Tools

Method52 Agile Analysis Platform

Method52 is a software platform for the analysis of unstructured text that we co-develop and use to undertake much of our applied research. It is also deployed commercially.

People: Simon Wibberley, Andy Robertson, Chris Inskip, Jeremy Reffin and David Weir

Sub-editor fact checking support

This research focus is in automated fact checking, the process of identifying claims needing verification, and the collection and selection of evidence necessary to determine a claim’s veracity. This work is being carried out in the context of an applications in journalism, specifically supporting the fact-checking elements of the process of sub-editing an article prior to publication. We are interested in understanding what kind of claims require access to external knowledge for verification, and which sources are most effective for this purpose. We are also investigating ways to classify claims based on their characteristics and how to best use these classifications in downstream tasks. 

People: Nestor Prieto Chavana, David Weir and Jeremy Reffin

Interview Support Technology (Intek): Extracting personal questions from in-personal data

At the intersection of Psychology, HCI and AI/NLP/IE, the Intek interview support application aims to provide interviewers with fact-based information in real-time, that can be used to generate near-episodic interview questions. Intek supports the Controlled Cognitive Engagement (CCE) interviewing method, which provides a framework for iterative topic questioning, each progressing from conversational to detailed test questions in order to detect deception. Intek takes a single snippet of information from each topic and returns a variety of cross-referenced information extracted from a variety of sources. The information returned ranges from summary, Q&A and tabular text along with some visual information. The aim is to support a general familiarisation and discourse about a topic as well as material for multiple multi-modal near-episodic tests. The real-time capability allows the interviewer to ask detailed questions about subjects they know little about, without previous preparation.

People: Colin Ashby, David Weir and Tom Ormerod

From Word to World: the MicroMacro Explorer

This project aims to enable humanist researchers to discover, analyse and ultimately to reanimate the lived experience that is represented in narrative accounts. The focus of the initial proof-of-concept work (2018-2020) has been the historical world of London and its environs as captured in the Old Bailey Trials from between, roughly, 1750 and 1820. The work rests on the core Method52 pipeline for classifying textual material, using diverse processes including supervised machine learning, but here incorporating an innovative ‘nested’ model of inheritance between levels of annotation. Building on this, an additional and novel exploratory environment allows for the reviewing and iterative refinement of annotated texts, using proximity as a proxy for semantic relevance, in combination with the visualisation of those complex events in urban space-time that have been identified in the corpus, via a map/timeline interface. Experiments, to date, have explored crime in relation to the worlds of Leisure and Entertainment, the Thames, and the local Watch, as well as attempting to capture the various pulses of the eighteenth century city.

People: Alex Butterworth, Simon Wibberley and David Weir