If we accept language as an endogenous system, then we can start discussing representations for language processing which have a basis in endogenous systems. This paper is a start in the direction of showing how such a model might be constructed—drawing from across disciplines such as psychology, psycholinguistics and neuroscience. What we propose, and show the start of, is a model of lexical items which rather than being a list of words and features, is a system of evidence points. We outline the underlying technologies supporting this and describe what each will bring to the model. We believe this model will better support data analysis where language values are involved and show how it could work in cross-language information analysis. The first area of application involves named entities—people, places, associations—which are frequently necessary intelligence analysis data points, but which have confounded systems designed to automatically incorporate them.
» Download Paper [PDF, 28KB]
