Multi-strategy In The Enterprise Integration Of Heterogeneous Data Semantic Matching

Semantic-based enterprise heterogeneous database integration needs to judge which attributes share the corresponding semantics. Having compared the approaches in and out of our country, we propose a multi-strategy approach for heterogeneous data semantic matching. This approach combines several strategies to insure the effectiveness of analysis by importing the holistic attribute information. The researches that we focus are as follows:1) Multi-strategy semantic matching approach for heterogeneous data: thereal semantic of attribute should be expressed by holistic attribute information. By combining several strategies to analyze the holistic attribute information, we can insure the effectiveness of analysis.2) Strategy A—Name matcher based on WordNet: By importing semantic dictionary: WordNet, we propose an approach for computing semantic similarity based on WordNet.3) Specific dictionary expression: In order to make up the flaw of WordNet, we build a specific dictionary to reinforce effectiveness of strategy A.4) Strategy B—Semantic classifier based on SOM neural network: Bytranslating schema information and data value of attributes into inputvectors to implement the classification of attributes. We propose a method to improve initialization of weight by computing the metadata's sensitivity.5) Merging method for multi-strategy semantic matching results: By using strategy A and B, we get two semantic matching sets. We propose a bi-directional revision method for merging multi-strategy results and get good jugement.6) Parse technology for XML document: We need to use the attribute information stored in the XML document, so we build a parser to implement the function.