ITR: American National Corpus: A Primary Resource for Linguistics Research

ITR：美国国家语料库：语言学研究的主要资源

• 批准号: 0218609
• 负责人: Nancy Ide
• 金额: $ 28.51万
• 依托单位: Vassar College
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2002
• 资助国家: 美国
• 起止时间: 2002-09-01 至 2005-12-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0218609&HistoricalAwards=false
• 关键词: ITR; American; National; Corpus; Primary

With National Science Foundation support, Drs. Nancy Ide and Randi Reppen will conduct a three-year project to annotate extensively a 10-million word portion of the American National Corpus (ANC). The ANC consists of both spoken and written language from North America across a range of registers, such as planned speeches, conversations, fiction, and newspapers. This research project uses techniques from both computational linguistics and corpus linguistics to annotate the ANC for a range of grammatical and semantic characteristics. Specifically the project seeks to accomplish three major objectives: 1) develop automatic tools for annotating various elements and structures in the corpus; 2) create a 'gold standard' portion of the ANC, consisting of 10 million words in which the markup, annotation, and parts of speech have been hand-validated; and 3) describe the conceptual and meaning relations among words in the ANC within the framework of the 'semantic web', thus greatly enhancing analysis and retrieval capabilities. The investigators are to carry out this research through a variety of software programs (many created specifically for this project), and through extensive human/computer interaction to hand-validate the computer assigned labels. This research project is important for several reasons. First, the resulting corpus will be the first publicly available tagged corpus of spoken and written American English. Second, because the annotation of the corpus will be hand-validated, the resulting product will approach 100% accuracy. With this carefully annotated 10-million word corpus, language researchers will be able to address a number of structural and linguistic relationships across texts that previously could not be addressed. Since the corpus will be hand-validated, researchers can use this information to develop models for processing previously unseen texts. The ANC corpus will be readily available to researchers via the web. In addition to the annotated corpus, the project will make available to researchers a suite of tools designed to retrieve information from the corpus.

在美国国家科学基金会的支持下，南希·艾德博士和兰迪·雷普恩博士将进行一个为期三年的项目，对美国国家语料库(ANC)1000万字的部分进行广泛注释。非国大由来自北美的一系列语域的口语和书面语组成，如计划中的演讲、对话、小说和报纸。本研究项目使用计算语言学和语料库语言学的方法对ANC进行一系列语法和语义特征的标注。具体地说，该项目寻求实现三个主要目标：1)开发用于标注语料库中各种元素和结构的自动工具；2)创建ANC的由1000万个单词组成的“黄金标准”部分，其中的标记、注释和词性已经过手工验证；以及3)在“语义网”的框架内描述ANC中单词之间的概念和意义关系，从而大大增强分析和检索能力。调查人员将通过各种软件程序(许多是专门为该项目创建的)进行这项研究，并通过广泛的人/机交互来手动验证计算机分配的标签。这个研究项目之所以重要，有几个原因。首先，由此产生的语料库将是第一个公开可用的美国英语口语和书面语的标记语料库。其次，由于语料库的标注将是手动验证的，因此结果产品将接近100%的准确率。有了这个精心注释的1000万字语料库，语言研究人员将能够解决以前无法解决的跨文本的一些结构和语言关系。由于语料库将是手工验证的，研究人员可以利用这些信息来开发处理以前未见过的文本的模型。研究人员将很容易通过网络获得非国大语料库。除了带注释的语料库外，该项目还将向研究人员提供一套工具，旨在从语料库中检索信息。