基于领域复合语义的产品结构非精确匹配模型与算法

Inexact product structure matching algorithm based on multiplex domain semantics

产品结构间的相似性问题是具有内部递归嵌套特性的复杂对象之间的非精确匹配问题。产品结构间的相似性不但与其组成部件的具体嵌套结构有关,而且还与每个节点相关联的属性集以及产品结构所依赖的企业特有的一套稳定的产品概念系统有关,产品概念系统提供了产品间相似的一类重要背景语义知识。构造了产品结构的属性集、产品概念系统、部件嵌套结构三元特征模型,并提出基于此三元特征的复合语义相似度计算模型。该计算模型是语义距离、属性集异同和下层结构的语义相似度复合3大组成部分的综合。分析证明了该模型可以充分、合理、全面地表达产品结构间的相似性。提出了相似度计算模型的非递归算法,分析了其执行性能,并讨论了该算法使用中的一些配套条件。

Similarity of product structures is an inexact matching problem among complex information objects with inductive structures. The similarities related not only to deductive nodes, but also to attribute set associated with each node, and specific background product concept （ontology） system belonging to enterprises. The product concept system offered a rich set of background semantics knowledge to estimate the similarity. A triple characteristics model was constructed which was composed of attribute sets of product structure, product inductive structure and product background concept system. Based on this model, a comprehensive similarity computation model based on multiplex domain semantics was constructed, which was the combination of several sub-- indicators, such as semantic distance, similarity among feature sets and similarity among inductive structures. Analysis showed that this model was a better choice to estimate the similarity among product structures. A non--recursion algorithm to realize the model was proposed, and the operation performance of the algorithm was studied, as well as the preconditions required for the algorithm to be applied in real world were also discussed.