An Extraction Method of Well-Matched Spot Pairs based on Co-occurrence of Experiences in Travel Plans
Junsaku Takada, Daisuke Kitayama
Proceedings of The International MultiConference of Engineers and Computer Scientists 2021(IMECS2021),pp.156–161,Hong Kong (online),2021年10月,査読有
論文PDF
概要
When travellers go sightseeing, they often know the main sightseeing spots to visit, but deciding on other interesting spots to visit can be time-consuming and difficult. Therefore, in this study, travellers input the main tourist spots that they want to visit and recommend other sights that are well-matched with those spots. In this study, well-matched is defined as a high rate of co-occurrence between experiences that appear in existing travel plans. The probability of co-occurrence is calculated based on the amount of mutual information. In this study, spot recommendations are based on the cosine similarity between the vector created by the probability of co-occurrence of experiences appearing in travel records of tourist sites and the vector created from the review texts of tourist sites. In this study, we use travel records generated from Flickr for trip planning. The combination of spots with well-matched experiences is perceived to be selected from the travel records. The experiences in this study are extracted from review texts available on travel websites.