WebPurify Profanity Filter, Image Moderation, Video Moderation APIs | Hacker News Reader

WebPurify Profanity Filter, Image Moderation, Video Moderation APIs

webpurify.com

13 pointsopusdie12 comments

Demo page:

http://www.webpurify.com/features/

Well, it passes the "Scunthorpe" test for that word at least.

http://en.wikipedia.org/wiki/Scunthorpe_problem

Oddly, it flags "sex" as profanity while finding "rape" not profanity. "sexual" is fine though. I found a false positive -- "cornhole" -- which is a sport most popular in the Midwest:

http://en.wikipedia.org/wiki/Cornhole

It also handles the "space in a swear word issue".

Upper Dicker, Lower Dicker and Cocking all get through as well.

Titty Hill - http://en.wikipedia.org/wiki/Titty_Hill - does unfortunately trigger the filter.

Scrotum is marked as profane too, when I am not sure it should be.

Hey Everyone,

I'm Jon (CTO) of WebPurify. I wanted to answer some of your questions.

1) There will be false positives and negatives with any profanity filter, we just try our best to minimize these. Our filter's algorithm is based on several years of feedback from our customers and you have the options of adding words to your custom allow/reject lists.

2) Most of the alternative image and video moderation services use crowdsourcing, which we have found to be generally inaccurate, not to mention there are security issues.

We have dedicated teams.

3) Simultaneous connections means the number of open connections you can have to our service at any one time. We chose this pricing model because it takes far more processing time to filter 100,000 characters than it does to filter 10 characters.

So basically, the longer the text you filter, the fewer requests you can make in a given period of time.

4) Our Video Moderation API will be publicly available in the next few months, until then contact sales@webpurify.com if you would like to discuss a custom project.

As always our Support Team (support@webpurify.com) is available to answer any questions you may have.

Thanks!

Do you really have people working 24x7 whose sole jobs are to look at images? Is approved or denied within a few seconds, or are they checking in batches? If I send 1000 images at once, how long would it take to hear back about all of them?

Our "standard criteria" team is staffed to operate 24/7 365 days a year.

We do our best to moderate every image within 5 minutes of submission. Today our average time to moderate was 47 seconds.

During times of very high volume it is possible that we may exceed our 5 minute goal, but this is rare.

We can also do custom projects if your criteria is different than our standard, or you require faster response times.

Feel free to get a free trial and give it try.

Is it active for all languages ? Does not seem to work at all with italian

If you are referring to the profanity filter demo on our website, that is only active for English. Here is a list of the languages our full version licenses support:

http://www.webpurify.com/documentation/additional/language/

Does anyone have experience with alternative services that provide this? CrowdFlower used to offer image moderation (http://techcrunch.com/2012/06/21/crowdflower-rtfm-2-0/) but they killed the product 18 months ago.

I realize you can build this with Mechanical Turk, etc, but am specifically interested in ones that offer out-of-box nudity detection.

Cool, I actually am on a project that needs exactly this. However, I find the pricing for profanity confusing. Per simultaneous requests is a bit strange and makes it hard to determine what we need. I think you should consider doing it by total volume, with some sort of automatic rate limiting if you must.

There doesn't seem to be any hint of information about an API for this. There are links to API docs for the other two services offered by Webpurify. But not this one.

sorry, I meant for the video service.

Reply on news.ycombinator.com