Context ontologies for recommending from the social web

Investigations into combining context and recommendation has resulted in much fruitful research which has improved recommender systems. Such contextual information has come in many forms and been used in different ways, successfully offering better in-situ suggestions. Factors such as location, time of recommendation, etc. have proven themselves as useful contributors to exploiting context.\ud \ud One issue, however, is the importance placed on each aspect of context, especially as new forms of recommendation are developed. Context is traditionally incorporated into recommenders at design-time, as a filter or as an integral part of how users are modelled, but the importance placed on each aspect is not often examined. Social recommenders and systems that draw on the wealth of data present in social networks frequently have access to far more contextual factors than traditional recommenders, making user relationships to these factors all the more important. \ud \ud The main contribution of this paper is to provide an examination of contextual priorities from the social web, which prove useful to recommender research in the area. This ontological examination of context shows that users have different priorities when it comes to context with a large variation in the suitability of each contextual factor in predicting good recommendations. In addition, this paper presents and discusses an approach to individually tailoring context ontologies (allowing for dynamically generated context sets), evaluating contextual factors in recommending from the social web