基于复杂网络的汉语相似词挖掘和相似度计算研究

Research of Chinese Similar Words Mining and Similarity Calculation Based on Complex Network

相似词是自然语言中普遍存在的现象，词汇相似度计算是情报学、自然语言处理和信息处理等领域的一项中间步骤。首先，基于大规模语料库构建了汉语词汇共现网络，进而利用复杂网络结构中节点相似的思想来计算词汇的相似度。其次，基于分布假设、上下文语境理论和词汇网络结构的特点，本文提出了基于贡献度折扣的词汇相似度计算方法，该方法不仅考虑了网络边的权重信息，还将节点的全局度特征融合进来。通过节点相似度实验发现，本文提出的基于贡献度折扣的相似度算法要明显好于共同邻居法、Jaccard方法和Sahon方法。最后文章就实验结果及其结论做了详细分析。

Similar words are a common phenomenon in human languages. Word similarity calculation is a middle step in informatics, natural languages processing and information processing. First, a giant language network is constructed based on Chinese large-scale corpus. And then, the node similarity in complex network is used to similar word mining and word similarity calculation. Under distributional hypothesis, context theory and the characteristics of word network, the paper propose a new algorithm named Contribution Discount Similarity algorithm （CDSim） , which can capture not only the edge weight, but also the global characteristic. Compared with the three typical methods of node similarity calculation, such as common neighbors, Jaecard and Sahon, CDSim performs best. Finally, some related experiments and conclusion are discussed.