| With the rapid expansion of the web information resources, the "information overload" problem is getting worse, and the search engine is the basic tool to solve this problem. Because the traditional search engine is for public users, and the search results are for all kinds of industries, so it is difficult for users to find the information they need. To a certain extent, the vertical search engine resolves the problems of low coverage of information for professional areas and specific subjects which a common search engine has. This article is studying how to resolve the key problems while developping a vertical search engine for the agricultural area. This research was supported by the national science and technology supporting plan under grant N0.2006BAD10A1410 named"an agriculture search engine based on Ontology",the national Science foundation of China under grant N0.60774096 named"an agriculture complex adaptive search engine".In the survey of the factors about dissatisfaction while using the earch engine in China, it reveals that the dissatisfaction with the results of the ranking makes up a large proportion of all the factors. This article made a careful analysis about the ranking algorithm. Most of the common search engines adopt the input-output responding model without considering the feedback from the users. This article gives an ameliorated algorithm based on the ranking algorithm of lucene, this algorithm considers multi-factors:(1)to show the importance of the web links,the algorithm takes the pagerank as a ranking factor;(2) the ranking of the website;(3)the feedback from the user: converting the clicks and web reading time to a score as a ranking factor.The experiments show that this algorithm achieved good results.An excellent ranking algorithm also needs a high-quality database,and the high-quality search results must rely on high-quality data.In order to avoid duplicated, similar or incompleted search results, the correctness of data is vital, otherwise the phenomenon of "a first-class search engine, a second-class database" will happen. So the work of data pre-processing is very important, and data cleaning has also been mentioned in a prominent position.In this paper, the data cleaning algorithm which is based on MD5 digital signature is designed. It meets the actual needs in precision, recall and response time, this paper also gives a method to process the incomplete information, namely, the use of other items' attribute value to estimate the missed attribute values as well as the most likely value to fill the missing value.The method has been successfully used in agriculture vertical search engine.Finally, a vertical search engine for agriculture was designed,and the article introduced the general design and the core modules in detail.A system testing shows that the vertical search engine has good performance in the precision, recall and response time and meets with the real needs of the project. |
