[4Xin1] Poster session 2

[4Xin1-06] Annotating Visiting Order in Travelogues toward Trajectory Extraction and Visualization

〇Aitaro Yamamoto¹, Shohei Higashiyama^2,1, Hiroki Ouchi^1,3, Hiroyuki Otomo¹, Yusuke Ide¹, Hiroyuki Shindo¹, Taro Watababe¹ (1.Nara Institute of Science and Technology, 2.National Institute of Information and Communications , 3.Institute of Physical and Chemical Research)

Keywords:Trajectory, Visiting Order, Corpus Annotation

In recent years, tourists' travel patterns have changed with the spread of COVID-19, and it has become more important for both tourism service providers and users to obtain and use information on human mobility trajectories. We are aiming to develop a system that can extract human mobility trajectories from text and visualize them on a map. As a first step of dataset construction for training and evaluating such trajectory analysis systems, we have annotated a Japanese travelogue dataset with information related to visited places. In this study, we focus on the visiting order in which travelers visited places and design criteria for visiting order annotation. We performed a trial annotation experiment based on the criteria and achieved a high inter-annotator agreement rate of approximately 0.8 F1. We also analyzed the discrepancy between annotation by two annotators, and the results suggest some issues for further improvement of the annotation criteria.