On Classifying Discussion Threads Using Travel Information Goal-Oriented Model

We study how to recommend discussion threads in the tourism domain to meet visitors’ travel information needs. This research-in-progress paper reports the first stage of our research, namely classifying discussion threads into travel goals. We propose an information goal-oriented model, which consists of four goals: Initiation, Attraction, Accommodation, and Route planning, that can be characterized using nine features. Seven of these nine features can be quantified based on lexicons, and the other two can be measured using the named entity recognition technique. Three lexicons can be further enhanced using WordNet. We conduct an experiment in evaluating the impact of these features on goal classification with a data set collected from TripAdvisor.com, the world's largest travelling website. The experimental results show that our approach generally has comparable or better performance than that of using purely lexical features, namely TF-IDF, for classification.