Study And Implementation On Intelligent Action Of Search Engine

Nowadays, the information on internet is updated with each passing second; and the searching result is expected to meet more and more precise requirements of users. All these invite challenges to the capacity of search engine. Therefore, the study on intelligent search engine deserves much attention in the current web data mining research field. In order to solve many problems existing in the search engine, such as inaccu-rateness etc, various search engines have been developed, including the directory search engine, the full-context search engine and the mate search engine. And in this way, artificial intelligence theories, technologies and methods are applied to this field. This paper applied the intelligent action adopted in books searches to the current research of search engine, and achieved the following results:1. Intelligent action of library information searches is included in five aspects: memory, comprehension, perception, logic reasoning, and practices. By analysing similarities between library inforamtion searches and internet information searches, five kinds of intelligent actions of the full-context search engine are found, and the three-net intelligent action model(Outernet→ Internet → Innernet) was proposed. An example points out that the model is useful.2. For the sake of matching the concept and meaning between the query and web pages, some researches about the web page of knowledge expression, and knowledge retrieval and knowledge inference were conducted with FCA(Formal Concept Analysis). They include three parts: (1) It was based on the definition of the LOW ADJACENCE set and UPPER ADJACENCE set in which words of every web page contain all or at least parts of the key words in user's query. Besides, starting from the equivalence class by the =, ∩ and (?) of attributes and objects of the formal concept, the method and algorithm constructing the concept lattice of LOW ADJACENCE were studied on many- continuous-valued formal context. (2).The method and algorithm retrieving Or-rules and AND-rules between user query and non-user query were analyzed from the concept lattice of the LOW ADJACENCE. (3). According as the basic theories of Formal Concept Lattice, the match model, method and algorithm between the user query and web pages in UPPER ADJACENCE set based on uncertainty reasoning were presented by Or-rules and And-rules. The time complexity of these algorithms is analyzed.3. In the perspective of morphology, syntax, semantics of natural language, a research work to match the english web page and the user query was discussed. This paper puts forward that the derivation process for identifying words by regular grammaris a linear resolution substantively. Furthermore, in certain condition, irregular grammar can be converted to regular one and the algorithm is given. In the light of regular grammar, we studied a method of morphology analysis, proved its completeness, and offered the algorithm to realize the program.