| 基于逆概念频率的词语相似度计算; Word Similarity Computing Based on Inverse Concept Frequencies | |
| 孙晶 ; 张东站 | |
| 2015-03-28 | |
| 关键词 | 知网 词语相似度 逆概念频率 义原权重 HowNet word similarity inverse concept frequency primitive weight |
| 英文摘要 | 词语相似性度量在服务选择、自然语言处理、文献检索等领域具有重要的作用,目前通用的词语相似度计算方法是利用《知网》对词的概念解释得出词语之间相似度.对《知网》结构进行分析,认为利用《知网》计算词的相似度的方法中概念的4项基本结构的权重应该动态产生,并提出区分度作为衡量4项基本结构的动态权重.在分析现有研究基础上,借鉴逆文档频率(Idf)权重计算思想,认为义原的区分度与义原在所有概念的相应位置中出现次数成反比,提出了一种基于义原出现频次的义原权重计算方法:逆概念频率(InVErSE COnCEPT frEQuEnCy,ICf).通过分析概念的组织结构,计算第一基本义原结构、其他基本义原结构、关系义原结构、关系符号结构中各义原的ICf权重,将4个基本结构中的最大义原ICf权重作为基本结构的ICf权重.利用动态ICf值逼近基本结构的区分度,进而计算词语相似度.通过对真实数据的实验对比可以看出ICf算法能有效提高计算词语相似度的准确率.相比较传统算法平均前160个词准确率从30.74%提高到72.28%,平均召回率从15.87%提高到49.64%.; The word similarity computation plays an important role in service selection,natural language processing,and literature retrieval.Current researches of word similarity are generally based on HowNet.By analyzing the structure of HowNet,we present an idea that the weight of four basic structures of the concept should be dynamically generated during computing the similarity between two words and a method of calculating the weight of primitive based on the frequency.We compute the ICF of each basic primitive in the first basic primitive,other basic primitives,relation primitive and mark primitive through concept structure analyzing,and take the maximum ICF as the ICF of the basic structure.Then we compute the word similarity by using dynamic ICF obtained as the weight of four basic structures.Experimental results show that the accuracy of word similarity calculation is effectively improved.The average accuracy of former 160 words rises from 30.74%to 72.28%,and the recall rises from 15.87%to 49.64%.; 国家自然科学基金(61303004); 福建省自然科学基金(2013J05099) |
| 语种 | zh_CN |
| 内容类型 | 期刊论文 |
| 源URL | [http://dspace.xmu.edu.cn/handle/2288/123241]  |
| 专题 | 信息技术－已发表论文 |
| 推荐引用方式 GB/T 7714 | 孙晶,张东站. 基于逆概念频率的词语相似度计算, Word Similarity Computing Based on Inverse Concept Frequencies[J],2015. |
| APA | 孙晶,&张东站.(2015).基于逆概念频率的词语相似度计算.. |
| MLA | 孙晶,et al."基于逆概念频率的词语相似度计算".(2015). |
| 个性服务 |
| 查看访问统计 |
| 相关权益政策 |
| 暂无数据 |
| 收藏/分享 |
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。
修改评论