Structured Event Retrieval over Microblog Archives

2012 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 646–655, Montr´eal, Canada, June 3-8, 2012. c⃝2012 Association for Computational Linguistics Structured Event Retrieval over Microblog Archives Donald Metzler, Congxing Cai, Eduard Hovy Information Sciences Institute University of Southern California 4676 Admiralty Way, Marina del Rey, CA 90292 Abstract Microblog streams often contain a consider- able amount of information about local, re- gional, national, and global events. Most ex- isting microblog search capabilities are fo- cused on recent happenings and do not provide the ability to search and explore past events. This paper proposes the problem of structured retrieval of historical event information over microblog archives. Rather than retrieving in- dividual microblog messages in response to an event query, we propose retrieving a ranked list of historical event summaries by distill- ing high quality event representations using a novel temporal query expansion technique. The results of an exploratory study carried out over a large archive of Twitter messages demonstrates both the value of the microblog event retrieval task and the effectiveness of our proposed search methodologies. 1 Introduction Real-time user generated content is one of the key driving forces behind the growing popularity of so- cial media-centric communication. The ability to in- stantly share, often from your mobile phone, your thoughts (via Twitter), your photos (via Facebook), your location (via Foursquare), and a variety of other information is changing the way that information is created, communicated, and consumed. There has been a substantial amount of research effort devoted to user generated content-related search tasks, including blog search, forum search, and community-based question answering. How- ever, there has been relatively little research on mi- croblog search. Microblog services, such as Tumblr and Twitter, provide users with the ability to broad- cast short messages in real-time. This is in contrast to traditional blogs that typically have considerably more content that is updated less frequently. By their very nature, microblog streams often contain a considerable amount of information about local, regional, national, and global news and events. A recent study found that over 85% of trending topics on Twitter are news-related (Kwak et al., 2010). An- other recent study by Teevan et al. that investigated the differences between microblog and Web search reported similar findings (Teevan et al., 2011). The study also found that microblog search queries are used to find information related to news and events, while Web search queries are more navigational in nature and used to find a variety of information on a specific topic. It is likely that microblogs have not received much attention because, unlike blog search, there is no well-defined microblog search task. Existing mi- croblog search services, such as those offered by Twitter and Google, only provide the ability to re- trieve individual microblog posts in response to a query. Unfortunately, this task has limited utility since very few real information needs can be satis- fied by a single short piece of text (e.g., the max- imum length of …

Content truncated to 500 words