Graph Databases and the Knowledge Graph

Evoloution of the Graph

  • Graph related problems, or those problems that are best analyzed, reported on or solved by using a graph are vast in nature.

  • In the past decade, social graphs have been the de facto standard in analyzing social networks.

  • Knowledge graphs (KGs) have also gained traction in recent years to represent important information in entities and their relationship to each other.

Why use a Knowledge Graph (KG) for your product or Enterprise ?

  • KGs have a wide array of applications that can give your teams significant boosts in productivity.

  • For products, enabling an effective KG within the product can be a game changer in providing market differentiating features.

Successful uses of KGs

google logo beside building near painted walls at daytime
google logo beside building near painted walls at daytime
Google

Google’s knowledge graph organizes and interconnects information about entities, enhancing the retrieval and presentation of relevant content to users. It enables presenting an overview of entities such as individuals, organizations, locations, or objects directly within the search results.

DBpedia

DB Pedia is a knowledge graph that is based on Wikipedia. DBpedia extracts structured data from over 100 Wikipedia language editions utilizing a mature ontology

a white and blue square with a blue and white facebook logo
a white and blue square with a blue and white facebook logo
Facebook

Facebook’s knowledge graph once featured publicly accessible search features based on both the content of the user and their friends’ profiles and the relationships between the user and their friends. Results were based on the friends and interests expressed on Facebook.

Medical Knowledge Graph

Applications include aiding in problems such as finding new treatments for existing drugs, diagnosing patients and identifying associations between diseases and biomolecules.

  • BenevolentAI's knowledge graph: An example of a successful MKG used for drug discovery.

  • COVID-KG: A KG created from multimodal data for drug repurposing related to COVID-19.

  • Patient-centric knowledge graphs (PCKGs): Knowledge graphs tailored to individual patient data for personalized care.

Knowledge Graph Benefits

a group of purple cubes hanging from a metal bar
a group of purple cubes hanging from a metal bar
a stack of stacked blue and white plates
a stack of stacked blue and white plates
What are some benefits ?

  • With your KG in place you can employ ML and LLM techniques that utilize the KG. They can predict or classify new data, answer questions, or fully automate processes based on instructions from the user.

  • One common example usage is through Graph Neural Networks (see the section on that for details)

  • Another example usage is in conjunction with LLMs where the KG and/or vector stores are utilized to provide context utilizing RAG (see the section on Knowledge RAG)

How to construct ?

  • A precursor for any KG is having the data available in some curated form. This can be in the form of a relational database or a set of linked documents in a hierarchy.

  • In constructing your KG it is important to understand the outcomes or use cases you are seeking to solve for. This will help to define a schema for your KG.

  • Extracting the KG from your source databases can be accomplished in a number of ways, including manual curation, rule based extraction, supervised and semi-supervised learning.

References

  1. Medical Knowledge Graph Creation and Usage

  2. Ali Ismayilov1, Dimitris Kontokostas2, S¨oren Auer, Jens Lehmann, Sebastian Hellmann. Wikidata through the Eyes of DBpedia arXiv:1507.04180 2015