Microtext Processing

缩微文本处理

• 批准号: 371717-2013
• 负责人: Khoury, Richard
• 金额: $ 1.09万
• 依托单位: Université Laval
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2017
• 资助国家: 加拿大
• 起止时间: 2017-01-01 至 2018-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=635051
• 关键词: Microtext; Processing

The term "microtext" describes a type of text document that is very short (typically one sentence or less), is written in an informal manner and unedited for quality, and includes some metadata. Microtexts have become omnipresent in today's world, and include online user comments, Facebook newsfeeds and Tweets, and search queries. By their very nature, microtexts are a very different challenge from traditional Natural Language Processing (NLP). Their short length hinders NLP methodologies that rely on word frequencies and statistics, and the abundance of spelling mistakes, typos, and unconventional abbreviations make standard linguistic resources unhelpful. Microtext processing should be seen as a new branch of NLP.This proposal brings together four parallel five-year microtext research programs. The first aims to develop microtext understanding algorithms specialized to the task of query understanding and answering using corporate knowledge bases. The second aims to develop microtext understanding algorithms specialized to the task of very precise event detection in Twitter. The third aims to develop microtext understanding algorithms specialized to the task of mining microblog discussions, specifically in the context of retrieving software development information. And the fourth aims to develop a "military Watson" that can monitor military chat and inform officers of crucial international events occurring. All four projects are done in collaboration with external research and industrial partners. Moreover, they all feed into a single unifying long-term objective: to pioneer and develop the new field of microtext processing, and create a core of microtext expertise. The development of microtext processing will be the next major milestone in Semantic Web and Web 2.0 development. The amount of microtext generated today is absolutely massive: for example, there are 4.7 billion web queries asked on Google and 340 million Tweets posted every day. Web users actively seek out and embrace new microtext technology. The strategic importance of this research for the Canadian economy cannot be overstated, and our multiple partner companies and organizations will benefit directly from our results.

术语“缩微文本”描述了一种非常短（通常是一句话或更少）的文本文档，以非正式方式编写且未经质量编辑，并且包含一些元数据。缩微文本在当今世界已变得无处不在，包括在线用户评论、Facebook 新闻和推文以及搜索查询。就其本质而言，缩微文本是与传统自然语言处理 (NLP) 截然不同的挑战。它们的长度较短阻碍了依赖词频和统计的 NLP 方法，并且大量的拼写错误、拼写错误和非常规缩写使标准语言资源毫无帮助。缩微文本处理应该被视为自然语言处理的一个新分支。该提案汇集了四个并行的五年缩微文本研究项目。第一个目标是开发专门用于使用企业知识库进行查询理解和回答任务的缩微文本理解算法。第二个目标是开发专门用于 Twitter 中非常精确的事件检测任务的缩微文本理解算法。第三个目标是开发专门用于挖掘微博讨论任务的微文本理解算法，特别是在检索软件开发信息的背景下。第四个目标是开发一个“军事沃森”，可以监控军事聊天并向军官通报正在发生的重要国际事件。所有四个项目都是与外部研究和工业合作伙伴合作完成的。此外，它们都致力于实现一个统一的长期目标：开拓和开发缩微文本处理的新领域，并创建缩微文本专业知识的核心。缩微文本处理的发展将成为语义网和Web 2.0 发展的下一个重要里程碑。如今生成的缩微文本数量绝对是巨大的：例如，每天在 Google 上提出的网络查询量为 47 亿次，发布的推文数量为 3.4 亿条。网络用户积极寻求并接受新的缩微文本技术。这项研究对加拿大经济的战略重要性怎么强调都不为过，我们的多个合作伙伴公司和组织将直接从我们的研究结果中受益。