Conversation

Karina Nguyen
@karinanguyen_
A demo of something I've been hacking on 👀 OSINT investigators have to parse large volume documents in their flow, looking for hidden clues and useful context. Definitely think we need more ML tools to help extract + visualize key information:
Embedded video
0:14
43.3K views
33
811
Karina Nguyen
@karinanguyen_
right now it's extracting key characters and their actions in the document & context, but thinking about creating timelines, relationships btw people, and possibly maps/locations.
5
62
Karina Nguyen
@karinanguyen_
This was much inspired from my personal investigative projects (both legal & journalistic) — supply chain analysis, military chain of command, illegal fishing. The tool shouldn’t drive investigations, rather the story should drive the tool’s flexibility.
1
15
Karina Nguyen
@karinanguyen_
Balancing journalist’s explorative control of the document is a key challenge here and one way to tackle this is to make AI decisions interpretable for humans to guide the direction of their investigations. I’m skeptical of making QA / prompt-based interfaces for this reason!
4
13
BenB
@benbvi3ws
Replying to
@karinanguyen_
Thanks for getting back to me! Yes this is something, a need, that arised a few times already, to identify war crimes individuais or to rule out different disinformation narratives. But right now, we're just stocking videos, not processing them in any way. Need more resources :/