Dimensions of emotional connotation

I only today discovered your group's existence having followed the progress of the "reasoning with uncertainty" group with interest from afar.

A quick scan of your group's first report reveals no reference to the extensive multicultural and multidimensional work of Charles Osgood on "semantic differential" theory (http://en.wikipedia.org/wiki/Semantic_differential ). That doesn't mean that the work you have cited ignores that work; indeed it may have subsumed it in fascinating ways of which I am ignorant having been away from that field for some thirty years. My comments are therefore quite naive, about the actual activities of your group, but I hope they will not be too unwelcome.

The magnitude of the corpus developed by him and like-minded empiricists in the 1950's through 70's was staggering. Work on establishing the viablity of his intrinsically three dimensional theory was conducted in several continents by hundreds of researchers.

In response to his claim that the "semantic differential captured meaning" some notable linguists replied convincingly that it only addressed "connotative" meaning rather than "denotative" meaning; a viewpoint I happen to share. My own methodologies on establishing the metric space of connotation (as might be appropriate to drive an inferential engine) began with rather simple computational methodologies in the early 1980's: use a monolingual thesaurus with the links between words therein to provide graph theoretic links. (see for example http://portal.acm.org/citation.cfm?id=12456 ) 

We might determine the graph which interconnects the terminology of emotion and then consider embeddings of that graph in various analytical metric spaces (via multidimensional and nondimensional scaling). A simple walk through thesaurus space (using an "improved" version of Roget's 1911 thesaurus for English) may be seen at http://marble.sru.edu/~ddailey/cgi/hyphens?wild . Methodologically, I would suppose that once one has established a metric space for emotively potent vocabulary, then one could investigate, using the typical psycholinguistic methodologies of George Lakoff, Donald Norman and others to develop inference trees and graphs. John angered Mary by playing her favorite love song --> John knew that [.....]

Ultimately, at some physiological level, emotion may be determined by some simple low dimensional space of vectors of neurotransmitters, but as emotion is filtered by cognition and language, and as it influences the inferences we make, it becomes much more graph theoretic and less multidimensional.

I am glad to see the W3C undertake such an initiative;
keep up the good work!

David Dailey

http://srufaculty.sru.edu/david.dailey/