Parsing of Grammatical Relations for Databases of Spoken Language

Despite the significant advances in syntactic parsing of written text, the application of these techniques to spontaneous spoken language has received more limited attention. The explosive growth of available corpora of transcribed spoken language opens up new opportunities in that direction. High accuracy parsers for spoken language will in turn provide a platform for development of a wide range of applications, as well as for advanced research on the nature of conversational interactions. One concrete field that is ripe for the application of such parsing tools is the study of child language acquisition. In this proposal, I describe a plan for developing a new approach for analyzing the syntactic structure of spontaneous conversational language in parent-child interactions. Specific emphasis is placed on the challenge of accurately annotating the English corpora in the CHILDES database with grammatical relations that are of particular interest and utility to researchers in child language acquisition. This work will involve rule-based and corpus-based natural language processing techniques, as well as a methodology for combining results from different approaches into a high-performance system. One practical application of this research is the automation of language competence measures used by clinicians and researchers of child language development. I will implement an automatic version of one such measurement scheme, providing not only a useful tool for the child language research community, but also a task-based evaluation framework for grammatical relation identification.