Pangeon的动态

Transcript

If you were to design A more general tool, the bridge again that same bridge that goes from unstructured to struction, how would you describe that tool? I think. That structuring, you know, you have to decide what fields you want from it and that can change overtime. You always link back to the source material. And I think that's a crucial part of this is that when we are extracting data, you can't just make a claim that a certain fish lives in a certain place and say we got it from a paper. You have to be able to attribute where that information came from. Ideally, you know, a paragraph number on a page on a specific document. And so I think the combination of pulling out these insights, keeping the link to the source information so you could go back and get it in a different shape or have someone audit it to see if it said the right thing is extremely important. And if you did that at scale, the types of queries and insights that we could run on these types of things would be something we've never seen before with the technology we've had in all of these unstructured format.