Abstract
We present a perception-based probabilistic model of language processing.
Our model employs frequency analysis and pragmatic information for
the interpretation of input sentences. The frequency analysis
relates words and semantic constructs, while pragmatics are used
to help identify specific semantic categories within a sentence.
Our model is realized as a semantic network that learns associations
between works and semantic categories from sample sentences. To
show the feasibility of the model we have developed a natural
language interface for a subset of DOS. This pilot system functions
effectively, providing evidence for the merit of our model.
