Unified Workbench for Knowledge Graph Management

This webpage is the supplementary material for Unified Workbench for Knowledge Graph Management (UWKGM) framework. In the video demo, we demonstrate the KG Population use case and the KG construction use case.

• KG Population : Transforming natural language text into structured triples and integrating such triples to an existing Knowledge Graph
• KG Construction : Adding new entities and new relations into a Knowledge Graph

*The video demonstration here is based on the study [1]

**If the video does not start at the first click, please click it again.

Apart from KG construction and KG population, we present the examples of KG revision, which can be fixed by FIXRVE [2] and also show the API screenshot of KG embedding from TorusE [3] for learning embedding representations. Note that, we are currently integrating both FIXRVE and TorusE to UWKGM. Also, we are planning to release the framework to the public soon.

• KG Revision : Correcting erroneous triples in KGs.

The following example is erroneous triple from DBpedia and Corrected triple from FIXRVE.

• Erroneous triple : dbr:The_Fighting_Devil_Dogs, dbo:language, dbr:John_English_(director)
• Corrected triple : dbr:The_Fighting_Devil_Dogs, dbo:language, dbr:English_Language

* dbr = http://dbpedia.org/resource and dbo = http://dbpedia.org/ontology

• KG Embedding : Learning embedding representations of entities and relations in KGs
  • The following screenshots are the demonstration of KG embedding API. In the left screenshot, a user posts data, including KG triples and the hyperparameters for learning embeddings, e.g. margin, learning rate, etc., to UWKGM, while in the right screenshot, UWKGM returns embeddings representations of the entities and the relations. The data exchange format is JSON.