I’m surprised, though, about it’s relatively low quality… Seriously, with all the time and effort which goes into spamming, can’t someone write a template-selection heuristic that delivers semi-intelligent comments for an article? It’s not that hard (speaking as the author of a program that writes out a customized “snark-o-gram” for pricing issues)…

To get you started, here are some basic principles for spambot 2.0:

1) The very best spam ends with a question, not a statement – ask the author to elaborate on their point with regards to a particular topic; you should be able to determine this using keyword density / prominance; this would be even better if someone could review it manually before sending (eg. computer screen shows spammer a page, writes recommended messages A/B/C, has them select one).

2)  Alternatively – open up up very aggressively with a non-sequitur challenge to the author’s position and ask them to explain how the topic relates to a sensitive subject. You generally need a table of these – associate certain keywords (music, RIAA, tax policy, healthcare) with a message asking how the article is helping address a relevant injustice or oppressive policy. Puts the author on the defensive but in fashion where they will likely feel obligated to respond, without deletion… to clear their name.

3) Write more than one line. In proper english. That isn’t totally generic.

Actually, the very, very best spam looks like… an on-topic comment (with a link)…

And while I’m at it – what’s with the domain names? Surely you can do better…

The post Towards Better Comment Spam – Designing Spambot 2.0 appeared first on Hanging Hyena.