Terms extraction from unstructured data silos

The major challenge that the big data era brings to the services computing landscape is debris of unstructured data. The high-dimensional data is in heterogeneous formats, schemaless, and requires multiple storage APIs is some cases. This situation has made it almost impractical to apply existing data mining techniques which are designed for schema-based data sources in a knowledge discovery in database (KDD) process. In this paper, a tool called TouchR is proposed which algorithmically relies on the Hidden Markov Model (HMM) to extract terms from data silos; specifically, distributed NoSQL databases- which we model as network graph. Our use case graph consists of storage nodes such as CouchDB, Neo4J, DynamoDB etc. The evaluation of TouchR shows high accuracy for terms extraction and organization.