Thanks for the thoughtful feedback. Answers to your questions/comments:
1. "Quality" of each source is a historical rating for each site. This varies as sites write higher or lower scoring articles over time. So not a judgment call we are making.
2. Factual tone is a set of NLP algorithms evaluating the degree and extent of emotional words used as a percentage of the total text. There are also other indicators like how many first person statements are present, unnecessary adverbs etc. Not perfect but directionally accurate and useful as part of the overall rubric.
3. Author expertise does not have subjectivity. Articles are classified into one of several subject areas and we look at how often the author's prior work falls in the same subject area and how well those prior articles rated.
I'll add to our FAQ after your response to see which issues remain unresolved after my comments above. Thanks again.
> 1. "Quality" of each source is a historical rating for each site ("cited" third site"). This varies as sites write higher or lower scoring articles over time.
So an average of the score of all articles you're previously processed, that cite that third site as a reference?
I can certainly think of plenty of sites where this would be non-problematic, but some sites I could see having very high variance. Do you account for this in any way?
> 2. Factual tone is a set of NLP algorithms evaluating the degree and extent of emotional words used as a percentage of the total text. There are also other indicators like how many first person statements are present, unnecessary adverbs etc. Not perfect but directionally accurate and useful as part of the overall rubric.
My intuition tells me a skilled writer may score ok on this, whereas a human reviewer might catch things.
> 3. Author expertise does not have subjectivity. Articles are classified into one of several subject areas and we look at how often the author's prior work falls in the same subject area and how well those prior articles rated.
I'm not saying this is necessarily a big deal, but there are some obvious shortcomings.
Overall, this seems like the type of tool that is sorely needed, so I hope you find success. I'm particularly interested in it for how contentious of an issue it is.
It would be cool (and good for credibility I'd think, as well as potentially providing you some free labor if you allowed feedback) if you had a public accessible page on your website where people could manually submit links and see how your engine ranks each of the 4 attributes. Any chance you might do such a thing?
1. "Quality" of each source is a historical rating for each site. This varies as sites write higher or lower scoring articles over time. So not a judgment call we are making.
2. Factual tone is a set of NLP algorithms evaluating the degree and extent of emotional words used as a percentage of the total text. There are also other indicators like how many first person statements are present, unnecessary adverbs etc. Not perfect but directionally accurate and useful as part of the overall rubric.
3. Author expertise does not have subjectivity. Articles are classified into one of several subject areas and we look at how often the author's prior work falls in the same subject area and how well those prior articles rated.
I'll add to our FAQ after your response to see which issues remain unresolved after my comments above. Thanks again.