1. KNOWLEDGE GRAPHS AT
ELSEVIER
Paul Groth (@pgroth)
Disruptive Technology Director
labs.elsevier.com
Knowledge Graph Industry Panel WWW 2015 http://www.www2015.it/industry-track/
data integration has been a central challenge for companies for a long time
We do data integration but for single domains
It’s already useful to identify and categorize entiies but now we need a global view. Indeed we have a hand curated taxonomies
1. Translational queries
2. Search is entity centric
3. People have become more central with the rise of the Social graph and the importance of personalization and profiles
4. Applications are data driven, we need more and higher quality data predictive models
5. Applications are data driven, extensible APIs
The flexible style of knowledge graphs
Vast fields of unstructured data is the norm
Effort to consume this data, arranging, input checking
Ever more data,
We can’t do all this by hand but we need to use the human annotators we do have more effectively especially when they are experts
Knowledge graphs make it easier to do automated extraction by having a reference point and framework for doing and managing such extractions and putting those puzzle pieces together
And also provides a powerful source of data that our monster models love
Helps solve two key challenges: data integration in a flexible way and most of the methods that we do help us acquire data
Now our there are a couple of things to worry about in this idealic view
Organic – data quality is a central issue with this
I think a powerful thing is the evolution of these knowledge graphs as they grow and change overtime