农业信息搜索可视化平台研究

RESEARCH ON VISUALISED PLATFORM OF AGRICULTURAL INFORMATION SEARCH

针对传统搜索引擎检索返回结果数量庞大、专业性差且只能为用户提供一维、线性搜索结果的问题,在分析研究农业垂直搜索引擎的基础上,构建农业信息搜索可视化服务平台。基于农业文献,对数据进行信息抽取、关联分析,并设计了一种基于最大距离法选取初始质心的K-means层次聚类算法来发现领域概念间关系;在此基础上,利用信息可视化模型与基于Java的Prefuse插件包为用户提供图形化的结果呈现方式,实现信息的交互控制,优化检索过程。通过实验验证,改进的层次聚类算法提高了领域概念间关系聚类效果的同时降低了聚类总耗时,平台满足用户检索的专业性需求。

Aiming at the problem of traditional search engines that they return a large number of retrieving results,be poor in professional capability and can only provide users with one-dimensional and linear search results,based on analysing and studying vertical agricultural search engines,we constructed the visualised service platform for agricultural information search. On the basis of agriculture literatures,we carried out the information extraction and association analysis on data,and designed a k-means hierarchical clustering algorithm,which is based on selecting initial centroid with maximum distance method,to discover the relationship between domain concepts. Based on this,we used the model of information visualisation and the Java-based Prefuse plugins pack to provide for users a graphical representation means for results,thus realised the interactive control of information,and optimised the retrieval process as well. It is verified through experiment that the improved hierarchical clustering algorithm in this paper improves the effect of correlation clustering between domain concepts and meanwhile reduces total clustering time consumption. The platform can meet the professional demand of users retrieval.