Research On The Semantic Annotation Of Domain With WordNet

The Internet is developing at an amazing speed now. It has become an important channel for accessing to the information and knowledge, and it is gradually becoming a part of people's modern life. At the same time, the expansion of web pages means the expansion of massive data. But the large quantity of valuable information is contained that is not easy to be found, because the data must get the reasonable and effective treatment to tap the valuable information. In order to enable non-structure or semi-structured data to be quickly understood by the computer and get the corresponding treatment, the people put forward the concept of semantic web. The semantic web aims to enable computers to understand the semantic of web documents, which can be shared and reused by the different data source so that people can conduct exchanges and cooperation with the computer. The realization of semantic web needs a wide range of semantic annotations for the massive data. The semantic annotation is actually a publishing process of semantic information that is based on the specific ontology. The semantic annotation is the cornerstone of semantic web.Most of the semantic annotation systems annotate universal concepts but not the special field. And most of them need more or less manually intervenes in the process. Automatic semantic accuracy rate is quiet low. According to defects and inadequacies of the existing semantic annotation system, the article has put forward a combination of the WordNet semantic annotation method in the wine field. First of all, the article introduces a kind of similarity calculation method is based on information capacity in the WordNet and combine it with the similarity calculation method that is based on the edit distance. So it can measure similarity between the named entity and the concept class or the instance in the wine ontology from grammatical and semantic aspects. Experimental results show that this similarity calculation method can get better precision and recall. In combination with the similarity calculation method which is based on edit distance can receive small increasing about precision and recall. After studying a wide range of similarity algorithm in WordNet, we find most of WordNet - based semantic similarity computation method depending on the tree hierarchy of nouns. The article improves a similarity calculation method based on sharing information in order to break the hierarchy. The more semantic elements can be considered in it. The experiment results show that use based on the improved shared information semantic annotation method can get the same precision with based on the information capacity of similarity computation. But, its recall is higher. After combining with the calculation method which is based on the edit distance, the precision and recall can be improved a little. In addition, the article adapts owl format to save the semantic annotation results in a non-embedded method. This way can reduce the difficulties to maintain the results and the results can be made according to demanding of different users comparing with the embedded method.