Seeing as how there have been a few instances of spam lately popping up on the feed, might it be worth looking into some form of spam detection using something like a Naive Bayes filter or doc2vec for descriptions and/or target link content? There should be enough good training data for either approach by now I would think.
Also along the lines of #53 it seems like some kind of manual "report" button would go hand-in-hand with an automated flagging approach.
If there's an interest, I'd be more than happy to look into it! I would think nx should make short work of something like this, or even leveraging some existing NLP package like penelope.
Seeing as how there have been a few instances of spam lately popping up on the feed, might it be worth looking into some form of spam detection using something like a Naive Bayes filter or doc2vec for descriptions and/or target link content? There should be enough good training data for either approach by now I would think.
Also along the lines of #53 it seems like some kind of manual "report" button would go hand-in-hand with an automated flagging approach.
If there's an interest, I'd be more than happy to look into it! I would think nx should make short work of something like this, or even leveraging some existing NLP package like penelope.