Natural Language Processing

点击翻译按钮获取中文摘要

This proposal explores ways to support people when they are speaking or writing in their mother tongue or during the process of acquiring these skills in an another language. This research, that we call Interactive Natural Language Generation, takes into account the specificities of the intended user, an adult already having a command of a first language, not a child learning to speak. It also addresses problems such as interface or navigation. Since an important goal of language is knowledge transfer, we deal with how knowledge is structured, represented, accessed and transformed. Language generation theory and practice have been described by two landmark books: one from Levelt, a psychologist describing mental processes, and another one from two computational linguists (Reiter and Dale) showing how to build tools for fully automated text generation. In this proposal, we focus on the production of language with the help of a machine. It is therefore a hybrid approach drawing on knowledge accumulated in psycholinguistics and linguistic engineering, relying additionally on human knowledge, intelligence and skills. This research program, which involves 2 PhDs, 4 MScs and a postdoctoral fellow, will propose solutions by referring to existing theoretical work and will develop practical web-based applications. Each task is independent but explores a specific facet of our overall framework so as to allow emulation and cooperation between students. The originality of this work lies in its difference from the existing body of work, drawn from different fields: natural language generation and natural language processing. It shows how current technology can be used to assist humans in producing language, or in acquiring this skill in a foreign language. We will focus on building tools and on showing what needs to be taken into account for people to generate language with the help of computers. In order to succeed in this research program, we will rely on our experience with natural language generation in related projects as well as on the preliminary explorations and prototypes we have developed within our current NSERC discovery grant.