Contextual Embeddings-Based Web Page Categorization Using the Fine-Tune BERT Model

The World Wide Web has revolutionized the way we live, causing number of web pages to increase exponentially. provides access a tremendous amount information, so it is difficult for internet users locate accurate and useful information on web. In order categorize accurately based queries users, methods categorizing need be developed. text content plays significant role in categorization pages. If word’s position altered within sentence, change interpretation that this phenomenon called polysemy. page categorization, polysemy property causes ambiguity referred as problem. This paper proposes fine-tuned model solve problem, using contextual embeddings created by symmetry multi-head encoder layer Bidirectional Encoder Representations from Transformers (BERT). effectiveness proposed was evaluated benchmark datasets i.e., WebKB DMOZ. Furthermore, experiment series also model’s hyperparameters achieve 96.00% 84.00% F1-Scores, respectively, demonstrating importance compared baseline approaches machine learning deep learning.