自然语言的自动分析与生成简介

A Brief Introduction to Natural Language Understanding and Generation

自然语言的自动分析(理解)和生成构成了自然语言处理研究的全部内容。但这两个直觉上对称的任务在实际中却并没有平等的地位,前者受到更多的关注和讨论,后者的研究则相对薄弱。本文对自然语言自动分析和生成各自的目标、面临的问题,所采用的基本方法等做了概要的介绍,并对如何认识二者之间的关系做了初步的讨论。

As a subfield of Artificial Intelligence, the aim of Natural Language Processing is to communicate between computers and human by natural language. In order to realizing it, computers should be able to understand natural language and generate natural language automatically as well. This paper gives a very brief introduction on the basic framework of NLU and NLG respectively, including the tasks, challenges and architectures of the two fields. Intuitively, the NLG can be viewed as the inverse process of NLU. The former, however, has received much less attention and research work than the latter. While there are many reasons which have been mentioned to explain why this might be so, this paper figures out a new image to illustrate the relation between them which is somewhat different from the past viewpoints.