A HYBrid symbolic-connectionist processor of natural language semantic relations

In the field of natural language processing (NLP), there are symbolic and connectionist approaches to account for semantic issues, such as the thematic role relationships between sentence constituents. A ldquohybridrdquo option merges both methods: a symbolic thematic theory is used to supply the connectionist network with initial knowledge. This way, benefits of connectionism, such as learning, generalization and fault tolerance are combined with representational symbolic features. A symbolic-connectionist hybrid system called HYBthetasPRED (hybrid symbolic-connectionist thematic (thetas) predictor) is proposed here. Its main purpose is to reveal the thematic grid assigned to a sentence. The connectionist architecture comprises, as input, a featural representation of the words (based on the verb/noun WordNet classification and on the classical semantic microfeature representation), and, as output, the thematic grid assigned to that sentence. HYBthetasPRED ldquopredictsrdquo thematic (semantic) roles assigned to words in a sentence context, adopting a psycholinguistic view of thematic theory.