摘要 为提高基于Agent的信息检索系统在海量的网络信息检索中查询准确率,提出了基于多兴趣Agent层次结构的检索系统模型（IRHOMIA,information retrieval system based on hierarchically organizedofmulti-interest Agent）,模型对查询信息进行了兴趣预测并生成了用户兴趣项权重向量,输入到训练过的神经网络并把输出层生成向量中的每个值与给定的阈值进行比较来确定将查询任务分配给其他兴趣Agent或者是拥有相应资源的查询工具。试验表明,IRHOMIA在预测用户兴趣的效果和推荐搜索信息的准确率方面比传统的检索系统以及单兴趣Agent检索系统IRHOIA有5%以上的提高。
Abstract：In order to increase the searching accuracy in the huge web resources for the information retrieval system based on Agent, an information retrieval system based on hierarchically organized multi-interest Agent（IRHOMIA） is presented. After making the interest predictions on user＇s query information, the model will generate weight vectors of user＇s interest item, then input the vectors to the trained neural network, and compare the given threshold value with each value of the vectors generated by the network＇ s output level. Through the comparison, the searching tasks will assign to the other interest Agent possessing the searching tools of corresponding resources. Through the experiment, it shows that, in the aspects of predicting users＇ interest, IRHOMIA is more effective than the traditional information retrieval system and the single-interest Agent retrieval system. Moreover, IRHOIA increases the accuracy over 5% in the recommendation of the searching information.
2008, Vol. 29 
