Automatic content moderation using validates_text_content | Hacker News Reader

Automatic content moderation using validates_text_content

blog.thingsaaronmade.com

7 pointsaarongough17 comments

I could see this causing problems on technical discussion sites as it might consider a block of C code a bad comment. Sometimes there are legitimate reasons to post something non-english, like a blob of numbers or some ascii (or unicode) art.

Slashdot has some filters like this on their comment system, but I think they are mostly targeted at blocking ascii art spam.

I can see that being a problem too. I think in that case you would have to change the validation code to look for those exceptional cases. In all honesty the validation as it is now is designed to handle content on more conversational sites (versus technical ones).

I'll be interested to see what people use the validation for and there shouldn't be any problem expanding the code to handle more edge-cases...

You could modify the plugin to exclude certain areas of the comment from validation. I'm thinking of Markdown, where indented lines are code -- you could just validate non-indented lines.

I thought about that after the top comment. The only issue with that approach is that it would be easier to game the system if it were setup this way. Simply include your content in an indented block to bypass the filters.

True, but the goal is to encourage quality comments. I think there is more benefit to allowing code markup in comments than there is potential downsides with people gaming the system. The system can be gamed anyway be a mildly persistent troll, so don't punish your good users by making commenting too hard!

The system achieves the goal of "good comments" by putting up a friction barrier to terrible comments like the ones on YouTube, not impassible barrier to someone who intends on posting a bad comment.

I guess going with a statistical tool will be easier in the medium to long term than adding special cases.

That's definitely a possibility. It's worth noting that the plugin was created with a specific use-case in mind (a non-technical user generated content site) and that outside of that role it's clearly not going to perform as well...

Yes, for such a limited use case special rules are quite good.

Just don't fall into a trap.

So you create a barrier by rejecting "bad" comments versus creating a barrier by using signup/captchas etc.

I'm not really buying it. The problem in communities isn't really bad writing or quick comments, it's usually trolls who will circumvent something like this quite easily.

I see what you mean about there still being a barrier, however I would describe it as a trailing barrier, rather than a leading barrier. Most people will never encounter it and if they do then they have already invested at least a little of their time to creating their content.

It's been my experience that once they are invested in the process they are more likely to commit to following through and completing the process, even if that means getting rejected several times (I get to see the progression because the systems in production email me every time content is rejected). Those that don't follow through are the one who were just creating throw-away content anyway, which is what I want to discourage.

From what I've seen the trolls do make it through a percentage of the time but they tend to be well-written, entertaining, trolls. Those who spout short comments full of racial slurs don't tend to understand why their comments are getting rejected and they either give up or their comment morphs into something marginally more useful.

In that case the voting and flagging mechanisms built-in to the posting system come into play, the majority of users who are reasonable tend to be able to swarm out the trolls, just like they do here on HN.

The pivotal piece there being that the rejection mechanism tends to discourage people from posting that have no investment in what they're writing, which means the remaining users have a vested interest in keeping the trolls at bay...

If you are interested in seeing examples of content progression from a live system just shoot me an email!

Do you explain the rules to the users? What kind of feedback do you get when your comment violates the rules?

The rules are not explained in advance to the users, that would make the system too easy to game, particularly if you were very specific.

If the content fails the validation this is the default message that is returned:

Correct capitalization and punctuation are required. (ie: All sentences should begin with a capital letter and end with a punctuation mark, etc.)

Most people seem to understand and simply clean up their sentence structure a bit. People that have written their entire comment in block caps tend to simply give up.

Before someone else says it: I know I should migrate to using Git, I just haven't found the motivation yet! I will try to migrate in the next few months though.

Git is fine. The other distributed version control systems (darcs, mercurial, bazaar, etc) are also usable.

Yeah it's just that the Ruby/Rails world is pretty tightly tied into Git. Most Gems these days are automatically created by GitHub/Gemcutter...

I'm actually perfectly happy with SVN, but in order to participate more in the open-source side of the Rails world I'll need to move to Git.

Interesting. Might try to port it to PHP.

Thanks! I don't expect it would be too hard, most of the logic can be taken pretty much as-is. Let me know how it goes aaron@aarongough.com

Reply on news.ycombinator.com