W-PathSim: Novel Approach of Weighted Similarity Measure in Content-Based Heterogeneous Information Networks by Applying LDA Topic Modeling

In information retrieval, similarity measure or top-k similarity searching had been extensively researched. Similarity search supports to find the most relevant information in a large-scale collection of datasets, especially, with large-scale heterogeneous information networks (HINs) which is composed by multiple types of object and relation. There are studies related to similarity search applied in HINs, the “PathSim” is one of a remarkable work of Sun et al. which is based on meta-path for calculating the similarity between objects in multi-typed linking information networks. However, there is also a shortcoming of PathSim in weighting the “path instance(s)” of defined meta-paths in similarity scoring between two objects. The shortage of evaluating the weight of linked connections between objects might influence the output quality. In this paper, we present W - PathSim model, which applies the L atent D irichlet A llocation (LDA) topic modeling for generating the weighting attribute for the object’s links. We conduct experiments on real DBLP and Aminer datasets in order to demonstrate the effectiveness of our proposed model.