Natural Language Generation and Analysis

近年来，自然语言处理的研究取得了很大的进展，并有了许多商业应用。语法校正器、机器翻译和信息提取系统等工具现在可以帮助人们并帮助节省时间和成本。它们让我们离人工智能的目标更近了一步。自动语音识别器和智能辅导系统等工具也导致了人机交互的进步。自然语言生成的子领域进展较慢，但自动文本生成开始在应用程序中使用，包括生成指令、医疗保健报告、交互式模拟（用于培训或娱乐）的文本消息等。

Recently, there was a lot of progress in the research area of Natural Language Processing, leading to many commercial applications. Tools such as grammar correctors, machine translators, and information extraction systems can now assist people and help saving time and costs. They are taking us one step closer to the goals of Artificial Intelligence. Tools such as automatic speech recognizers and intelligent tutoring systems also lead to progress in Human-Computer Interaction. The sub-area of Natural Language Generation has seen a slower progress, but automatic text generation started to be used in applications, including generating instructions, health care reports, text messages for interactive simulations (for training purposes or for entertainment), etc. The long term objective of the proposed research program is to facilitate further advances in these directions and to fill in the gap in the area of Natural Language Generation. The proposed short-term objectives are to develop new methods for automatic text generation and for text analysis and classification, by using the results of the analysis stage to guide the generation stage. The novelty of the proposed approach consists in the ways the results and the insights from the text analysis stage will be used in order to develop new methods for text generation. I will focus on several applications for English and French, for which I did some preliminary work in current and past projects: classification of texts by the emotion expressed (such as happiness, anger, fear, surprise, disgust, and sadness) and generation of texts that express a desired emotion; classification of texts by style and generation of texts with a particular style (such as formal or informal); classification of medical texts and generation of texts for medical reports; intelligent text retrieval and database access using keywords or natural language sentences, and generation of natural language descriptions for existing database reports. I anticipate that the new methods and the proposed applications will advance the state-of-the-art in research, while also allowing for technology transfer into industrial applications.