas PaLM, BERT and GPT models are all large language models (LLM) that have been developed in the recent years. These models can learn the statistical relationships between words and phrases, which allows them to identify the topics that are discussed in a document with greater accuracy. They can also process large amounts of text data quickly, which is important for applications that require real-time topic categorization. Additionally, they are more robust to noise and errors than traditional methods. As these models continue to develop, they are likely to become even more effective at topic categorization. A comparative study has been made among PaLM, BERT and GPT language models for topic categorization using AG News Dataset in this paper. The results are promising and encouraging where the suggested models provides more precise categorization for text of the dataset.