Measuring Semantic Relatedness using Mined Semantic Analysis

Mined Semantic Analysis (MSA) is a novel distributional semantics approach which employs data mining techniques. MSA embraces knowledge-driven analysis of natural languages. It uncovers implicit relations between concepts by mining for their associations in target encyclopedic corpora. MSA exploits not only target corpus content but also its knowledge graph (e.g., "See also" link graph of Wikipedia). Empirical results show competitive performance of MSA compared to prior state-of-the-art methods for measuring semantic relatedness on benchmark data sets. Additionally, we introduce the first analytical study to examine statistical significance of results reported by different semantic relatedness methods. Our study shows that, top performing results could be statistically equivalent though mathematically different. The study positions MSA as one of state-of-the-art methods for measuring semantic relatedness.