The example [1] travel diary is not perfect though -- not a criticism, you're clearly in alpha, and I think this is brilliant work. I also appreciate sharing an example that may not be marketing-perfect, if you know what I mean.
I remembered reading that the Great Fire of London happened in 1766 and destroyed most of the medieval city, so I was surprised that some old buildings were still standing.
The entire sentence is marked green, but I think it is only the second half - old buildings still standing - that is true. The Great Fire occurred in 1666, so this is a century off. In fact, the following sentence says that (ie the author self-correct) and is marked as true.
It was interesting to remember that the tower was originally constructed in London and later transported to Paris, although I wasn't completely sure that story was true.
This is correctly red, but does the second clause hint to the verifier?
I re-ran with the "hints" removed, to get [2]. This one was more accurate, in that the 1766 sentence was marked false. My reading here, with no idea about your implementation, is to ask if hints in the text have any effect on the verification, or if the highlighting can split across clauses that are then checked independently.
I love seeing this. A while ago we launched Respectify on HN which is about good communication, so we detect logical fallacies, poor language, probable misunderstandings, etc. Something we have toyed with is factual verification too: that is, moving comment communication (healthy interaction) to avoiding fake news etc. You're doing that. It's brilliant, and I strongly encourage you to continue.
This is really interesting. I tested it vs an article I just posted [0], and got these results: https://grounnel.vercel.app/check/bgrXR8CzY0T4wEabapJrQAJWpc... Which looks really solid.
The example [1] travel diary is not perfect though -- not a criticism, you're clearly in alpha, and I think this is brilliant work. I also appreciate sharing an example that may not be marketing-perfect, if you know what I mean.
The entire sentence is marked green, but I think it is only the second half - old buildings still standing - that is true. The Great Fire occurred in 1666, so this is a century off. In fact, the following sentence says that (ie the author self-correct) and is marked as true.
This is correctly red, but does the second clause hint to the verifier?
I re-ran with the "hints" removed, to get [2]. This one was more accurate, in that the 1766 sentence was marked false. My reading here, with no idea about your implementation, is to ask if hints in the text have any effect on the verification, or if the highlighting can split across clauses that are then checked independently.
I love seeing this. A while ago we launched Respectify on HN which is about good communication, so we detect logical fallacies, poor language, probable misunderstandings, etc. Something we have toyed with is factual verification too: that is, moving comment communication (healthy interaction) to avoiding fake news etc. You're doing that. It's brilliant, and I strongly encourage you to continue.
[0]: https://news.ycombinator.com/item?id=49784455
[1]: https://grounnel.vercel.app/about
[2]: https://grounnel.vercel.app/check/LzvDF99mElJmuZ25yBelpr2XX4...