基于结构和属性特征的政府开放数据检索方法研究

Retrieval Method for Open Government Data Based on Structure and Attribute

[目的/意义]排序是信息检索中的一个重要环节,其核心是如何利用特征构建更为有效的排序函数。目前已提出了多种文档、网页特征用于构建排序函数,但由于检索对象上的差异,针对非结构化信息检索对象的特征难以有效地为政府开放数据的检索排序进行服务。[方法/过程]针对这一问题,提出了一种基于结构和属性特征的政府开放数据检索方法,对政府开放数据检索对象的特点进行分析,设计出尽可能全面反映检索对象的特征,利用新特征学习排序函数,并构建二维偏移量的倒排索引对选择的特征进行索引。[结果/结论]实验结果显示,基于结构和属性特征的政府开放数据检索性能优于传统的本地检索方法。

[ Purpose/Significance ] Ranking is an essential part of information retrieve, and it is a hot research topic that how to use fea- tures to construct more efficient ranking functions. Nowadays there are a lot of features of documents and web pages for constructing rank- ing function, however, the searchable objects to the information retrieval and data retrieval are different, existing features can＇t be effec- tively applied to construct ranking function of the open government data retrieval. [ Method/Process~ In this paper, an open government data retrieval based on features of structure and attribute is proposed. Firstly, analyze the object features of data retrieval. Secondly, design features that can fully as possible to reflect the object of retrieval. And finally ranking functions are learned through new design features. [ Result/Contusion ] To achieve the algorithms, the inverted index are established, where offset is represented by two-dimensional coor- dinates, compared with traditional method. The experimental results show the feature rank model is more feasible than local rank model.