As the amount of news publications increases each day, so does the need for effective search algorithms. Because simple word-based approaches are inherently limited, ignoring much of the information in natural language, in this paper we propose a linguistic approach called Destiny, which utilizes this information to improve search results. The major difference from approaches that represent text as a bag-of-words is that Destiny represents sentences as graphs, with words as nodes and the grammatical relations between words as edges. The proposed algorithm is evaluated using a custom corpus of user-rated sentences and compared to a TF-IDF baseline, performs significantly better in terms of Mean Average Precision, normalized Discounted Cumulative Gain, and Spearman’s Rho.

, ,,
Communications in Computer and Information Science
Erasmus University Rotterdam

Schouten, K., & Frasincar, F. (2016). Web news sentence searching using linguistic graph similarity. In Communications in Computer and Information Science. doi:10.1007/978-3-319-40180-5_22