Location Name Extraction from Targeted Text Streams using Gazetteer-based Statistical Language Models

Proceedings of the 27th International Conference on Computational Linguistics, pages 1986–1997 Santa Fe, New Mexico, USA, August 20-26, 2018. 1986 Location Name Extraction from Targeted Text Streams using Gazetteer-based Statistical Language Models Hussein S. Al-Olimat, Krishnaprasad Thirunarayan, Valerie L. Shalin and Amit Sheth Kno.e.sis Center, Wright State University, Dayton, OH {hussein;tkprasad;valerie;amit}@knoesis.org Abstract Extracting location names from informal and unstructured social media data requires the identifi- cation of referent boundaries and partitioning compound names. Variability, particularly system- atic variability in location names (Carroll, 1983), challenges the identification task. Some of this variability can be anticipated as operations within a statistical language model, in this case drawn from gazetteers such as OpenStreetMap (OSM), Geonames, and DBpedia. This permits evalua- tion of an observed n-gram in Twitter targeted text as a legitimate location name variant from the same location-context. Using n-gram statistics and location-related dictionaries, our Location Name Extraction tool (LNEx) handles abbreviations and automatically filters and augments the location names in gazetteers (handling name contractions and auxiliary contents) to help detect the boundaries of multi-word location names and thereby delimit them in texts. We evaluated our approach on 4,500 event-specific tweets from three targeted streams to compare the performance of LNEx against that of ten state-of-the-art taggers that rely on standard seman- tic, syntactic and/or orthographic features. LNEx improved the average F-Score by 33-179%, outperforming all taggers. Further, LNEx is capable of stream processing.1 1 Introduction In context-aware computing, location is a fundamental component that supports a wide-range of appli- cations (Hazas et al., 2004; Licht et al., 2017). During natural disasters, location is crucial for situational awareness during disaster response (Son et al., 2008). When available, targeted streams of social me- dia data are therefore particularly valuable for disaster response (Munro, 2011)2. For example, the tweet “water level in Ganapathy Colony is around 2 m” refers to a location. But, unless we know where “Gana- pathy Colony” is, the water level data cannot enhance situational awareness and inform disaster response applications such as storm surge modeling or forecasting. However, pragmatic influences on writing style shorten names to reduce redundant content in social media. We call this the location name contraction problem. For example, “Balalok School”, appears in the Chennai flood tweets in contrast to the full gazetteer name–“Balalok Matriculation Higher Sec- ondary School”. Carroll (1983) examined the complex phenomenon of alternate name forms (called Nameheads). He distinguishes between four shortening processes: (1) Appellation Formation, (2) Ex- plicit Metonomy, (3) Category Ellipsis, and (4) Location Ellipsis. Appellation Formation occurs when, for example, the author refers to the location name “The Erie Canal” as “The Canal”. People may also refer to the only airport in the affected area as just “The Airport”. Referring to “University of Michigan” as “Michigan” is an example of Explicit Metonomy. Common ground or shared understanding between the author and the recipient establishes the referent (Resnick et al., 1991). Both Appellation formation and Metonomy pose disambiguation problems, and require context such as the author’s location to resolve. This work …

Content truncated to 500 words