So @hichris here's my concept. I want to see what you think about it:
It takes a comment in. It then goes through and separates it into individual thoughts (i.e. I like cats, but I like dogs better -> array(I like cats, I like dogs better)). It then makes everything capitalized so I can ignore the capitalization of each section.
Anyway, I'd remove some common words, like "like", "he", "she", etc. so it'd just be a few keywords
I'd then search a database for different "thoughts" that are whitelisted/blacklisted to see if it has any value.
@hichris123 yeah, that's why I had separate it into multiple parts so if one part is good, even if it does have "edit" in it, the comment is left alone
@hichris123 like you mean destroy context?
@hichris123 For collecting data, I think that if there's an unindexed string, it will be pulled for a review. Not the whole comment, just that snippet of text.
@hichris123 Yeah, like "can you edit that into the post?" might be deleted. We'll just require it to be a near perfect match to consider it "good"
The concept is a great idea IMHO, but it'll need a ton of tweaking to be accurate. This isn't just a quick keyword search thing.
@hichris123 As far as data, we might be able to collect some comments that got deleted and pull the most commonly used "thoughts" from those comments and put them in the database manually
@AnnonomusPenguin Regarding chat.stackexchange.com/transcript/message/16357806#16357806, when you have a comment like that there's usually ones like "edit plz" then "done" then "it still doesn't do anything1!11" and "oh, fixed that, try now" "it works!"
@AnnonomusPenguin Yes, but it would take a lot of work.
@hichris123 Still a little confused. I understand you'll have to delete multiple sometimes, are you just concerned that not all will be deleted, leaving a "fragmented" conversation?
Hmm... I hope that by that point that most of these things will be whitelisted/blacklisted
we might need to get some junk data to feed into it from a SE admin
Line breaks would suffice on comments since I'm pretty sure SE weeds out linebreaks in comments. However, that renders this code useless if we want to use it for questions/answers.
PHPPHP by ircmaxell
<?php // ?>
