Microplanning and Linguistic Realization for Natural Language Generation in a Biomedical Domain

Abstract

The goal of natural language generation (NLG) is to produce understandable spoken or written texts in human language automatically. Microplanning and linguistic realization are important tasks for natural language generation. In this thesis, the microplanning and linguistic realization process for an implementation of a prototype natural language generation system called GenIE are discussed. GenIE uses information about clinical genetics and specific knowledge about a patient given by a genetic counselor to produce the first draft of a patient letter. GenIE uses four modules to accomplish the tasks of microplanning and linguistic realization. Also, a study was conducted to test the writing quality of the letters. Participants were asked to read a copy of a generated letter and make editing suggestions. The changes made to our generated letter were categorized according to what type of microplanning task the change required in order to determine what changes could be made to better the current version of GenIE.