A 943 page document was released on January 3, 2024 filled with court transcripts related to the Epstein case.
Kineviz just released SightXR, a web application which uses Generative AI to create Knowledge Maps, which are holistic representations of documents. They make document exploration significantly faster. They also make AI powered Q&A more reliable, by grounding generated responses with real-world evidence. Knowledge Maps capture the relationships between people, organizations, locations, events, and other concepts. We call those “observations”.
Here’s what I discovered in an hour. An example of Q&A is at the end.
It took about 18 minutes to generate this Knowledge Map from a 943 page document.
A degree calculation gives us the most frequently mentioned Persons: Defendant, Jeffrey Epstein, Ghislaine Maxwell, Ms. Giuffre, Plaintiff, Ms. Maxwell, Sigrid McCawley, and Virginia Giuffre. Here I’ve selected those Persons and their neighboring nodes.
From our selection, we can export a spreadsheet of all the observations generated by the AI about these Persons. There are 1941 observations in the selection.
Here’s a simple example. The observation is “Philip Barfden and Ross Gow sent an email to gmaxl@ellmax.com”. The observation is linked to gmaxl@ellmax.com and “Philip Barden, Ross Gow”.
I’ve expanded the observation to the chunk it was derived from. We can check the chunk text for more context and to verify the observation.
Another way to start exploring is by searching the database for “Jeffrey Epstein”. Here I pulled an observation “It is unclear if Emmy had any sexual involvement with Jeffrey Epstein” and then expanded on Emmy.
I asked “Who were Epstein’s accomplices?” Then I pulled all the context into the workspace, arranged the entities in a grid, and then applied a ring layout.
By transforming this document into a knowledge map, we can quickly get a gist of the content without sacrificing on depth because the context is always present.
Find out more about Kineviz at https://www.kineviz.com/ or contact me at ben@kineviz.com for more information.
