Major Life Event Extraction from Twitter based on Congratulations/Condolences Speech Acts

Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 1997–2007, October 25-29, 2014, Doha, Qatar. c⃝2014 Association for Computational Linguistics Major Life Event Extraction from Twitter based on Congratulations/Condolences Speech Acts Jiwei Li1, Alan Ritter2, Claire Cardie3 and Eduard Hovy4 1Computer Science Department, Stanford University, Stanford, CA 94305, USA 2Department of Computer Science and Engineering, the Ohio State University, OH 43210, USA 3Computer Science Department, Cornell University, Ithaca, NY 14853, USA 4Language Technology Institute, Carnegie Mellon University, PA 15213, USA jiweil@stanford.edu ritter.1492@osu.edu cardie@cs.cornell.edu ehovy@andrew.cmu.edu Abstract Social media websites provide a platform for anyone to describe significant events taking place in their lives in realtime. Currently, the majority of personal news and life events are published in a tex- tual format, motivating information ex- traction systems that can provide a struc- tured representations of major life events (weddings, graduation, etc. . . ). This pa- per demonstrates the feasibility of accu- rately extracting major life events. Our system extracts a fine-grained description of users’ life events based on their pub- lished tweets. We are optimistic that our system can help Twitter users more easily grasp information from users they take in- terest in following and also facilitate many downstream applications, for example re- altime friend recommendation. 1 Introduction Social networking websites such as Facebook and Twitter have recently challenged mainstream me- dia as the freshest source of information on im- portant news events. In addition to an important source for breaking news, social media presents a unique source of information on private events, for example a friend’s engagement or college gradua- tion (examples are presented in Figure 1). While a significant amount of previous work has inves- tigated event extraction from Twitter (e.g., (Rit- ter et al., 2012; Diao et al., 2012)), existing ap- proaches mostly focus on public bursty event ex- traction, and little progress has been made towards the problem of automatically extracting the major life events of ordinary users. A system which can automatically extract ma- jor life events and generate fine-grained descrip- tions as in Figure 1 will not only help Twitter users with the problem of information overload by summarizing important events taking place in their friends lives, but could also facilitate downstream applications such as friend recommendation (e.g., friend recommendation in realtime to people who were just admitted into the same university, get the same jobs or internships), targeted online ad- vertising (e.g., recommend baby care products to newly expecting mothers, or wedding services to new couples), information extraction, etc. Before getting started, we first identify a num- ber of key challenges in extracting significant life events from user-generated text, which account the reason for the lack of previous work in this area: Challenge 1: Ambiguous Definition for Ma- jor Life Events Major life event identification is an open-domain problem. While many types of events (e.g., marriage, engagement, finding a new job, giving birth) are universally agreed to be im- portant, it is difficult to robustly predefine a list of …

Content truncated to 500 words