An Effective XML Documents Clustering Method Using Word Embeddings for Heterogeneous Collections
As the size of XML repositories is growing, XML data management becomes challenging as how these documents can be stored and retrieved. One way of resolving such issues is to group the documents into clusters so that documents within the same cluster are more related than documents in different clusters. This became necessary in order to aid indexing and retrieval of XML documents. Traditional documents clustering methods represents documents with models that fails to consider the semantic relation between words. In this paper, WEClusterX is proposed to semantically cluster XML documents. The idea behind WEClusterX is to pinpoint which concept is represented by a particular context. Firstly, a pre-trained Bidirectional Encoder Representations from Transformers (BERT) is used to extract and cluster embeddings. Then, a Context-Document matrix is generated from the cluster of embeddings. Finally, clusters were formed using the famous k-means algorithm. The method combines the statistical importance of words with their contextualized representation in documents in order to forms meaningful clusters. The proposed WEClusterX is evaluated using extensive experiments. Experimental results have demonstrated that our proposed clustering solution achieved better performance in terms of purity and entropy.
Authors
- B.A. Bodinga
Institutions
- Usmanu Danfodiyo University (NG)
Publication Details
- Journal
- INTERNATIONAL JOURNAL OF COMPUTER SCIENCE AND MATHEMATICAL THEORY E-ISSN
- Published
- 2026-11-13
- DOI
- https://doi.org/10.56201/ijcsmt.v10.no2.2024.pg120.140
- Primary Topic
- Advanced Clustering Algorithms Research
- Type
- article
- Field-Weighted Citation Impact
- 0.00