QMUL-IMC

PhD proposals within IMC

A computational approach to perceptual aesthetics in images
Peter McOwan [e]: The human eye is attracted to beauty. When studying the world around us our brains have the ability to extract information on the aesthetics of what we observe, for example a pretty face stands out in a crowd. Aesthetics is a multifaceted concept; elements by which human observers assess aesthetics in an observed scene include low-level visual measures of symmetry, spatial proportions such as the golden ratio, and global measures such as overall composition. In this project we will develop biologically plausible mathematical models for aesthetic appreciation and test these models predictions against subject performance. This interdisciplinary approach combining philosophy, psychology and computer science allows us to develop a better model of the perceptual process and to invent computer artefacts with a sense of aesthetics for example to assist in automated product design.

Design features for creativity
Nick Bryan-Kinns [e]: Analysis of collaboration has typically focussed on the logistics of information sharing, role assignment, and so on. The next step in understanding complex group activities is to assess the features of collaboration over and above the basic operations required to produce artefacts ? the glue that makes collaboration engaging, enjoyable, and rewarding for the participants. This PhD investigates features of group design in order to develop a framework for understanding creative engagement between people, and to inform the design of systems aimed at supporting group creativity.

Improvisation as a way of understanding groups
Nick Bryan-Kinns [e]: Collaborative improvisation, composition, and performance of music constitute a basic and distinctive form of human interaction in which engagement, innovation, and ensemble co-ordination are key objects of the interaction. Like face-to-face conversation, collaborative improvisation - which can usefully be considered as instant composition - is typically co?present and multimodal, combining musical signals with verbal and visual cues. These properties make musical collaboration uniquely interesting as a basic, underexplored form of human interaction. The nature of collaborative music improvisation means that current approaches to understanding group work, collaboration, and communication are not immediately appropriate in this context. This PhD investigates how developing understandings of group music improvisation can inform understandings of creative collaborations in general.

Multimodal / Multilingual Computer Assisted Language Learning (CALL)
Marie-Luce Bourguet [e]: Educational software systems dedicated to first or second language acquisition are common, but very few systems have been designed to address the needs of young children that acquire multiple languages from birth. In response to this situation, the aim of this research is to explore new interaction paradigms for the design of interactive systems that will specifically support multilingualism acquisition and development. Multimodal interaction refers to the usage of several natural modalities of interaction such as speech, gestures and gaze (input modalities); speech synthesis, prosody and smart graphics (output modalities) to interact with systems. We propose to study how multimodal interaction techniques (modality fusion and modality fission) can be used to address common multilingual phenomena such as language mixing (interactions between the child?s developing language systems) and language dominance or specialization (unequal developments of the languages).

Tools, methodologies and guidelines to support the design, development and evaluation of multimodal interaction
Marie-Luce Bourguet [e]: Recent developments in recognition-based interaction technologies (e.g. speech and gesture recognition) have opened a myriad of new possibilities for the design and implementation of multimodal systems. However, our lack of understanding of how these new modes of interaction can be best combined in the user interface often leads to interface designs with poor usability. Furthermore, building multimodal system prototypes is not an easy task: currently, no tools are available to support the implementation of modality fusion (the integration of several communication channels that operate in parallel) and modality fission (the partition of information sets for the generation of multimodal presentations). Evaluating multimodal prototypes is an even more difficult task as no methodologies or guidelines are available. The main objectives of this research are: (1) To research and develop tools that will support the task of designing and prototyping multimodal systems, across a wide range of modalities, devices and application domains (these tools will take advantage of several modality fusion and fission techniques that will be carefully evaluated in terms of efficiency, flexibility and error robustness); (2) To devise evaluation guidelines, methodologies and metrics for multimodal interaction.

XML for Historical Research
Graham White [e]: This research would be concerned with the development and evaluation of XML-based data formats and inferential tools to aid historical research. Two main problems will be addressed:
a) The representation of the objects of historical research -- people, events, and institutions -- in XML, using a suitably developed ontology.
b) A rigorous separation between objects and evidence: historical evidence is always partial and tentative, and the conclusions drawn from this evidence should always be capable of being modified, or revoked, in the light of new evidence.

The Statistics of Internet Chat
Graham White [e]: The department runs a MUD (http://tcz.dcs.qmul.ac.uk/) which has detailed logging facilities: we have accumulated several GB of logs of interaction in this MUD. It is an extremely large corpus, which we are using extensively for research. The statistics of this corpus are both theoretically and practically important.
Connectivity on the internet is known to have interesting theoretical properties: it is what is known as a scale-free network. It would be theoretically very interesting to know whether or not interaction patterns in this MUD followed a similar pattern: practically, it is an important question, because any empirical work that we might do on interaction in the MUD depends on the statistical properties of the interaction.
Work on this problem will be concerned with
i) empirical investigation of the statistical properties of the corpus: distribution of node degrees, correlations between the activity of participants, and so on.
ii) the development of a model for the statistics of the corpus, and
iii) the validation of the model against the observed properties of the corpus.