Attivio’s Join Finder Patent

Attivio’s latest patent covers one of the most interesting features in the Semantic Data Catalog and the one that always gets wows when we demo it – the automatic join finder. Under the hood, the technology replaces manual processes that could take hours or days with a quick, easy process that takes minutes.

In short, in the Attivio Semantic Data Catalog you select the data sources you are interested in and add them to your data mart. Then you go to the “Unify” menu, and Attivio automatically creates a data model. Right off the bat, join finder will automatically suggest linkages between sources if available.  We source this information from a variety of sources including referential integrity from a database, user feedback and a semantic and statistical understanding of the data.  But as you can see, even with all of this information often there isn’t a one-hop connection available between data sources.

And that is where the real magic of join finder comes into play.  Because we understand all possible relationships between all possible sources, we are able to suggest additional sources to add to the mart.  We call it “Find Linked Sources” in the menu bar.  

Attivio produces the graph using machine learning and infers new foreign and primary keys by analyzing the features of the columns and looking at intersections, column types, and data cardinality. Attivio graphs relationships across databases and data types. For example, Attivio can find references to locations and customers in unstructured content and graph their relationship to structured data sources such as customer orders from a different system.

Because of the underlying graph structure, Attivio can detect hidden relationships between disparate sources.

Once the recommended sources are added, Attivio completes the model. The data mart can then be published as a virtual mart and shared with other teams, or it can be provisioned to analytics and BI tools. 

This technology is so unique that it was awarded US patent # 9507824 B2, titled “Automated creation of join graphs for unrelated data sets among relational databases.” Read the press announcement about the patent.

To see for yourself how the join finder works, you can sign up for a free trial of the Semantic Data Catalog.