A Context-aware Approach to News Article Retrieval

Abstract

In the past few years the number of publishing Urdu news online has grown rapidly. There are a number of online Urdu newspapers and their readers have grown. However, recommending relevant stories to the readers still remains an issue. There are various approaches to develop a recommender system such as content-based, collaborative filtering (CF) and hybrid approaches. However, the approaches are developed for content in other languages such as English, Spanish and German. There is still a need to develop an approach for Urdu language considering the specific characteristics of the language. Many researchers have recognized the importance of the contextual information in various areas including information retrieval. However, the research community has already made a significant amount of contribution in the field of recommender systems. Mostly, existing approaches do not consider the contextual information while making recommendations. In this study, a context-aware approach is proposed. It is an alternative to content and collaborative filtering based approaches. In context-aware approach the filtering can be made by using item rating or content similarity by considering the context. The proposed approach is focused on news content, context and similarity index to filter the relevant articles and to make recommendations. In the proposed methodology, similarity on the basis of content, by combining content and context and ensembling the output of both cosine similarity and pearson coefficient is computed. It is observed that an ensemble is significantly more effective than any individual models used in this study. The precision and recall obtained by ensembling both similarity measurement approaches are 96% and 97.9% respectively.