A Lightweight Approach to Extract Interschema Properties from Structured, Semi-Structured and Unstructured Sources in a Big Data Scenario

Abstract

The knowledge of interschema properties (e.g., synonymies, homonymies, hyponymies and subschema similarities) plays a key role for allowing decision-making in sources characterized by disparate formats. In the past, wide amount and variety of approaches to derive interschema properties from structured and semi-structured data have been proposed. However, currently, it is esteemed that more than 80% of data sources are unstructured. Furthermore, the number of sources generally involved in an interaction is much higher than in the past. As a consequence, the necessity arises of new approaches to address the interschema property derivation issue in this new scenario. In this paper, we aim at providing a contribution in this setting by proposing an approach capable of uniformly extracting interschema properties from a huge number of structured, semi-structured and unstructured sources.