Biologically Plausible Connectionist Prediction of Natural Language Thematic Relations

In Natural Language Processing (NLP) symbolic systems, several linguisticphenomena, for instance, the thematic role relationships between sentence constituents,such as agent, patient, and location, can be accounted for by the employment of arule-based grammar. Another approach to NLP concerns the use of the connectionistmodel, which has the benefits of learning, generalization and fault tolerance, amongothers. A third option merges the two previous approaches into a hybrid one: a symbolicthematic theory is used to supply the connectionist network with initial knowledge. Inspiredon neuroscience, it is proposed a symbolic-connectionist hybrid system calledBioèPred (Biologically plausible thematic (è) symbolic-connectionist Predictor), designedto reveal the thematic grid assigned to a sentence. Its connectionist architecturecomprises, as input, a featural representation of the words (based on the verb/nounWordNet classification and on the classical semantic microfeature representation), and,as output, the thematic grid assigned to the sentence. BioèPred is designed to “predict”thematic (semantic) roles assigned to words in a sentence context, employingbiologically inspired training algorithm and architecture, and adopting a psycholinguisticview of thematic theory.