Novel Association of Word Embeddings, PCA and K-means for Text Clustering

This paper addresses the problem of the increasing difficulty for users to find the information they want due to the explosive growth of documents on various topics in the modern web environment.Document clustering is a powerful tool that can improve the accessibility and usability of information by grouping documents with similar features, helping users quickly and easily find the information they need.In this paper, we proposed a novel method called PCA-KM (Principal Component Analysis -Kmeans) for document clustering.PCA-KM reduces dimensionality by applying principal component analysis to high-dimensional vectors of documents obtained from word embeddings, followed by the iterative application of modified K-means.Several clustering performance metrics were used to compare the proposed method with the traditional one, combining word embeddings and K-means.The results showed that the proposed method performed similarly or better on these metrics.Therefore, we expect that the proposed method will contribute to developing more efficient document retrieval services.