Andrea Salfinger, Caroline Salfinger, Birgit Pröll, Werner Retschitzegger, Wieland Schwinger,
"Pinpointing the Eye of the Hurricane - Creating A Gold-Standard Corpus for Situative Geo-Coding of Crisis Tweets Based on Linked Open Data"
, in John P. McCrae, Christian Chiarcos et al.: Proceedings of the LREC 2016 Workshop "LDL 2016 ? 5th Workshop on Linked Data in Linguistics: Managing, Building and Using Linked Language Resources", Seite(n) 27 - 35, 2016
Original Titel:
Pinpointing the Eye of the Hurricane - Creating A Gold-Standard Corpus for Situative Geo-Coding of Crisis Tweets Based on Linked Open Data
Sprache des Titels:
Englisch
Original Buchtitel:
Proceedings of the LREC 2016 Workshop "LDL 2016 ? 5th Workshop on Linked Data in Linguistics: Managing, Building and Using Linked Language Resources"
Original Kurzfassung:
Crisis management systems would benefit from exploiting human observations of disaster sites shared in near-real time via microblogs, however, utterly require location information in order to make use of these. Whereas the popularity of microblogging services, such as Twitter, is on the rise, the percentage of GPS-stamped Twitter microblog articles (i.e., tweets) is stagnating. Geo-coding techniques, which extract location information from text, represent a promising means to overcome this limitation. However, whereas geo-coding of news articles represents a well-studied area, the brevity, informal nature and lack of context encountered in tweets introduces novel challenges on their geo-coding. Few efforts so far have been devoted to analyzing the different types of geographical information users mention in tweets, and the challenges of geo-coding these in the light of omitted context by exploiting situative information. To overcome this limitation, we propose a gold-standard corpus building approach for evaluating such situative geo-coding, and contribute a human-curated, geo-referenced tweet corpus covering a real-world crisis event, suited for benchmarking of geo-coding tools. We demonstrate how incorporating a semantically rich Linked Open Data resource facilitates the analysis of types and prevalence of geo-spatial information encountered in crisis-related tweets, thereby highlighting directions for further research.
Forschungs-