How Entity-Centric Learning Improves Entity Resolution Accuracy

Matching data within or across data sources isn’t easy. Watch this video to learn more as Jeff Jonas explains how entity centric-learning improves the speed and accuracy of entity resolution, especially for large data sets.

Entity-centric learning learning is a superior method for resolving entity data that ultimately results in faster, more accurate and smarter entity resolution. In an entity-centric system, resolved records are viewed as single holistic entities. New records are compared to each resolved entity versus record-to-record matching, which is a less accurate approach.

With entity-centric learning systems get smarter as more records are received. In addition, the system detects non-obvious relationships that humans easily miss. Learn more about Senzing Smarter Entity Resolution®. Schedule a call to speak with an entity resolution expert.

Video Transcript

Timestamps
0:00 Intro
0:18 Entity-Centric Learning Improves Entity Resolution Accuracy: Record Matching
0:28 Entity-Centric Learning Improves Entity Resolution Accuracy: Entity Resolution with Entity-Centric Learning
1:17 Entity-Centric Learning Improves Entity Resolution Accuracy: Entity-Centric Learning for Fraud
1:50 Data Matching Entity Bundles with Entity-Centric Learning
2:06 Entity Resolution for Marketing Data Matching
2:19 Entity-Centric Learning

There’s a couple of different methods of entity resolution. One very popular one that the world’s been doing the most of and for the longest time from the beginning of time is record matching.

Entity-Centric Learning Improves Entity Resolution Accuracy: 0:18 Record Matching

That means when a new record shows up, you check all the records you’ve previously seen and see if it matches any of those records. It’s fine sometimes, but not all the time.

0:28 Entity-Centric Learning Improves Entity Resolution Accuracy: Entity Resolution with Entity-Centric Learning

The other method is called entity-centric learning. That’s where when a group of records comes together, and imagine they are rubber banded together, so when a new record comes in, you’re not trying to find the record it matches, you’re trying to find the entity that matches.

I learned this, by the way, years ago when I automated my very first entity resolution system for a credit bureau. They had all of the different records of the different people on cards and they sat in a packet. On the cover of the packet was a summary of all the names and a summary of all the addresses and phone numbers that existed on the cards.

I didn’t know how important that was. But, it turns out, when you get a new record in, you don’t want to check it against each of the cards, you wanted to check it against the summary all of the names and addresses and stuff. That is entity-centric learning.

1:17 Entity-Centric Learning Improves Entity Resolution Accuracy: Entity Centric-Learning for Fraud

In the case of fraud, entity-centric learning is utterly essential. In fact, I would argue to say, if you’re doing record matching, you can’t catch clever bad people, and that’s because clever bad actors don’t use the same name, phone, address, passport on every record.

In entity-centric learning, as you assemble the bits of data and the records that you know about somebody, you’re accumulating all the different ways they spell their name, their nicknames, their aliases, AKAs, the different addresses they’ve used, the different variations they’ve used of their date of birth, maybe sometimes they use their brother’s date of birth, that’s in there too.

1:50 Data Matching Entity Bundles with Entity-Centric Learning

And it’s assembled so you’re really matching a union of features. An inbound record comes in, you’re not looking for records, you’re looking for identities. And entity-centric learning is that process of matching records to bundles, if you will. And when you can do that, you can find clever bad people.

2:06 Entity Resolution for Marketing Data Matching

By the way, while it’s essential for finding bad actors – you have to have it – it turns out it’s also helped with this darn messy marketing data. So, it’s a win-win for everybody.

2:19 Entity-Centric Learning

Entity-centric learning, once you see it, you can’t unsee it. You’re like wow, how did it match this record to this record? There’s nothing in common, and it turns out it’s all the glue and the records in between.

This is something you’re not going to find in synthetic data. You’re going to find this in your real data. You’re going to find this as you add more data and get more diversity and more data sources, and Senzing starts to sing and get smarter and smarter because of things like entity-centric learning. Try it for yourself, run real data, keep an eye out for it. It’s a special thing and it’s a must-have.

Interested in what we're up to?
Subscribe to email updates from Senzing.

Please add your email address to opt-in to be subscribed to our email marketing list. You can unsubscribe at any time. For further information, please view our full Privacy Notice.