Abstract: The old object-based rank algorithms are not satisfactory because of ignoring the Web features of the object and neglecting the correlation between objects and user’s searching. An improved H-PopRank algorithm is proposed in this paper aiming at these shortcomings by taking the following factors into account such as the Web features of the object, the correlation between objects and searching. The experiments indicate that the H-PopRank algorithm can improve recall level precision average and user’s acceptance.

Key words: object-level rank, correlation, recall level precision average

摘要： 当前的对象级排序算法或是忽略对象的Web特征，或是忽略与用户查询的相关度，效果都不太理想。针对这些不足，综合考虑对象的Web特征以及与查询词的相关度等因素，给出改进的H-PopRank算法。实验表明，该算法可以提高查全率级平均查准率和用户满 意度。

关键词: 对象级排序, 相关度, 查全率级平均查准率

CLC Number: 

• TP301.6