automatic classificationof previously unseen proper noun（自动classificationof以前看不见的专有名词）.docVIP

下载本文档

0
0
约2.72万字
约 16页
2017-01-24 发布于四川
举报
版权申诉

automatic classificationof previously unseen proper noun（自动classificationof以前看不见的专有名词）.doc

1、原创力文档（book118）网站文档一经付费（服务费），不意味着购买了该文档的版权，仅供个人/单位学习、研究之用，不得用于商业用途，未经授权，严禁复制、发行、汇编、翻译或者网络传播等，侵权必究。。
2、本站所有内容均由合作方或网友上传，本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺！文档内容仅供研究参考，付费前请自行鉴别。如您付费，意味着您自己接受本站规则且自行承担风险，本站不退款、不进行额外附加服务；查看《如何避免下载的几个坑》。如果您已付费下载过本站文档，您可以点击这里二次下载。
3、如文档侵犯商业秘密、侵犯著作权、侵犯人身权等，请点击“版权申诉”（推荐），也可以打举报电话：400-050-0827(电话支持时间：9:00-18:30)。
4、该文档为VIP文档，如果想要下载，成为VIP会员后，下载免费。
5、成为VIP后，下载本文档将扣除1次下载权益。下载后，不支持退款、换文档。如有疑问请联系我们。
6、成为VIP后，您将拥有八大权益，权益包括：VIP文档下载权益、阅读免打扰、文档格式转换、高级专利检索、专属身份标志、高级客服、多端互通、版权登记。
7、VIP文档为合作方或网友上传，每下载1次，网站将根据用户上传文档的质量评分、类型等，对文档贡献者给予高额补贴、流量扶持。如果你也想贡献VIP文档。上传文档

Automatic Classification of Previously Unseen Proper Noun Phrases into Semantic Categories Using an N-Gram Letter Model Stephen Patel spatel37@ Joseph Smarr jsmarr@ CS 224N Final Project, Stanford University, Spring 2001 Abstract We investigate the ability to automatically classify previously unseen proper noun phrases as drug names, company names, movie titles, and place names. The classifier relies on learning probabilistic features of each of the categories, the most important being a linearly interpolated 4-gram letter model of the proper nouns. In addition, prior probabilities, suffix probabilities, phrase length, number of words per phrase, and na?ve bayes over common word(s) in the phrase all contributed to system performance. Overall system performance was at or near that of human trials conducted, typically in the 94-98+% range for most classification tasks. 1 Introduction 1.1 Overview The performance of many NLP systems suffers because no statistical data has been generated for previously unseen words. This problem is particularly acute in domains where there are large numbers of specific names, or where new names are constantly being generated, such as names of drugs, names of companies, names of places, and so on. Despite having never seen these words before, most people seem fairly capable of identifying them as drug names, company names, etc., which suggests that these classes of words can be modeled using word-level features, such as n-gram letter distributions, suffixes, word lengths and other features. Being able to classify new words in this way would allow natural language text-processing systems to continue unhindered when unknown phrases are encountered. It would particularly aid in the semantic processing of sentences, since assigning an unknown word to a category like “drug names” would impart a great deal of semantic information not present in “unknown word. It could also be useful for automating tasks such as finding stock p