System and method for building relationship hierarchy

摘要

The various embodiments herein provide a method and system for building a relationship hierarchy from big data. The method comprises extracting a plurality of relationships defined between entities from a big data, building relationship recognition models adapted to identify different forms of generic relationships, resolving the relationships by grouping the similar relationships together and separating the relationships which are syntactically and semantically dissimilar and reconciling the resolved relationships to build the relationship hierarchy. The relationship hierarchy comprises groups and subgroups of relationships created based on generic relationship similarity based on a contextual aspect and a specialization aspect using a Language and Domain model. The method of extraction of plurality of relationships from the unstructured data is a self-learning process which uses open information extraction techniques for learning new relationships.

主权项

1. A method for building a relationship hierarchy comprises:
obtaining a plurality of relationships from training samples drawn from big data through open information extraction models; grouping the obtained relationships under generic relationships based on similarity; creating sub-groups of the grouped relationships; tagging the training sample with relationship tags to the occurrences of relationships in training samples; building relationship extraction models from the tagged training samples using one or more machine learning approaches; and providing new training samples from the big data; extracting new relationships using natural language processing based open information extraction models; reconciling the new relationships under at least one of the existing generic relationships or a new generic relationship; annotating training samples with the relationships extracted through the open information extraction models; create one or more new relationship extraction models; populating the relationship extraction models with new relationship extraction models that can successfully identify new relationships; and passing the text data through new relationship recognition models to identify new relationships from data using one or more machine Learning models; extract relationships from the big data using the relationship extraction models; building relationship recognition models adapted to identify different forms of generic relationships; resolving the relationships by grouping the similar relationships together and separating the relationships which are syntactically and semantically dissimilar; and reconciling the resolved relationships to build the relationship hierarchy, wherein the relationship hierarchy comprises groups and subgroups of relationships created based on generic relationship similarity based on a contextual aspect and a specialization aspect using a Language and Domain model.