START Conference Manager    

Context-based Natural Language Processing for GIS-based Vague Region Visualization

Wei Chen

ACL Workshop on Language Technology and Computational Social Science (ACL LACSS 2014)
Baltimore, Maryland, USA, June 26 - 26, 2014


Abstract

Vernacular regions such as central Ohio are commonly used in everyday languages, but the fact that their boundaries are vague and indeterministic limits our ability of communicating these regions. This paper introduces a context-based natural language processing approach to retrieve geographic entities from news articles. These entities are used as behavioral samples to map out the location and extent of the vernacular region central Ohio. Particularly, part of speech tagging and tree parsing are employed to filter out candidate geographic entities from sentences. The prepositional logic of context (PLC) from artificial intelligence (AI) is adapted to build a contextual model to decide the memberships of named entities. Results are visualized in GIS using both graduated symbol and kernel density maps.


START Conference Manager (V2.61.0 - Rev. 3312)