University of Glasgow at TREC 2014: Experiments with Terrier in Contextual Suggestion, Temporal Summarisation and Web Tracks

In TREC 2014, we focus on tackling the challenges posed by the Contextual Suggestion and Temporal Summarisation tracks, as well as enhancing our existing technologies to tackle risk-sensitivity as part of the Web track, building upon our Terrier Information Retrieval Platform. In particular, for the Contextual Suggestion track, we propose a novel bundled venue retrieval approach and experiment with text-based summarisation for building the venue description. For our participation to the Temporal Summarisation track, we propose a general framework for performing summarisation over time and two new real-time filtering approaches that leverage the semi-structured nature of news articles to enhance summary coverage. For the TREC Web track, we investigated a novel risk-sensitive learning to rank approach that is based on hypothesis testing and examined approaches that selectively apply di↵erent retrieval techniques based upon the query, with the aim of minimising risk.