Routing of Queries in Social Information Retrieval Using Latent and Explicit Semantic Cues

Abstract

Social Information Retrieval can be interpreted as querying the private information spaces of others within one's social network. One of the crucial steps in such a search approach is to identify the set of potential information providers to route the query to. In this experiment, we compare various routing mechanisms based on topic models (Latent Dirichlet Allocation, LDA), Explicit Semantic Analysis (ESA), and traditional metrics like Term Frequency (TF) and Term Frequency-Inverse Document Frequency (TF-IDF) to identify expertise using a publicly available data collection with 1, 400 scientific abstracts including author information, queries, and relevance judgments. The abstracts are interpreted as knowledge profile in a social information retrieval scenario. Our results suggest that both LDA and ESA can solve the routing problem, whereas the LDA-based approach and a new ESA approach considering links between semantic concepts perform best on the tested dataset.