How to process heterogeneous data sources (structure, semi structure and unstructured) in Neo4Js using Ontology driven data integration? Any help /reference is highly appreciated


(Dileepdas830) #1

Hi All

We are trying to do a POC by using Neo4JS to build a knowledge base ,We are trying to integrate heterogeneous data sources(structured , semi structured , Unstructured) in to Neo4 JS using Ontology driven data integration . Why we are preferring Ontology based data integration , we want to inject the domain specific knowledge. We have gone though may of Neo4JS documentation , we could not figure out right direction yet ,any help /reference / direction is highly appreciated .

(Andrew Bowman) #2

Could you provide more detail? I'm not personally familiar with this means of integration, looks like it's usually associated with RDF triple stores.

Relevant articles, though perhaps not exactly what you're looking for:

If you can provide some concrete examples of what you're trying to achieve we can see what options are available.

(Dileepdas830) #3

Andrew Thanks for the response . Let me explain the context. We are trying to do a project POC.

We are trying to integrate multiple data sources (claims , medications, provider details ,authorization details ) , the source are in different format and different scheme and trying to provide the consolidate view of all the activities of a provider - like a how many claims processed/approved/rejected , what is trend of medication of provider - how many generic /branded ?,How many prior authentication got approved/rejected) . We are trying to minimize the manual intervention and automate all the data integration process to Graph DB . If a new sources comes , we should be able to integrate without much manual intervention. Since these are related to health care domain , for the consistency and avoid duplication , we are trying to use ONTOLOGY based data integration of multiple source . What should be best method to use? we are trying to use NEO4J as graph DB .

Any help /Direction is highly appreciated ?

(Andrew Bowman) #4

I did find something that may be a bit more relevant for your case, as it has procedures for ontology mapping:

@jesus.barrasa may be better equipped to provide advice for this use case.

(Jesus Barrasa) #5

Hi, when you say ontology based data, you mean data pre-annotated according to specific ontologies?
If that's the case, I assume it will be RDF data. You can import RDF data using the extension that @andrew.bowman has shared with you.
Once you've imported all the data sources into neo4j, I'd recommend implementing the specific entity resolution rules (or any integration logic) as microinferencing stored procedures that use a representation of your ontologies in Neo4j.
You have some basic examples for SubClassOf and SubPropertyOf type inferences here.

Also related, you can have a look at how to implement ontology-based consistency checking in neo4j in this post. It's a bit old and implementation should actually be based on stored procedures (not available at the time it was written), still, you can get the concept from it.