Automatic Construction of Expiration Time Expression Dataset from Retweets
by Hirotaka Nagashima, Keishi Tajima
Abstract
Documents describing information with expiration time often include
time expressions specifying the expiration time. To train a
classifier determining if a time expression represents an expiration
time, we need a labeled dataset. We propose a method of automatically
constructing such a dataset. Our method collects tweets including
time expressions, and automatically determines whether the time
expressions represent expiration times based on temporal changes in
the frequency of retweets. Our experimental result shows that our
method produces an effective dataset.
/ Fax: +81(Japan) 75-753-5978 / Office: Research Bldg. #7, room 404