Project Grant 2219712

This two-year, $262,952 project grant from the National Science Foundation's Division of Computer and Network Systems, under the Computer and Information Science and Engineering program (CFDA 47.070), will support the development of a hybrid-architecture symbolic parser and neural lexicon system (HASPNEL) at the University of Puerto Rico. HASPNEL uses a feature-unification parser and structure generator encoded with symbolic AI to judge whether an utterance is grammatical and detect ambiguity at the word or sentence level. A neural network trained on an annotated synthetic corpus constructs a feature-rich lexicon by properly identifying and tagging lexical items and estimating category likelihoods. The system parses only grammatical utterances, recognizes ambiguous utterances by producing multiple syntactic representations, and calculates the likelihood of each. HASPNEL can account for syntactic variation through minor parametric adjustments to grammatical and lexical features. This project aims to advance human language cognitive models, technologies, and applications in education and industry.