Research On Multi-Keywords Retrieval Based On HowNet

With the rapid expansion of Internet information, Internet is becoming the main channel for accessing information, and information retrieval has become one of the main purposes as people browse internet. However, how to get exact information that user really what from such huge of information resources has become an important problem. The existing search engines often use mechanical string matching search technology, when in dealing with single keyword, it can basically meet user's requirements, but when in dealing with the multi-keywords, the problem emerged. First, because this search technology only based on string matching, the keywords which on the later place are more likely to be ignored. Furthermore, even if the all of keywords are matched, but it did not pay attention to each keyword's importance degree, this caused the search results inconsistent with the intention of users, thus the precision of information retrieval is reduced.As multi-keywords must be dealt with, a powerful knowledge database system is needed. Hownet is a bilingual general knowledge-base describing relation between concepts and relations between the attributes of concepts. It provides enrichment resource for Language Processing. Therefore, This paper selects HowNet to process the multi-keywords, so as to make retrieval is no longer based on a simple mechanical keyword matching, but on the relationship between the keywords,to solve the low precision problem existing in the current search engines.According to seriously study HowNet, a model about Multi-keywords Awareness based on HowNet about Chinese is proposed. There are three modules which are applied to this model: Word Sense Disambiguation, Multi-keywords Semantic Relationship Awareness, and Compound Word Processing. Word Sense Disambiguation: the phenomena of polysemy word are prevalent in the natural language, it restricts and affects the Semantic Relationship Awareness. Therefor, this paper poposes the five correlation factors of Semantic Relativity, and disambiguates the polysemy words by calculating semantic relativity. Multi-keywords Semantic Relationship Awareness: according to Chinese characteristics, this paper poposes HowNet-based multi-keywords semantic relationships, establishes the"kernel key-word"and"multi-level weight"concepts, and sets different multi-level weight according to different semantic relationship, so as to improve the accuracy of query. Compound Word Processing: as for the unknown words in HowNet database, called compound words, a new Conceptual Combination based on semantic relationgship awareness is presented which to determine the correct DEF, It effectively solve the problem which compound word can't analyze semantic relationship.In order to validate retionality of the model, the three modules are experimented, and a Chinese meta-search engine model based on multi-keywords awareness is put out. From the results of the experiment, the multi-keywords awareness modle is proved right.The search engine with awareness modle has high precision.