Research

I research human spoken and non-verbal interaction and human-system interaction using Computational Linguistics/Natural Language Processing/AI techniques, with the aim of improving human-artificial agent interaction in terms of quality and ethical outcomes. I mainly focus on trying to answer the following questions:

How do people repair their own speech in dialogue when encountering trouble or when they are disfluent, and how can we automatically detect these events? (See this long-ish talk if very interested, or shorter 25 minute talk and discussion if slightly interested.)
How do people combine speech and physical action when communicating in task-orientated settings? (See this paper on physical coaching or this paper on directed puzzle building)
What are the best models for putting this information into machines (robots, avatars, in-device speech-based systems) to make interaction more natural? (See this 3 minute talk or this paper on human-robot interaction.)
How can such models help us automatically diagnose different types of mental health conditions and cognitive decline? (See this paper on automatically detecting Alzheimer's dementia from speech)

For publications and regular senior collaborators please see my Google Scholar page or the Publications page of this site.

If you are interested in doing a PhD with me, please email me making it clear what your interest is. If your interests don't overlap with mine very much (see my papers and projects), unfortunately I can't guarantee getting back to you.

Current Funded Projects

FLUIDITY in simulated human-robot interaction with speech interfaces - EPSRC 2023-2026, Principal Investigator
April 2023: JOB OPPORTUNITY: 3-year Post-doc Research Officer in Interactive Robotics in VR on the FLUIDITY project at Swansea University
- See the FLUIDITY project's objectives.
- Email me at julian.hough@swansea.ac.uk with an Expression of Interest and CV.
ARCIDUCA (Annotating Reference and Coreference In Dialogue Using Conversational Agents in games) (PI Massimo Poesio) - EPSRC 2022-25, Co-Investigator

Previous Funded Research:

Learning collaboration affordances for intuitive human-robot interaction (PI Kaspar Althoefer) - Alan Turing Institute 2019-21, Co-Investigator
SEVERE (SEVErity Detection in REal Care Comments) (PI Matt Purver) - Care Quality Commission 2018, Named Lead Analyst
RISER (Robust Incremental Semantic Resources for Dialogue) (PI Matt Purver) - EPSRC 2012-13, Named PhD Researcher

People

Research students. Primary/co-supervisor:

Jorge del Bosque Trevino (→ Nanu) - PhD, Modelling Analogy Deployment for Conversational AI Tutoring Systems
Anna Nagele - PhD, Post-human Analyses of Sleep Tracking Devices

Active second supervisor:

Shamila Nasreen - PhD, NLP for dementia diagnosis
Georgia Brown - PhD, The Performed Speech Style of Vivien Leigh

Previous Research students:

Tom Gurion (→ Mintel) - PhD, Head Movement in Conversation
Morteza Rohanian (→ University of Zurich) - PhD, Multimodal Assessment of Cognitive Decline: Applications in Alzheimer's Disease and Depression

Previous Experience

From Jan 2018-Jan 2023 I was a Lecturer at the School of Electronic Engineering and Computer Science at Queen Mary University of London, co-coordinator of the Human Interaction Lab (with Pat Healey) and faculty member of the Computational Linguistics Lab within the Cognitive Science research group in the school of EECS.

From 2019-2021 I was a Co-Investigator on the Alan Turing Institute grant Learning collaboration affordances for intuitive human-robot interaction (PI Kaspar Althoefer), and was a Turing Fellow throughout this.

In 2018, at QMUL I was a researcher on the SEVERE (SEVErity Detection in REal Care Comments) funded by the Care Quality Commission (PI Matt Purver).

From April 2014-December 2017 I was a Wissenschaftlicher Mitarbeiter (post-doc research associate) at Bielefeld University in the Dialogue Systems Group headed by David Schlangen working mainly on the DUEL project, where together with Ye Tian and Jonathan Ginzburg in Paris we investigated disfluency and laughter empirically from real data, by recording and analysing people in conversations which formed the DUEL corpus. The other part of my job was helping to build a virtual coach within the ICSPACE project working with David, Iwan de Kok and Stefan Kopp.

From May 2010-April 2014 I was a PhD student and research assistant in CogSci (or IMC as it was originally) at QMUL with my supervisor Matt Purver. My PhD put self-repair capabilities into incremental parsing and generation. While at Queen Mary I also worked on the DynDial project with Ruth Kempson, Pat Healey, Eleni Gregoromichelaki, Chris Howes and particularly with my supervisor Matt and Arash Eshghi in developing the DyLan dialogue system. I also worked on incremental semantic grammar learning within the RISER project.