| Summary: | 碩士 === 國立政治大學 === 資訊科學學系 === 99 === Trip planning is an important and time-consuming step for backpackers. Most research focuses on finding the travel sequence from different data sources such as blog, photos and GPS. Although these approaches can recommend some popular travel sequences for a tourist, but tourist’s place preferences and temporal constraints are not considered. In this thesis, we propose an approach for personalized trip planning which takes tourist’s preference and temporal constraint into consideration.
In the proposed approach, first the place names are extracted from travelogues with the aid of the Wikipedia. Then the travel sequences are extracted from travelogues. Based on the relationship of mutual reinforcement between the authority of a place and the hub of a travelogue author, the authority of each place is derived. Moreover, the stay time of each place is estimated from the information of travel photos of a web album. Finally, based on the user specified place preference and temporal constraints, this thesis presents the algorithms to arrange a personalized trip for a user. The experiments show that the place name extraction achieves 92% precision and 100% recall. For the estimation of place stay time, the error is 3.16% compared with the ground truth collected from well-known backpacker site.
|
|---|
