JISE


  [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15]


Journal of Information Science and Engineering, Vol. 39 No. 3, pp. 655-670


EventGo! Mining Events through Semi-Supervised Event Title Recognition and Pattern-based Venue/Date Coupling


YUAN-HAO LIN1, CHIA-HUI CHANG1,+ AND HSIU-MIN CHUANG2
1Department of Computer Science and Information Engineering
National Central University
Taoyuan, 320 Taiwan

2Department of Information and Computer Engineering
Chung Yuan Christian University
Taoyuan, 320 Taiwan
E-mail: luff543@gmail.com; chia@csie.ncu.edu.tw; showmin1205@gmail.com


Looking for local activities and events is a common task for most users during travel or daily life. Events are usually announced on the event organizers’ website or spread by posting on social networks such as Facebook Event or Facebook Fanpages. Integrating all these activities/events allows us to explore the city and understand its dynamics. In this article, we study the problem of event extraction, including event title recognition, venue extraction, and relationship coupling. Although distant supervision is a common technique for generating annotated training data, how to choose proper seed entities depends on the nature of the entities to be extracted, and the automatic labeling strategy adopted. To improve the performance, we proposed model-based distant supervision for event title recognition and Point Of Interest (POI) extraction, which reached 0.565 and 0.536 F1, respectively. Meanwhile, we conduct sequential pattern mining from Facebook event posts to determine the event venue and start/end date when multiple addresses/POIs or temporal expressions are recognized in a message. Overall, the average F1 of the proposed model in event extraction is 0.620.


Keywords: event title extraction, venue recognition, social event search, relation coupling, semi-supervised learning

  Retrieve PDF document (JISE_202303_14.pdf)