Inducing Information Structures for Data-driven Text Analysis
Andrew Salway, Samia Touileb and Endre Tvinnereim
ACL Workshop on Language Technology and Computational Social Science (ACL LACSS 2014)
Baltimore, Maryland, USA, June 26 - 26, 2014
Abstract
We report ongoing work that is aiming to develop a data-driven approach to text analysis for computational social science. The novel feature is the use of a grammar induction algorithm to identify salient information structures from an unannotated text corpus. The structures provide richer representations of text content than keywords, by capturing patterning related to what is written about key terms. Here we show how information structures were induced from texts that record political negotiations, and explore their potential use for analyzing relations between countries and negotiation positions.
START
Conference Manager (V2.61.0 - Rev. 3312)