Vector space methods that measure semantic similarity and relatedness oftenrely on distributional information such as co--occurrence frequencies orstatistical measures of association to weight the importance of particularco--occurrences. In this paper, we extend these methods by incorporating ameasure of semantic similarity based on a human curated taxonomy into asecond--order vector representation. This results in a measure of semanticrelatedness that combines both the contextual information available in acorpus--based vector space representation with the semantic knowledge found ina biomedical ontology. Our results show that incorporating semantic similarityinto a second order co--occurrence matrices improves correlation with humanjudgments for both similarity and relatedness, and that our method comparesfavorably to various different word embedding methods that have recently beenevaluated on the same reference standards we have used.
展开▼
