Towards an embedding-based semantic similarity measure designed for mediation caching

Determining a measure of inter-query semantic similarity is an interesting and challenging task. A reliable inter-query similarity measure has several application domains, including mediation systems, semantic caches, retrieval systems, and recommendation systems. Caching in the context of information mediation will be the focus of this study. We will propose a new semantic similarity measure for queries based on query descriptions, our knowledge of the global schema and source schemas, and embedding approaches. Using unsupervised embedding techniques, we will assign each query a vector in a “semantic vector space”. In this space, embedding vectors of semantically similar queries are designed to be close to each other.