Progress Report: 8_29_08

Nick Davis

Here is a review of some of the activity that has built up to where we are now.

- Last semester we reviewed multiple robotic literatures (for a reference list see: http://cognitiverobotics.pbwiki.com/Other%20work%20in%20the%20field ) , with emphasis on Paul Cohen. His work has to do with creating a natural semantic robot (creating meanings based on the robots experiences as opposed to functional semantics: pre programmed reaction based meanings). A review of this work can be found at: http://cognitiverobotics.pbwiki.com/General+Project+Proposal this description includes a summary of image schemas and perceptual symbol systems, two important theoretical concepts to consider when thinking about creating a natural semantic AI.

-Jim Suffolk did a good deal of programming last year creating an agent. A summary of this work can be found at: http://cognitiverobotics.pbwiki.com/Jim+Suffolk

-As Dr. Brandt mentioned in class, we are also striving to make the agent have a third person representation of itself in a separate pane so to speak, a kind of internal map with a self representation. We feel this is the first step to creating a naturally semantic agent. ---There are many aspects to consider when dealing with these issues, so everyone will be able to contribute in some way; to offer some insight into this process. The theoretical group will hopefully guide the programmers, as the programmers ground the theory. It should work out nicely.

-One overarching goal of the cognitive robotics project is to integrate the AI research with the natural language processor that Matt Elliott has created. The parser, as Matt explained in class, is based on a brute force approach: exhausting every possible permutation of word class ( noun, article, adverb etc.) into phrases ( noun phrase, verb phrase etc.) into overarching phrases like ditransitive phrase. After this process, the parser will see where these words can fall in a stemmatic realization and again, it tries every possible combination. This is a computationally expensive process, and we are currently working towards getting the funding for a powerful computer so we can implement the parser on a large scale. The current parser objectives are creating a stemmatic database (which requires parsing a large amount of text…hence the need of a new computer) This data base will begin to construct a worldy knowledge base from the propositional content of sentences. If the sentence “The red barn is next to the green house” is parsed, the program now knows that barns can be red and this particular barn is red, barns can be next to houses etc. and it creates a knowledge network based on this paradigm. For more information on the intricacies of the parser see: http://stemma.pbwiki.com/A+Stemmatic+Parsing+Project

-There is also another project in parallel with the parser which is called temporal diagramming. The premise here is that we create a program that records natural diagram production and is able to play that diagramming back to the user in order to present information in a more dynamic way. This will also serve to research the components of diagrams overall. We can research this creative process by labeling or annotating certain termporal portions of the diagram. For example, when to things approach each other we highlight a certain amount of time and type “with” or if one object is moving to another we type “to.” Next, we enter these word into the stemmatic parser’s database so that when the language processor parses a text, it can diagram out words that it has the diagrammatic correlate based on manually entered data from the temporal diagramming program. Lately, Dr. Brandt has been entertaining the idea that creativity has a strong correlate with diagrammatics. If we are able to enable the agent some access to these diagrammatic components in a way that is naturally semantic, we may have some interesting results.

Essentially, there are some different projects that are going on in the laboratory for applied research in cognitive semiotics (LARCS : http://www.case.edu/artsci/dmll/larcs/ ) that could inform and benefit from the work on robotics we are doing in class.

Stemmatic Syntax will likely play an important role in this class. For this reason, it may be pertinent to visit: http://stemma.pbwiki.com/Stemmatic+Theory and the stemmatic syntax wiki in general.

I am more than happy to meet with anyone to discuss potential projects, contributions, stemmatic theory, or general concerns. Email: nxd32@case.edu or call 440-506-0399