Treffer: Parallel implementation of information retrieval clustering models

Information Retrieval (IR) is fundamental nowadays, and more since the appearance of the Internet and huge amount of information in electronic format. All this information is not useful unless its search is efficient and effective. With large collections parallelization is important because the data volume is enormous. Hence, usually, only one computer is not sufficient to manage all data, and more in a reasonable time. The parallelization also is important because in many situations the document collection is already distributed and its centralization is not a good idea. This is the reason why we present parallel algorithms in information retrieval systems. We propose two parallel clustering algorithms: α-Bisecting K-Means and α-Bisecting Spherical K-Means. Moreover, we have prepared a set of experiments to compare the computation performance of the algorithms. These studies have been accomplished in a cluster of PCs with 20 bi-processor nodes and two different collections.