Unsupervised Grammar Induction of Clinical Report Sublanguage

Clinical reports are written using a subset of natural language while employing many domain-specific terms, such a language is also known as a sub language for a scientific or a technical domain. In this paper, we present a method which automatically induces a grammar for the sub language of a given genre of clinical reports from a corpus of reports with no annotations. The method first identifies the semantic classes of the clinical terms used in the reports, then it induces a grammar that is based on these semantic classes and part-of-speech tags. Experiments show that the induced grammar is able to parse novel sentences and obtains a reasonable accuracy.