Defense: Yunhao Fan
Oral examination for Project Option (Option III)
December 1, 12 p.m. (noon)
via Zoom: https://go.unl.edu/zo60
Meeting ID: 990 6381 8530
Passcode: 816882
Twitter is a widely used social media platform and has become an important source of information to monitor a variety of social processes. Extracting 5Ws (where, when, who, what, and why) information from the tweets will facilitate the transformation of millions of daily tweets into actionable intelligence. However, the performance of current 5Ws algorithms have been developed for structured text, does not translate to tweets because the tweets are short, unstructured, and informal in nature. In this research, we have developed extensions to the 5Ws extraction algorithms to adapt to tweets by leveraging the syntactical and semantic relationships within the text. To examine the performance of the modified algorithms, we manually annotate a collection of tweets due to the absence of ground truth. Overall, the performance of the revised algorithms on tweets does not match that of the original algorithms on structured new articles. The performance of modified “where,” “when,” and “why” algorithms are efficient when compared to original algorithms. However, for “who” and “what,” the revised algorithms are not as efficient and need improvement.