Dr Timothy Cribbin

2001-present : Lecturer, DISC, Brunel University
2007 : PGCert (Brunel) Learning and Teaching in Higher Education
2005 : PhD (Brunel) "Classifying complex topics using spatial-semantic document visualization: an evaluation of an interaction model to support open-ended search tasks" BURA
2000-2001 : LIC funded research assistant at DISC, Brunel University (PI: Dr Chaomei Chen)
1998-2000 : EPSRC funded research assistant at Psychology Institute, Aston University (PI: Dr Stephen Westerman)
1996 : MSc (Hull) Industrial Psychology, awarded Tom Hoyes memorial prize for dissertation (DOI)
1994 : BSc (Portsmouth) Psychology

Research Interests

Visual text analytics, social media analysis, bibliometric analysis, search user interfaces, information visualisation, human-computer interaction

My main interest is in the field visual text analytics (VTA). VTA is sub-class of data mining that aims to create visualizations of semantic structure that lies latent within unstructured or semi-structured collections of textual data. Most of my work has focused on the spatial-semantic or spatialization approach. Spatialization invokes a spatial-semantic (distance-similarity) metaphor, to produce point maps or node-link graphs that summarise the general semantic relationships between documents in a corpus. Spatializations provide the searcher or analyst with both a thematic overview of the corpus and an intuitive context in which to search and explore.

Recent projects have investigated ways to optimise the cognitive plausibility (validity) of spatial-semantic structures as well as to improve the efficiency/scalability of content-based similarity computation. Earlier work involved foundational work on the usability of spatializations for a variety of information seeking tasks (e.g. see Westerman and Cribbin, 2000; Cribbin and Chen, 2001; Chen et al., 2002).

I use a 'bag of words' vector space model approach as the basis for document similarity modelling. There are many ways to improve both the structural properties of both the dis(similarity) matrix and the resulting spatial layout. A particularly effect method for improving the similarity matrix is second-order similarity analysis (SOSA). SOSA transforms the first-order (i.e. term overlap) document similarity matrix to one of mutual neighbourhood (second-order) similarities. This transformation uncovers latent relationships that are not detected by first-order similarity metrics. The more near neighbours two documents share, the more likely they are to be about the same topic, which is reflected in the SOS coefficient. Recent experimental results (Cribbin, 2011) showed that SOSA can produce significantly better topic clustering than latent semantic analysis (LSA) whilst being simpler to apply because it is parameter free. A drawback of SOSA is that the run-time grows cubically with N. However, my experiments also showed that it is possible to reduce run-times significantly, without harming similarity measures, by truncating the similarity vectors prior to matrix multiplication.

Regardless of the quality of the similarity matrix, projecting these complex, high-dimensional spaces onto a 2D plane is a difficult problem. My work has focused on graph-theoretic approaches, which help to deal with the non-linearity and violation of metric assumptions associated with these spaces. I have observed good results using the Isomap method, in which the original dissimilarities are transformed to geodesic distances. These distances are estimated by computing shortest paths within a neighbourhood graph of some kind. Traditionally this might be a k-nn graph, although this presents the problem of selecting the k-parameter - too large and the graph will contain disruptive "short-circuits"; too small and the graph becomes disconnected. My work has shown that minimum-spanning trees (MST) or minimal (q=N-1) pathfinder networks tend to produce equally good spatial-semantic results without the need to tune any parameters (Cribbin, 2006, 2010).

A second, current line of my research is Citation Chain Aggregation (CCA: Cribbin, 2011). CCA is an interaction model I developed to support search and analysis tasks within citation networks. Traditionally, citation chaining activity (footnote chasing and citation searching) is conducted within page-based hypertext interfaces. This results in the focus and context problem, whereby the user is attempting to gain an overview of a complex network of citations (i.e. to find relevant items and determine their relations and relative importance) but can only see a small part of that network at any one time. CCA attempts to solve this problem by means of a three-list view, which displays the aggregation of first-order citation chains (cited<-article<-citing) surrounding a set or 'pearl' of known relevant articles (see below). As more items are added to pearl, differences in the incidence of overlap between their cited and citing articles provide a form of relevance feedback, drastically reducing the size of the search space and avoiding the need to 'navigate', node by node, through the network. See the paper and poster for a more detailed explanation of the concept. You can try CCA for yourself by downloading Oyster search here.

I have recently begun to explore the potential of VTA, including sentiment analysis, as a means of supporting the analysis of social media data (e.g. Tweets, blogs, forums etc.). See this blog for a summary of the work-in-progress on our Chorus infrastructure, which comprises the Tweetcatcher, for managing long-term Twitter queries and Tweetvis for analysing the retrieved datasets. So far, the focus has been within the domain of health informatics, looking for example at public perceptions and responses to health-crises such as the 2011 e-coli outbreak (see the foodRisC project) and recent/current h1n1 flu pandemics. We are also interested in the potential of social media as a source of information to support the evaluation of medical devices (see the MATCH project). A key long-term goal here is to develop tools for social (and mainstream) media analysis that enable social scientists to readily leverage the powerful methods developed by the text mining and information visualization communities within the last two decades.

