Content-based news recommendation is traditionally performed using the cosine similarity and TF-IDF weighting scheme for terms occurring in news messages and user profiles. Semantics-driven variants such as SF-IDF additionally take into account term meaning by exploiting synsets from semantic lexicons. However, they ignore the various semantic relationships between synsets, providing only for a limited understanding of news semantics. Moreover, semantics-based weighting techniques are not able to handle -- often crucial -- named entities, which are usually not present in semantic lexicons. Hence, we extend SF-IDF by also considering the synset semantic relationships, and by employing named entity similarities using Bing page counts. Our proposed method, Bing-SF-IDF+, outperforms TF-IDF and SF-IDF in terms of F1 scores and kappa statistics.
|Number of pages||2|
|Publication status||Published - 7 Apr 2014|
|Event||Twenty-Third International World Wide Web Conference (WWW 2014) - |
Duration: 7 Apr 2014 → 11 Apr 2014
|Conference||Twenty-Third International World Wide Web Conference (WWW 2014)|
|Period||7/04/14 → 11/04/14|