Abstract
In this paper, we propose an application called TaskGuideRoid(TGR). TGR recommends YouTube contents with task-based descriptions. To collect the task description with both high coverage and precision, we mine tasks from the web given the input of lexico-syntactic pattern using sightseeing spot names. A user evaluation shows that TGR improved the frequency of watching videos by about 15%. From the results of an interview, 90.0% of the subjects gave TGR high marks in terms of finding new enjoyable videos. Furthermore, We show that TGR allows users to find sightseeing spots through task based descriptions at unknown places.