I currently belong to the People and Interactivity Research Centre (PANDI) and the Centre for Intelligent Data Analysis (CIDA) within the Department of Information Systems and Computing and am also a co-investigator on the MATCH project. If you are interested in reading for a PhD in any of the topic areas mentioned above, please email me.

de Folter, J & Cribbin, T (2012). Facilitating insight into a simulation model using visualization and dynamic model previews. Journal of Visual Languages and Computing, 23(6), 344-353. DOI

Cribbin, T (2011). Citation Chain Aggregation: an interaction model to support citation cycling. In the proceedings of the 20th ACM International Conference on Information and Knowledge Management (CIKM '11), Glasgow (October 24-28, 2011). DOI BURA

Cribbin, T (2011) Discovering latent topical structure by second-order similarity analysis. Journal of the American Society for Information Science and Technology, 62(6), 1188-1207. DOI

Westerman, S.J., Cribbin, T. & Collins, J. (2010). Human assessments of document similarity. Journal of the American Society for Information Science and Technology, 61(8), 1535-1542.DOI BURA (post-print)

Cribbin, T. (2010). Visualising the structure of document search results: a comparison of graph theoretic approaches. Information Visualization, 9(2), 83-97. DOI

_{Cribbin, T. (2006).
Classifying complex topics using spatial-semantic document
visualization: an evaluation of an interaction model to support
open-ended search tasks. Doctoral dissertation, Brunel University,
Uxbridge, UK. BURA}

_{Westerman, S. J.,
Collins, J., & Cribbin, T. (2005). Browsing a document
collection represented in two- and three-dimensional virtual
information spaces. International Journal of Human-Computer
Studies, 62(6), 713-736.
DOI}

_{Morar, S. S., Macredie, R., &
Cribbin, T. (2002). An investigation of visual cues used to create
and support frames of reference and visual search tasks in desktop
virtual environments. Virtual Reality, 6(3), 140-150.
BURA
DOI}

_{Chen, C., Cribbin, T., Kuljis, J.,
& Macredie, R. (2002). Footprints of Information Foragers:
Behaviour Semantics of Visual Exploration. International Journal
of Human-Computer Studies, 57(2), 139-163.

DOI
BURA
(post-print)}

_{Chen, C., Cribbin, T., Morar, S.
S., & Macredie, R. (2002). Visualizing and Tracking the Growth
of Competing Paradigms: Two Case Studies. Journal of the American
Society for Information Science and Technology, 53(8),
678-689.
DOI}

_{Westerman, S. J., Cribbin, T.,
& Wilson, R. (2001). Virtual information space navigation:
Evaluating the use of head tracking. Behaviour and Information
Technology, 20(6), 419-426.

DOI}

_{Morar, S. S., Macredie, R., &
Cribbin, T. (2001). Perceiving depth in desktop virtual
environments: Effects of motion parallax and object placement.
Paper presented at INTERACT 2001, Tokyo,
Japan.}

_{Morar, S. S., Macredie, R. D.,
& Cribbin, T. (2001). A Study of the Relative Importance of
Visual Cues in Desktop Virtual Environments. Paper presented at
HCI International 2001, New Orleans, USA.}

_{Cribbin, T., & Chen, C. (2001,
5-10 August). A study of navigation strategies in
spatial-semantic visualisations. Paper presented at the HCI
International 2001, New Orleans, USA.
BURA
(post-print)}

_{Cribbin, T., & Chen, C. (2001,
9-13 July). Exploring Cognitive Issues in Visual Information
Retrieval. Paper presented at the Eighth IFIP TC.13 Conference
on Human-Computer Interaction, INTERACT 2001, Tokyo, Japan.

BURA
(pre-print)}

_{Cribbin, T., & Chen, C. (2001,
January 21-26). Visual-Spatial Exploration of Thematic Spaces: A
Comparative Study of Three Visualisation Models. Paper presented
at Electronic Imaging 2001: Visual Data Exploration and Analysis
VIII, San Jose, CA.
PDF
(post-print)}

_{Chen, C., & Cribbin, T.
(2001). Visualising and animating visual information foraging in
context. Paper presented at HCI International 2001, New Orleans.
BURA
(post-print)}

_{Westerman, S. J., & Cribbin,
T. (2000). Cognitive ability and information retrieval: When less is
more. Virtual Reality, 5(1), 1-7.
DOI}

_{Westerman, S. J., & Cribbin,
T. (2000). Mapping semantic information in virtual space:
Dimensions, variance, and individual differences. International
Journal of Human-Computer Studies, 53(5), 765-788.

DOI}

_{Westerman, S. J., & Cribbin,
T. (1999). Navigating Virtual Information Spaces: Individual
Differences in Cognitive Maps. Paper presented at UK Virtual
Reality Special Interest Group Conference, Salford,
England.}

_{Cribbin, T., & Westerman, S.
J. (1999, August 30-September 3). Spatial Data Management
Systems: Mapping Semantic Distance. Paper presented at INTERACT
99, IFIP TC.13 International Conference on Human-Computer
Interaction, Edinburgh, Scotland.}

_{Cribbin, T. (1999, August
30-September 3). Spatial Data Management Systems: Human Factors
Perspectives. Paper presented at INTERACT 99, IFIP TC.13
International Conference on Human-Computer Interaction, Edinburgh,
Scotland.}

_{Westerman, S. J., & Cribbin,
T. (1998). Individual differences in the use of depth cues:
Implications for computer- and video-based tasks. Acta
Psychologica, 99(3), 293-310.

DOI}