References for biologically plausible models of knowledge representation?

Question

I'm looking for references that deal with the issue of how various kinds of semantic knowledge are (or might be) represented neurally. Most of the discussion of this topic seems skewed by social psychologists, who talk in very high-level terms of how various semantic constructs (such as goals) might be represented based on behavioral observations (cf. Eitam & Higgins, 2010). But it's been hard to find accounts that are both closer to the metal (that consider neural plausibility) and also high-level (that treat semantics more complex than associative strength between generic nodes.)

Can anyone point me to good material on this topic?

References:

• Eitam, B., & Higgins, E. T. (2010). Motivation in Mental Accessibility: Relevance of a Representation (ROAR) as a New Framework. Social and Personality Psychology Compass, Relevance of a Representation, 4(10), 951–967, free pdf

Answer 1

In my experience, the term "semantic knowledge" (or semantic memory or conceptual knowledge) is generally used to refer to knowledge of objects, word meanings, facts and people, without connection to any particular time or place. The neural basis of this kind of knowledge is more or less agreed to depend on a distributed network of cortical brain regions (e.g., Martin, 2007; Patterson, Nestor, & Rogers, 2007). Tim Rogers and Jay McClelland (e.g., Rogers & McClelland, 2004) are among the leaders in developing biologically plausible computational models of semantic memory (and many others are working on it as well).

Goals are usually put in the domain of "executive functions" or "cognitive control" and generally associated with prefrontal (and frontal) cortical regions. I know somewhat less about this area of research, but there are definitely biologically plausible models being developed, including by Botvinick and Plaut (2004) and you might find what you're looking for in the work of Munakata and O'Reilly (e.g., Munakata et al., 2011)

References

• Botvinick, M. M., & Plaut, D. C. (2004). Doing Without Schema Hierarchies: A Recurrent Connectionist Approach to Normal and Impaired Routine Sequential Action. Psychological Review, 111(2), 395-429, free pdf.

• Martin, A. (2007). The representation of object concepts in the brain. Annual Review of Psychology, 58, 25-45, free pdf.

• Munakata, Y., Herd, S. A., Chatham, C. H., Depue, B. E., Banich, M. T., & O’Reilly, R. C. (2011). A unified framework for inhibitory control. Trends in Cognitive Sciences, 15(10), 453-459.

• Patterson, K. E., Nestor, P. J., & Rogers, T. T. (2007). Where do you know what you know? The representation of semantic knowledge in the human brain. Nature Reviews Neuroscience, 8, 976-987, free pdf.

• Rogers, T. T., & McClelland, J. L. (2004). Semantic Cognition: A Parallel Distributed Processing Approach. Cambridge, MA: MIT Press.

Answer 2

I think this recent paper fits your requirements. It considers biological plausibility by showing that the number of neurons required in the proposed method is within a reasonable size for the human brain, and dismisses a series of unreasonable models. Specifically, they create a neural network using the Neural Engineering Framework (NEF) and the Semantic Pointer Architecture (SPA). The network contains 2.5 million neurons to contain a network of 100,000 concepts, which would only require 14.7mm$^2$, in contrast to another model discussed in the paper, which would required 500cm$^2$.

It's also "high-level" enough that it does not just consider relative strengths of nodes, but applies them to analysing vocabulary semantic information. For example, one of the three tests discussed in the paper is to see if the network is "smart" enough to determine that the word "dog" is a synonym for "canine".

The paper itself is kind of hard to swallow if you don't have the background with the NEF and SPA to approach it, so I recommend reading either this tutorial paper to get started or "How to Build a Brain" by Chris Eliasmith.

Reference

Eric Crawford, Matthew Gingerich, and Chris Eliasmith. Biologically plausible, human-scale knowledge representation. In 35th Annual Conference of the Cognitive Science Society, 412–417. 2013.