融合《知网》和搜索引擎的词汇语义相似度计算

发布时间：2018-03-22 02:05

本文选题：语义相似度　切入点：知网　出处：《计算机应用》2017年04期 　论文类型：期刊论文

【摘要】：针对当前《知网》的词语语义描述与人们对词汇的主观认知之间存在诸多不匹配的问题,在充分利用丰富的网络知识的背景下,提出了一种融合《知网》和搜索引擎的词汇语义相似度计算方法。首先,考虑了词语与词语义原之间的包含关系,利用改进的概念相似度计算方法得到初步的词语语义相似度结果;然后,利用基于搜索引擎的相关性双重检测算法和点互信息法得出进一步的语义相似度结果;最后,设计了拟合函数并利用批量梯度下降法学习权值参数,融合前两步的相似度计算结果。实验结果表明,与单纯的基于《知网》和基于搜索引擎的改进方法相比,融合方法的斯皮尔曼系数和皮尔逊系数均提升了5%,同时提升了具体词语义描述与人们对词汇的主观认知之间的匹配度,验证了将网络知识背景融入到概念相似度计算方法中能有效提高中文词汇语义相似度的计算性能。
[Abstract]:In view of the mismatch between the semantic description of the words and the subjective cognition of the words in the current Web of knowledge, under the background of making full use of the abundant knowledge of the network, there are many mismatches between the semantic description of the words and the subjective cognition of the words. In this paper, a method for calculating semantic similarity of words is proposed. Firstly, the inclusion relationship between words and semantic elements is considered. The improved concept similarity calculation method is used to obtain the preliminary semantic similarity results. Then, the search engine based correlation double detection algorithm and the point mutual information method are used to obtain the further semantic similarity results. Finally, The fitting function is designed and the weight parameters are studied by the batch gradient descent method, and the results of similarity calculation of the first two steps are fused. The experimental results show that compared with the improved methods based on knowledge net and search engine, The Spelman coefficient and Pearson coefficient of the fusion method were increased by 5, and the matching degree between the semantic description of specific words and the subjective cognition of the words was improved. It is verified that the integration of the network knowledge background into the conceptual similarity calculation method can effectively improve the computational performance of semantic similarity of Chinese vocabulary.
【作者单位】：南华大学计算机科学与技术学院;
【基金】：国家自然科学基金资助项目(61402220,61502221) 湖南省教育厅科研项目(16C1378,14B153,15C1186) 湖南省哲学社会科学基金资助项目(14YBA335)~~
【分类号】：TP391.3

【相似文献】