Abstract

In this talk we will describe two areas of Natural Language Processing (NLP) which may be of interest to the Networks research group. First, Stephen will describe some recent work (of Brian Harrington) which attempts to build very large semantic networks automatically from large bodies of text. These networks are intended to be general meaning representations which can potentially be used for a variety of NLP tasks and applications; in particular we will focus on the task of determining whether two words or concepts in the network are semantically related.

Second, Sandro will describe his ongoing MPhil research work (a collaboration between the Networks and NLIP groups) which applies topic modelling techniques from NLP to a large collection of Tweets from a social networking tool, which also have geographical information encoded as meta-data. Some interesting patterns are observed with respect to how the conversations Twitter that users have vary according to geographical location.