• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>IEEE Annual Information Technology ,Electronics and Mobile Communication Conference >A knowledge-based word sense disambiguation algorithm utilizing syntactic dependency relation
【24h】

A knowledge-based word sense disambiguation algorithm utilizing syntactic dependency relation

机译：一种基于知识的词语义消除歧义算法，利用句法依赖关系

获取原文
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Accurate recognition of the meaning of words in the given document fragments has a long historical standing in computational linguistics. Because of its importance in understanding natural language semantics, it becomes one of the most challenging issues within this field. In this paper, we present an enhanced knowledge-based word sense disambiguation (WSD) algorithm that manipulates by calculating the semantic similarity between definitions of the target word and an extended syntactic context vector. The actual meaning of the target word is identified as that for which the semantic similarity between definition and extended syntactic context vector is highest. Therefore, while traditional WSD approaches are based on semantic information, our method is built on both semantic and syntactic relations between sentences. This allows it to take advantage of a better use of comprehensive knowledge and more in line with the human being way of thinking. For the sake of improving the performance of WSD, we combine the syntactic importance of the words within the original document piece to provide a sequence that can be used to decide which words should be disambiguated. Finally, the evaluation performed on two standard datasets shows that our algorithm goes better beyond the baselines and has more desirable features than other popular algorithms.
机译：准确地识别给定的文档片段中的单词的含义在计算语言学中具有很长的历史。由于其在理解自然语言语义方面的重要性，它成为该领域最具挑战性问题之一。在本文中，我们介绍了一种增强的知识型词感测消歧（WSD）算法，其通过计算目标字的定义与扩展语法上下文向量之间的语义相似性来操纵。目标词的实际含义被标识为哪个定义和扩展语法上下文向量之间的语义相似性最高。因此，虽然传统的WSD方法基于语义信息，但我们的方法是在句子之间的语义和句法关系中构建的。这允许它利用更好地利用综合知识，更符合人类的思维方式。为了提高WSD的性能，我们将单词的句法重要性与原始文档件中的单词相结合，提供了一种可用于确定应该消除的序列的序列。最后，对两个标准数据集执行的评估表明，我们的算法超出基线更好，并且具有比其他流行算法更好的特征。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号