Research On Correlation Sorting Algorithm Based On The Object-Level Search Results In Relational Database

With the development of the Internet, web search engine has achieved great success, the user can use simple keywords to find information which they need. The relational database is the main form of database, it uses the structured query language to make retrieval, and requires the user to master certain query language and database model. However, keywords retrieval can let the users get rid of the SQL statement bondage. This is a very natural need, that is, let the relational database support efficient keyword queries.When compared to the Internet search engine, relational database has new features, for example:semantic relationships exist between tuples; The attribute value of database hidden the equivalent or transfer relationship; The tuple of database is the short-text, and so on. So some information retrieval methods are not suitable for relational database, there is a need to research a set of correlation sorting algorithms which are suitable for relational database characteristics. But these are tuple-level keywords retrieval in relation database. This article designs a kind of object-level correlation sorting algorithm in view of the relationship between the characteristics of the database and the characteristics of information retrieval. This method solves the problem of decentralized information which is the disadvantage of tuple-level information retrieval. So, the technical route of this thesis is, building full text index on relational database, and according to the schema graph of database to make information integration to get the object. And then, do the keywords retrieval on the constructed object, and make correlation sort according to the results of the retrieval.The correlation of sorting algorithm, which is put forward in this thesis, first, it needs to find the transmission relationships between attribute values. The more attribute values appear, the more closely relationship between the attribute value and the keywords. Then, using the information entropy method for the attribute weight distribution. The size of information entropy is related to the distribution of the situation, through calculating the information entropy to reflect the current attribute value distribution, then, can find the relevant situation between attribute value and keywords, getting the relevance score of the information retrieval. Second, it needs to consider the structure characteristics of each object, including the tuple and the edge between tuples to get correlation score of database structure. Putting them together to get correlate score.This thesis uses this method to design an overall framework of object-level correlation sort on the retrieval results of relational database, and implement a alogrithm. Making experiment on the data set which in the field of mobile phone for the alogrithm, the results verify that the availability and the feasibility of the algorithm. This sorting process can get not only object information which contain the keywords, but also can make the difference between the objects which contain the same keywords. Compared to the traditional sorting algorithm, the method which this thesis used can effectively improve the effect of the relational database keywords retrieval sort.