Abstract:
Information retrieval often yields an overwhelming number of results, making effective data utilization challenging. Clustering techniques enhance navigation by grouping similar results. However, Arabic text clustering research often neglects semantic relationships. This study introduces the Bag of Tokens (BOT) model, which integrates semantic information, unlike the traditional Bag of Words (BOW) model. BOT relies on entities and relations extracted from text, enabling both improved clustering and dimensionality reduction. The approach was evaluated on an Arabic text dataset from the Wikipedia Monolingual Corpora, focusing on Sudanese politicians' biographies. Results demonstrated that BOT achieves lower dimensionality and enhances cluster coherence and separation across various clustering algorithms and distance metrics compared to BOW
