up vote 1 down vote favorite
share [g+] share [fb]

From an answer I gave to another question:

There are comment filtering services out there that can analyse comments in a manner similar to mail spam filters (all links to the client API page, organised from simplest API to most complex):

  1. Steve Kemp (again) has an xml-rpc-based comment filter: it's how Debian filters comments, and the code is free software, meaning you can run your own comment filtering server if you like;
  2. There's Akismet, which is from the WordPress universe;
  3. There's Mollom, which has an impressive list of users. It's closed source; it might say "not sure" about comments, intended to suggest offering a captcha to check the user.

For myself, I'm happy with offline by-hand filtering, but I suggested Kemp's service to someone who had an underwhelming experience with Mollom, and I'd like to pass on more reports from anyone who has tried these or other services.

link|improve this question
83% accept rate
I'm in the process of overhauling the blogspam testing facility so any feedback is most welcome and extra-timely. – Steve Kemp May 7 '10 at 15:47
Not a direct answer to the main question, but I've found Web-Of-Trust very good for filtering URLs submitted in comments. WOT returns a reputation score for different factors like trustworthy-ness and child-safety. They have a fairly simple JSON/REST API – sam Jul 22 '11 at 8:28
feedback

1 Answer

up vote 1 down vote

There is an API at www.atlbl.com that you can query for IP address, hostname, domain, full URL, and even email address, to help combat comment spam etc.

link|improve this answer
Good to know. There isn't anything I can find that would help me evaluate the service: How complex is the interface? How good is their algorithm in terms of getting an acceptable false +ve/-ve rate? – Charles Stewart Jun 10 '10 at 8:19
You can get an account for free, API is just simply REST-XML or you can use the DNSBL's. Data is all reputation based and scored, so you can build in your own trigger/threshold points. – user45348 Jun 11 '10 at 10:50
@cwopen: The site says "invitation only"... There are three things specifically I am interested in, in order: (1. most important) low incidence of false negatives, (2) how much work is it to integrate the API into existing weblog software, particularly hand-crafted software, and get it to work the way we want, and (3. least) low false positive rate. – Charles Stewart Jun 13 '10 at 17:30
Where did you see "invitation only" ? It's been out of beta for quite some quite now. You should just need to drop them an email to get an account set-up. False negative's should be low if at all as it's a dynamic lookup, rather than an hourly download or something. The API is very simple, just a HTML call, so just a couple of lines of code in php,perl,whatever. False positive's should be very low as the delisting is automatic and you can setup your own threshold's anyway. – user45348 Jun 15 '10 at 8:38
feedback

Your Answer

 
or
required, but never shown
discard

Not the answer you're looking for? Browse other questions tagged or ask your own question.