基于模糊聚类的自然语言语义特征  

Semantic feature of natural language based on fuzzy cluster

作　　者：毛婷[1] 杨敬辉[2] 杨晶东[3] 

机构地区：[1]上海第二工业大学国际交流学院,上海201209 [2]上海第二工业大学经济与管理学院,上海201209 [3]上海理工大学光电信息与计算机工程学院,上海200093

出　　处：《辽宁工程技术大学学报（自然科学版）》2013年第1期81-84,共4页Journal of Liaoning Technical University (Natural Science)

基　　金：国家自然科学基金资助项目(60874002)

摘　　要：针对复杂语境下自然语言语义特征提取、匹配精度和实时性较差的问题,提出了模糊聚类、单亲遗传搜索匹配算法相结合的新方法,通过对候选特征点进行模糊聚类处理,使其分布在高斯差分图像的灰度轮廓线边缘,利用单亲遗传算法找到满足约束条件全局最优语义特征,把所有语义特征进行分类,并给出分类依据.研究结果表明:此语义特征匹配算法在未知语境环境、语义特征频繁变化的环境具有很强的鲁棒性,能够在自然语言处理过程中实时准确识别段落中的语义特征.In a complex language environment, there are some problems associated with the semantic feature extraction from natural language, matching accuracy and real-time capability. This study proposes a new feature extracting algorithm which combines fuzzy cluster and partheno-genetic. Firstly, the candidate features are processed using the fuzzy cluster algorithm to enable those extracted features to be distributed around gray outline of Gaussian difference image. Secondly, the globe best semantic features, which satisfy constraints, are found based on partheno genetic algorithm. Finally, all those founded features are classified and the classification reasons are also given. The study shows that the algorithm is strongly robust under unknown language environment, such as part of speech and semantic feature. Therefore, the algorithm is able to extract semantic features from the paragraphs in real-time and accurately.

关 键 词：模糊聚类 单亲遗传 局部最优 特征匹配 语义特征 模板匹配 隶属度函数 免疫平衡 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]