Clustering of news texts using hyperdimensional computing

Abstract

The computational treatment of unstructured data presents a formidable challenge owing to the multitude of implicit elements that must be considered. The purview of such processes extends to computational tasks such as clustering, classification, and administrative functions. Furthermore, the volume of this data is experiencing exponential growth, necessitating a corresponding increase in processing capabilities. By carefully selecting numerical indicators obtained from the same text, it is possible to group them by their theme. However, managing these indicators together is not trivial. An alternative solution is the use of hypervectors and specific operators to integrate a solution for handling such texts. This work presents the process of generating and evaluating text identifiers from news agencies for grouping them without the need for human reading.