一种上下文驱动的信息检索多维匹配模型研究

摘要

为改善信息检索服务效果,提出一种多维匹配模型,在内容维度之外,将信息检索过程中的上下文作为独立维度进行匹配.隐性匹配将各类上下文归为两类:与用户相关的和与文档相关的,两者一一对应.实验表明该模型能大大提高返回结果相关度,亦验证了计算文档相关度时隐性维度有更大权重,即在特定信息需求下,上下文相对于内容具有更大决定性.%To improve the effect of information retrieval services, a multidimensional matching model is proposed in the paper, it matches in information retrieval process the context as an independent dimension besides the content dimensions. The implicit matching classifies all the contexts into two parts: the users correlated and the documents correlated, they are in correspondence one to one. Experiment indicates that this model is able to improve the relevance of returned results greatly, and has also demonstrated that the implicit dimension owns higher weight when calculating the document relevance, i. e. when in specific information requirement, the context is of greater decisive in contrast to the content.