Evaluating User Knowledge in Large Scale Online Knowledge Communities

摘要

It is an important knowledge management task to evaluate a user's domain knowledge in a knowledge community. We present a new domain knowledge representation method that considers both user-document associations and document-topic relevance. We provide three alternative user-document association models with varying syntactic and semantic inferences and two alternative document-topic relevance models with different assumptions on knowledge diffusion. We compare the effectiveness of different knowledge representation methods using the combinations of these alternatives. Using a real data set collected from the Sun forums, we find that the vector space model for user-document association combined with the medium-level diffusion model outperform all other model combinations.