Executing Knowledge Graph Initiatives in Organizations – A Field Guide

Half-Day Tutorial @ 15th European Semantic Web Conference 2018

June 3rd, 2018, Heraklion, Greece

 

What is this about

Ever since Google announced that “their knowledge graph allowed searching for things, not strings”, the term “knowledge graph” has been widely adopted, both by the academia and industry, to denote any graph-like network of interrelated typed entities and concepts that can be used to integrate, share and exploit data and knowledge in one or more domains.  Apart from Google, knowledge graphs are found and developed within several prominent companies, including Microsoft, Apple, LinkedIn, Thomson Reuters and others, as an enabling technology for data integration and analytics, semantic search and question answering, and other cognitive applications.

In this tutorial I describe the technical, business and organizational dimensions and challenges that Knowledge Graph Architects need to be aware of before launching a Knowledge Graph initiative in an organization. More importantly, I provide a framework to guide the successful execution of a knowledge graph project, combining state-of-the-art techniques with practical advice and lessons learned from real-world case studies.

What will you learn

During the session I will walk you through 8 key stages of introducing, developing, delivering and evolving Knowledge Graphs in an organization. These are:

Stage 1 – “Knowing where you are getting into”:  In this stage I clarify and disambiguate the notion of knowledge graph, describe what kind of organizations build knowledge graphs and why, and identify the key factors that make the building of a knowledge graph an easier or more difficult task. This will help you judge whether a knowledge graph is an appropriate solution for a given organization and business problem, and have a first idea of its feasibility and viability.

Stage 2 – ”Setting up the stage”:  In this stage I describe 5 key questions that need to have a relatively clear and specific answer before the actual building of the knowledge graph starts: “What are we building, why are we building it, how are we building it, who is building it, and who cares?” For each of these questions I provide some typical answers (drawn from the literature and own experience) but, most importantly, I outline the actions needed to seek and discover those answers. This will help you effectively prepare for the launch of a knowledge graph initiative and have a more accurate and comprehensive idea of its feasibility and viability.

Stage 3 – “Deciding what to build”: In this stage I dive into the details of knowledge graph specification and I show how competency questions can be used to perform a gap analysis between an organization’s knowledge capabilities and needs, as well as how these needs can be scoped and prioritized. This will help you to effectively determine what the knowledge graph needs to contain, as well as in what order this content should/can be realistically delivered.

Stage 4 – “Giving it a shape”: In this stage I discuss the task of schema design for knowledge graphs, using notions and techniques from Ontology Representation and Engineering. My focus is not on explaining a particular language or methodology but rather on identifying conceptual modeling best practices, dilemmas and pitfalls, derived from real-world cases. Particular focus is given to the challenge of uncertainty and vagueness that characterizes real-world knowledge and on approaches for tackling it within knowledge graphs. This will help you design better schemas that are error-free and more maintainable and semantically inter-operable.

Stage 5 – “Giving it substance”: In this stage I discuss the challenging problem of (automated) knowledge graph population by a) describing the different population tasks, b) identifying the factors that make them less or more difficult, and c) providing guidelines for evaluating and selecting already available knowledge extraction systems and tools for this purpose. This will help you to design optimal population pipelines for your graph.

Stage 6 – “Ensuring it’s good”: In this stage I discuss the problem of assessing the quality of a knowledge graph, not only by defining relevant dimensions and metrics, but also by describing and providing insights about important quality trade-offs that have to be made. This will help you not only measure quality but also effectively prioritize the dimensions you need to focus on.

Stage 7 – “Making it useful”: In this stage I discuss two typical applications of knowledge graphs, namely entity disambiguation and semantic search, providing guidelines and best practices for optimizing the latter’s performance through proper knowledge graph design and adaptation. This will help you maximize the usefulness and value of the knowledge graph.

Stage 8 – “Making it last”: In this stage I discuss the challenging problem of knowledge graph maintenance and evolution, providing a framework for detecting, measuring and monitoring concept drift, and suggesting best practices for enabling continuous improvement. This will help you design mechanisms for effectively preventing a developed and deployed knowledge graph from becoming irrelevant over time.

Why should you attend

The tutorial will particularly benefit you if:

You are planning a knowledge graph initiative (academic, industrial or mixed) and you don’t know where to start.

You are planning or already implementing a knowledge graph initiative and you think you have covered all bases.

You work in one or more Semantic Web research areas and you want to know how this work fits into the knowledge graph narrative.

You work in one or more Semantic Web research areas and you want to know what it’s like to apply this work in an organizational context.

You love Semantic Web Languages, Standards and Technologies and you assume everyone applies them.

You believe effective conceptual modeling is only a matter of using the correct language and tooling.

You have managed to maximize your system’s F1 or other performance score for a competition, challenge or paper and you believe this is enough to apply it in an industry setting.

You believe uncertainty, vagueness and other non-Boolean phenomena are not problems that should concern knowledge graphs.

You believe knowledge graph quality is only a matter of defining and calculating standardized metrics.

How can you attend

To attend the tutorial you need to register at the ESWC 2018 conference website.

Tutorial Material

 

Who am I

See About Me

How to learn more

Feel free to contact me for:

  • Getting more details about the scope and content of the tutorial
  • Suggesting particular topics, cases or questions that you would like to see covered in the tutorial.
  • Any other Knowledge Graph related matter.

    Share this