Exploration of Ranking-based Strategy for Contextual Suggestion.

摘要

We describe our efforts in TREC 2012 Contextual Suggestion Track. The task is to provide suggestions to users based on their personal interests as well as their contexts. To tackle the problem, we propose to rank candidate suggestions based on their similarity to the personal profile and that to the contexts (i.e. geographic and temporal information). The ranking function is computed based on the similarity between a suggestion and the places that the user like and the dis-similarity between the suggestion and the places disliked by the user. The similarities are computed based on the either the category or description of the suggestions.