Introducing the First Conversational AI for Entity Resolution
By Jeff Jonas, published June 6, 2023
There’s a new breakthrough in the field of entity resolution called, “Conversational AI for Entity Resolution.”
Large Language Models (LLMs) and generative AI are fundamentally changing how people interact with computers. Everything. Everywhere. Even entity resolution.
Today, we are announcing another first: Conversational AI for Entity Resolution™ or Conversational Entity Resolution™.
What is Conversational ER?
Want to find a person or organization in your system? Want to understand more about that entity and its connections? Or get an explanation about why the records matched? Just ask!
If you’re the administrator, use natural language to prepare and load your next dozen data sources. You can also perform data stewardship conversationally e.g., forcing two records apart or two records together by asking.
Organizations deploying Conversational ER™ with Senzing® real-time AI for entity resolution are going to get even faster time to value and gain a significant competitive advantage.
BACKSTORY: I’ve been building entity resolution systems into business workflows since the early 1980s. That’s 40 years! Over these four decades, there’s been lots of, what I’ll call, “incremental progress” in entity resolution methodology. The biggest leap before today, at least on my journey, was in 2009 when I invented principle-based entity resolution. This is what makes our Senzing technology the unique critter it is: a purpose-built real-time AI for entity resolution that is smart out-of-the-box and self-trains, self-tunes, fixes the past … all while scaling like mad.
Conversational ER is the next giant leap forward.
This is exactly what we are doing.
Later this week we will start releasing open source code that demonstrates the power of Conversational ER. Your data scientists and developers will find this inspirational as they build their projects.
Stay in the know – sign up here.
Want to see what we’ve done with a beta Senzing plugin for ChatGPT?
Locate Data Sources Conversationally
All entity resolution projects start with loading data. Imagine if you could find available data sets and prep them for loading conversationally? Popular commercial systems, such as CRMs will likely need no guidance. Proprietary homegrown systems will get mapped faster with Conversational ER.
Want to experiment with some public data sets? That conversation might look something like this:
Load Data Conversationally
Load all data sets at once or one at a time. Here’s what a conversation to load the first one might look like using Conversational AI for Entity Resolution:
Review Sample Matches Conversationally
To see example MATCHES, even as data is loading, you could use Conversational ER to ask the following:
Find Possible Matches Conversationally
Curious about records that did NOT match, but were very close? Here’s what asking for POSSIBLE MATCHES might look like using Conversational AI for Entity Resolution:
Review an Entity's Details Conversationally
Want to know more about an entity’s details? It’s easy with Conversational ER:
Learn How an Entity Was Created Conversationally
Here’s how to find out how an entity was constructed using Conversational AI for Entity Resolution:
Discover Who is Related to Whom Conversationally
Wondering who is “related” to an entity? It’s easy to explore an entity-resolved graph with Conversational ER. In this example, we’re using a simple natural language request to find out if other entities share an address:
Govern Data Conversationally
Sometimes humans need to override an entity resolution decision, e.g, join records together or take them apart. Here’s how an authorized users will execute a data governance request conversationally:
Manage Operations Conversationally
Administrators will use natural language requests to manage a system. You’ll be able to ask questions about data sources already loaded, total number of records in the system, current load performance and more. In this example, the question is about the status of the latest load request: