| With the speed of stepping into"information society", full-text retrieval has been widely used in many fields such as enterprise information platform etc. But most of the full-text retrieval systems use static index database or part-dynamic database. This kind of database can't fulfill the request for government or newspaper, because it can't be updated at real-time after the data has been inputted, or can only be updated at a pre-set time. So, it is very needed for the dynamic function of full-text index system.The dynamic function of full-text index system is based on the dynamic function of the creation and updating of full-text index. By contrasting to the traditional model and the emerging model, analyzing each model's merits and defects we discover that inter-relevant successive trees model has splendid time efficiency and spatial efficiency, but the dynamic update efficiency of the index data is not very ideal. In order to improve the dynamic function of the system, this paper made researches in three aspects like the optimization of storage structure, the design of an updating index structure and the application of the distributional memory and parallel retrieval. In the first aspect, we present a new structure by the"page"of operation system and define some algorithms on block, record and document. Data shows that the new structure can make the update of index more flexible and can improve on the dynamic function of the system. In the second aspect, the index structure consists of main index, additional index and deleted file table. The experiment indicates that it has good performance in solving the index renewal problem. In the third aspect, we propose the way of distributional memory and parallel retrieval to avoid the weakness of inter-relevant successive trees. In the end, a full-text retrieval system frame is constructed by integrating all the schemas mentioned above, which based on inter-relevant successive trees and it has nice function. |
PDF Full Text Request