Semantic similarity attributes for Data Cloud - A Case Study in MIDAS

• Babacar Mane UFBA
• Wita dos Santos Rocha UFBA
• Edmilson Lima UFBA
• Daniela Barreiro Claro UFBA

Resumo

Distinct Data as a Service (DaaS) and Database as a Service (DBaaS) providers can store and manage heterogeneous data in the cloud. Data are accessed from the Software as a Service (SaaS) through an automated solution named MIDAS (Middleware for DaaS/DBaaS and SaaS), which is an intermediate communication layer between SaaS and DaaS/DBaaS. During the providers' evolution, a SaaS query may be affected by DaaS attributes' change. The challenge is to ensure data cloud services consistency for consumers can access the correct attribute when querying a DaaS. Thus, we propose a method to match semantic similarity among DaaS attributes. The method combines two steps to track semantic similarities of DaaS attributes. The first step is done by Cosine and Jaccard measures (edge-counting approach) to identify the number of edges linking two attributes to find the syntactic or lexical similarity, and the second step performed by Information Content (IC)-based measures with WordNet to assess the semantic similarity if edge-counting measures are unable to provide parity between DaaS attributes. Our method is implemented in MIDAS middleware as a proof of concept, and we perform some experiments to evaluate three criteria: overhead, performance, and correctness of our approach. Results showed that we are in a potential direction to provide semantic interoperability between SaaS and DaaS.

Palavras-chave: Semantic Similarity, DaaS, SaaS, DaaS Attributes