Extract potential facts from text?

[Chaddeus]

What might be a good way to analyze a passage of text to determine if there might be things an author should double check before publishing their work?

Entities looks promising, but it's returning almost too much. Some parts are perfect, but some are things like "one" -- which could be important to check, but is likely a false-positive.

I wonder if filters might help, but unsure how to implement it in this fashion.

Thank you.

[sanjayaksaxena]

Hello @Chaddeus can you share one/two representative examples to help us understand the question better... Sanjaya

[Chaddeus]

Interesting. What magic is going on to determine one sentence has importance over others in the passage?

[sanjayaksaxena]

See // Compute relative importance of each sentence as a numerical value between 0 and 1. section there...

[Chaddeus]

Nice. Is sentenceWiseNormalizedWeights based on number of words in the sentence? Is there any sort of analysis of content/context happening to determine "importance?"

[sanjayaksaxena]

It is not on the number of words in the sentence... please go thru the logic.

[Chaddeus]

Ah, I see it now. Interesting. Thank you.

[rachnachakraborty]

Hi @Chaddeus,

Would be wonderful to mark this thread as answered if discussion has been helpful to you.

Your focus on NLP based features makes us curious about your application, would be happy to learn more about it.

Looking forward to hearing from you.

Best,
Rachna

[Chaddeus]

Oh, oops. I thought this was a "discussion" - didn't realize there was a "mark as answer" toggle.