The Handbook of Discourse Analysis
The Handbook of Discourse Analysis
复制标题
DOI：
10.1002/9780470753460.ch42
复制
发表时间：
2005
期刊：
影响因子：
--
通讯作者：
B. Webber
中科院分区：
文献类型：
--
作者：
B. Webber
关于话语和对话的计算工作反映了自然语言处理的两个总体目标：·根据计算过程系统对人类理解和自然语言生成进行建模。这一领域的工作通常被称为计算语言学。·使计算机能够分析和生成自然语言，以便提供有用的服务。这一领域的工作被称为应用自然语言处理、自然语言工程，或最近的语言技术。这些目标可以追溯到自然语言处理(NLP)的最早研究和发展，它始于20世纪50年代初的机器翻译工作。早期的机器翻译工作指出了在弱限定领域中处理无限制的、扩展的文本时存在的严重问题。这导致20世纪60年代和70年代初的NLP研究人员将重点放在受限领域的问答对话上，例如格林等人(1961)的棒球比赛、伍兹(1968)的航班时刻表、伍兹等人(1972)的月球岩石分析以及Winograd(1973)的“积木世界”。但是，随着有效语言处理所需的意义表示和推理的发展变得越来越少的语言问题，自然语言处理研究人员的注意力从开发自然语言系统转向解决与个别语言相关的问题--例如开发更快、更高效的解析器；开发更弱的、因此更现实的语法，其复杂性仅略高于上下文无关(参见。Joshi 1999)；开发处理指称表达的方法；模拟通信目标和计划以及它们在语言中的实现等。但现在我们又回到了原点，最近在计算机网络上可用的信息的爆炸性增长，以及语音技术的进步使得对不那么令人沮丧的基于电话的服务设施的需求成为可能，使人们重新将兴趣集中在处理不受限制的扩展文本和对话上。
Computational work on discourse and dialog reflects the two general aims of natural language processing:• that of modeling human understanding and generation of natural language in terms of a system of computational processes. Work in this area is usually called computational linguistics.• that of enabling computers to analyze and generate natural language in order to provide a useful service. Work in this area has been called applied natural language processing, natural language engineering, or more recently language technology.These aims go back as far as the earliest research and development in natural language processing (NLP), which began with work on machine translation in the early 1950s. Early machine translation work pointed out serious problems in trying to deal with unrestricted, extended text in weakly circumscribed domains. This led NLP researchers in the 1960s and early 1970s to focus on question-answering dialogs in restricted domains, such as baseball games in Green et al.(1961), airline schedules in Woods (1968), analyses of lunar rocks in Woods et al.(1972), and a “blocks world” in Winograd (1973). But as the development of meaning representations and reasoning needed for effective language processing became less and less language issues, the attention of NLP researchers shifted from developing natural language systems to solving individual language-related problems–eg developing faster, more efficient parsers; developing “weaker” and hence more realistic grammars whose complexity is only slightly more than context-free (cf. Joshi 1999); developing ways of handling referring expressions; modeling communicative goals and plans and their realization in language, etc. But now we have come full circle, and the recent explosion in information available over computer networks, and demands for less frustrating automated telephone-based service facilities made possible by advances in speech technology, have refocused interest on dealing with unrestricted extended text and dialog.