Random and Irregular Effusions of Dave Cochran's Thinking Meats

Dave W.H. Cochran

(Permanently) Under Construction!

Welcome to my rambling page of gibberish.

When I'm not dressed as an evil faery, I'm a PhD student in the Cognitive Systems Group in the School of Computer Science at the University of St. Andrews, on a 3.5 year funded Studentship under the Supervision of Jedi Master, Prof. Rens Bod. My second supervisor is the Sith Lord, Dr. W. Tecumseh Fitch. I've just completed the MSc in Developmental Linguistics, in Linguistics and English Language at the University of Edinburgh, and it is my intention to maintain my Edinburgh connections, in order to bring together the computational, Data-Oriented approaches of the Cognitive Systems Group with the theoretical and experimental work of the Developmental Linguistics Research Group and the Language Evolution and Computation Research Group in Edinburgh. This is part of my ongoing master plan to never leave university - because study rocks, research rocks, and real jobs suck.

My CV is here. Last updated, 6th May 2007.

RESEARCH INTERESTS

This list doesn't get any shorter, except perhaps by using more general terms as a form of data compression - helpful terms like "everything in the entire world". Some of these are things I actually know a fair bit about, others are things I'm just starting to get into;

My primary research interest is in developing Data-Oriented approaches to language (in which the regular behaviour of cognitive systems, such as syntactic processing, are driven directly by the statistical regularities present in a corpus of concrete previous experiences, without any extraction of abstract representations therefrom) into a full-spectrum theory of the human language faculty, and in particular in developing it as an approach to Historical Linguistics. To this end, I am interested in developing Iterated Learning Models using Data-Oriented language learning and using agents. This is a long-term goal, which requires that a number of subgoals first be achieved:

Data-Oriented models of First Language Acquisition; this in turn requires three subcomponents;
- Vocal babbling, as the emergence of crossmodal connections between Data-Oriented models of acoustic perception and motor control of the vocal tract.* This comes in three sub-parts.
  - Darwinised Data-Oriented Parsing (DDOP): An unsupervised DOP model designed as a Genetic Algorithm, in which subtrees are replicators under a selection pressure for generalisability. This is what I ~~ought to be~~ am working on right now.
  - Motor Babbling; the redeployment of DDOP to a model of Motor Control. I have a notion to implement this in a six-legged Lego robot, and have it learn to walk.
  - Vocal Babbling; the redeployment of the Motor Babbling model to an artificial vocal tract - learning the mapping between the phonological patterns of the ambient language and motor commands.
- The emergence of syntax, as driven by the need to handle mismatches between phonological and semantic constituency.*
- The development of social cognition, as the emergence of crossmodal connections between Data-Oriented models of vision, motor control, proprioception and affect.**
Data-Oriented Generation. I am especially interested the relation of language to other cognitive modalities. This was the topic of my MSc thesis, but requires considerable further attention.
Generally, the integration of multiple cognitive modalities in Data-Oriented modelling; Data-Oriented models of consciousness.
The development of suitable co-operative language-mediated tasks for virtual agents in Data-Oriented Iterated Learning Models.
The development of mathematical and computational tools for the analysis of the behaviour of attractors in epigenetic and glossogenetic subtree-space.

Besides this, my latest fun obsession is the question of whether we can identify common principles of informational and representational dynamics among computationally and substratally diverse "world modeling systems"; that is to say, systems, such as individual human brains, whole scientific communities, machine learning systems, ant collonies, etc, which model a limitlessly complex world using limited resources.

I am also interested in Indo-European Studies, with a particular emphasis on methodological issues arising from the interaction of linguistics with other disciplines, in the Philosophy of Mind, particularly concerning the philosophical basis of Cognitive Computational Modelling, and also the development of a form of Neutral Monism as a response to the problem of qualia, and in the Philosophy of Science, particularly regarding the role of distributed cognition/swarm intelligence in the scientific process, the strategic nature of theory-construction, the role of algorithmic concepts in the special sciences, and the roles of emergence, reduction and epiphenomena in scientific explanation.

