How Senzing Uses Machine Learning for Entity Resolution
Did you know that entity resolution delivers more accurate results when it uses machine learning? Watch this video as Senzing CEO Jeff Jonas discusses the different types of machine learning that Senzing® entity resolution uses to get better results out of the box and then to get smarter over time.
Machine learning is often defined as systems that learn through experience. Senzing entity resolution has different styles of machine learning built into its technology, including real time continuous learning and Entity Centric Learning™.
Doing real time learning at thousands of transactions a second over billions of records is non-trivial. It’s why Senzing is the first purpose built real time AI for entity resolution.
Try Senzing entity resolution for free. It’s easy to get started. Take a look at 3 Quick Ways to Explore Senzing Entity Resolution and see for yourself how Senzing machine learning improves the quality of your data.
Edited Video Transcript
Timestamps
0:00 Introduction
0:09 Senzing Entity Resolution Uses Various Types of Machine Learning
0:32 Real Time Continuous Learning for Entity Resolution
0:54 Self Correcting Machine Learning for Entity Resolution
1:25 Real Time Entity Centric Learning with Senzing Entity Resolution
2:02 Senzing: The First Real Time AI for Entity Resolution
2:31 The Senzing Proof of Concept (PoC)
I get asked from time to time, what kinds of machine learning does Senzing use in its entity resolution?
0:09 Senzing Entity Resolution Uses Various Types of Machine Learning
We use one kind of machine learning for how we compare names. We use a different kind of machine learning in the way that we parse global addresses. And then we have our own really specialized real time machine learning – our principle based entity resolution – for when we’re making decisions about when to match two people (entities) together.
0:32 Real Time Continuous Learning for Entity Resolution
Let me say a little bit more about that because it’s what makes Senzing special. I did say real time learning. As Senzing is ingesting data, it’s keeping a statistical distribution of how often it sees an attribute like a phone number or date of birth – and not just how often, but how many people have it – and Senzing uses that information second by second with thousands of transactions per second.
0:54 Self-Correcting Machine Learning for Entity Resolution
This enables Senzing to figure out things in real time like – while passports might be great for entity resolution most of the time – this one passport number (say, 321321321) is no longer very discriminating because 50 people use it. The moment Senzing figures this out, it stops making decisions the same way on that one passport. And it looks backwards in time and asks, now that I know that, should these earlier assertions be reviewed?
1:25 Real Time Entity Centric Learning with Senzing Entity Resolution
That’s one kind of machine learning inside of Senzing. The second kind of real time learning that Senzing does is local to an entity, to a profile. Because of Entity Centric Learning, as new records are being combined and understood to be the same person (entity), what you end up learning over time is things like nicknames.
You may never have seen this in any training data. You have to learn it locally in the moment. You might learn that Ken has a nickname called “Slim” or “Slimmy Boy,” whatever. The moment Senzing learns that it goes, wow, now that we know that, are there any other decisions we have ever made that could have been better, and if so, it fixes them in the moment.
2:02 Senzing: The First Real Time AI for Entity Resolution
Doing real time learning like that at thousands of transactions a second over billions of records is non-trivial. It’s why Senzing is the first real time AI for entity resolution. We’re very excited about that, very excited. In fact, our customers are even more excited about it, because we have organizations downloading our software and finding out that out of the box it’s outperforming their team of people (often a small army) that have been doing homegrown custom entity resolution on their data for a decade.
2:31 The Senzing Proof of Concept (PoC)
Imagine that you pop Senzing out of the box and we’re outperforming your homegrown entity resolution. How? Because we have all these different styles and types of machine learning built into the Senzing technology, and all are pre-configured. So you don’t have to spend any time doing it yourself. We’re excited about that. You can download Senzing and try it for free today. You can do an entity resolution Proof of Concept on a Tuesday.
