Emergency-relief coordination on social media: Automatically matching resource requests and offers

Abstract

Community Question Answering has emerged as a popular and effective paradigm for a wide range of information needs. For example, to find out an obscure piece of trivia, it is now possible and even very effective to post a question on a popular community QA site such as Yahoo! Answers, and to rely on other users to provide answers, often within minutes. The importance of such community QA sites is magnified as they create archives of millions of questions and hundreds of millions of answers, many of which are invaluable for the information needs of other searchers. However, to make this immense body of knowledge accessible, effective answer retrieval is required. In particular, as any user can contribute an answer to a question, the majority of the content reflects personal, often unsubstantiated opinions. A ranking that combines both relevance and quality is required to make such archives usable for factual information retrieval. This task is challenging, as the structure and the contents of community QA archives differ significantly from the web setting. To address this problem we present a general ranking framework for factual information retrieval from social media. Results of a large scale evaluation demonstrate that our method is highly effective at retrieving well-formed, factual answers to questions, as evaluated on a standard factoid QA benchmark. We also show that our learning framework can be tuned with the minimum of manual labeling. Finally, we provide result analysis to gain deeper understanding of which features are significant for social media search and retrieval. Our system can be used as a crucial building block for combining results from a variety of social media content with general web search results, and to better integrate social media content for effective information access.

Intent Classification of Short-Text on Social Media

“…Therefore, our research question is "How can we mine relevant social intent from an ambiguous, unconstrained natural language short-text document?" By definition, the relevant intent classes meet actionable information needs of an organization in a given context, e.g., resource seeking during crisis response coordination [3], [4]. Table I shows examples of short-text documents and associated potential intent.…”

“…First, informal language use causes ambiguity in interpreting user expressions in short-text messages, weakening predictor-class relationships (e.g., 'wanna help' appears as a strong intent signal but exists in messages of two complementary intent classes, 'seeking' and 'offering'). Second, sparsity of instances of specific intent classes in the corpus creates data imbalance (e.g., our prior study on binary intent classification [4] observed that expressions of 'offering' intent were only a fraction of those with 'requesting' intent (1:7 ratio) during Hurricane Sandy event in 2012). Furthermore, both intent classes of 'seeking' and 'offering' may co-occur within a single message.…”

“…Therefore, the nature of other kinds of complex intent (such as the broad intent class of helping) requires more thorough investigation. The closest works on crisis data analytics on Twitter has dealt with the identification of problems-aid report [3], and request-offer messages [4], [15] using only binary classifiers.…”

“…Using a keyword-based crawling method, we collected a set of short-text documents as tweets from the Twitter Streaming API like prior studies [4]. For a real-world event, we defined a set of relevant keywords.…”

“…Within the crisis domain, our prior study of binary intent classification [4] created request-offer behavior patterns from search logs of domain experts at American Red Cross. In another study [15], we used characteristics of communication from psycholinguistics to define a pattern set of intent expressions.…”