* These two form the basis of my PhD work.
** Don't you think this would make a lovely basis for a post-doc?

WRITINGS AND DOINGS

27th May 2005, "Using Stochastic Tree-Substitution Grammar in Iterated Learning Simulations as a way of Approaching Issues in Diachronic Syntax"; this is a talk I gave to the MSc DevLing group, with invitees from the MSc Evolution of Language and Cognition group, telling them all about what research I want to do from now until I'm old and grey. I gave a condensed version of the same at the College of Arts and Social Sciences Postgraduate Conference at the University of Aberdeen on Thursday the 23rd of June 2005.

31st May 2006, "Exemplar-Based Phonological Bootstrapping"; Talk given at "Computational Linguistics meets Language Acquisition", at the University of St. Andrews, introducing a novel hypothetical account of First Language bootstrapping. Actually doing the modelling and experimental work to back this up will probably be most if not all of my PhD. An earlier version was given as a talk to the Developmental Linguistics Research Group at Edinburgh on the 11th January.

15th-16th of June 2006, "Neutral Monism, or something near enough"; poster presented at MIND 2006, at the University of Sussex, outlining my ideas regarding the mind/body problem, which I reckon can be satisfactorily resolved by a non-reductive form of Neutral Monism.

August 2006, I completed my first big modelling project, for my MSc dissertation, supervised by Simon Kirby; developing the Unmediated Data-Oriented Generation model outlined in my 2005 DevLing talk above, but with a couple of simplifications to make it doable in 3 months; that the visual stimuli to be described are just simple 1-D arrangements of lines and dots, and it only generates NP’s, not whole sentences.

7th December 2006, "Data-Oriented Picture Parsing", talk given at the 3rd Scottish Perception Meeting at the University of Stirling, presenting my idea for a Data-Oriented Image Parsing algorithm.

10th March 2007, "Algorithmic Thinking and the Strategic Use of Abstraction", talk given at Sci-Phi'07, the Science and Philosophy Symposium, at Stony Brook University, Manhattan. -Working through some thoughts about the role of algorithmic concepts in the special sciences. Abstract here.

24th March 2007, "Access-consciousness as network structure in a system of fragmentable, recombinable multi-modal exemplars", talk given at PPNB'07 the CONTACT Graduate Conference on Philosophy of Psychology, Neuroscience and Biology, at the University of Bristol. Abstract here.

26th May 2007, "Unmediated Data-Oriented Generation", talk at NODALIDA 2007, the 16th Nordic Conference of Computational Linguistics, University of Tartu, Estonia (24th-26th May 2007), giving an overview of my MSc Dissertation work.

21st June 2007, "Darwinised Data-Oriented Parsing: Subtrees as replicators, DOP as a Genetic Algorithm", poster at "Machine Learning and Cognitive Science of Language Acquisition", University College London (21st-22nd June 2007). Abstract here

23rd June 2007, "How Computers can turn Philosophy into Science; Wittgenstein, Data-Oriented Modeling and the Virtuality of the Concept", talk at ECAP 2007, the European Computing and Philosophy Conference, University of Twente, Enschede, Netherlands (21st-23rd June 2007).

2-5th July 2007, "Emerging Network Structure in a Darwinised Data-Oriented Parser", at Complex'07, the 8th Asia-Pacific Complex Systems Conference, Gold Coast Australia; to be confirmed. Abstract here.

13th-17th August 2007, Rens and ran a workshop, "Exemplar-Based Approaches to Language Acquisition and Use", at ESSLLI 2007, Trinity College Dublin (6th-17th August 2007). If you've read this far without getting a brain haemorrhage, you should probably have gone to it.

Comments are always welcome on any of my papers.

CONTACT

You can email me at davec at cs dot st-andrews dot ac dot uk. Go on, it'll be a laugh. Just remember, there are only two stupid questions in the world - the one that's never asked, and the one that you don't bother to check on Google before bugging someone about it.

"Morality Ruins Art"