4 ms·

WebPurify Profanity Filter, Image Moderation, Video Moderation APIs

13 points by opusdie 12y ago

chadlundgren 12y ago
Demo page: http://www.webpurify.com/features/ http://www.webpurify.com/features/ Well, it passes the "Scunthorpe" test for that word at least. http://en.wikipedia.org/wiki/Scunthorpe_problem 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 http://en.wikipedia.org/wiki/Cornhole It also handles the "space in a swear word issue".
- Nursie 12y ago
  Upper Dicker, Lower Dicker and Cocking all get through as well. Titty Hill - http://en.wikipedia.org/wiki/Titty_Hill http://en.wikipedia.org/wiki/Titty_Hill - does unfortunately trigger the filter.
  - bhayden 12y ago
    Scrotum is marked as profane too, when I am not sure it should be.
dberlind 12y ago
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.
- dberlind 12y ago
  sorry, I meant for the video service.
MichaelGG 12y ago
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.
esilverberg2 12y ago
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/ 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.
webpurify 12y ago
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!
- bhayden 12y ago
  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?
  - webpurify 12y ago
    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.
- bonif 12y ago
  Is it active for all languages ? Does not seem to work at all with italian

13 points by opusdie 12y ago

Demo page: http://www.webpurify.com/features/ http://www.webpurify.com/features/ Well, it passes the "Scunthorpe" test for that word at least. http://en.wikipedia.org/wiki/Scunthorpe_problem 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 http://en.wikipedia.org/wiki/Cornhole It also handles the "space in a swear word issue".

Nursie 12y ago
Upper Dicker, Lower Dicker and Cocking all get through as well. Titty Hill - http://en.wikipedia.org/wiki/Titty_Hill http://en.wikipedia.org/wiki/Titty_Hill - does unfortunately trigger the filter.
- bhayden 12y ago
  Scrotum is marked as profane too, when I am not sure it should be.

dberlind 12y ago

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.

dberlind 12y ago
sorry, I meant for the video service.

MichaelGG 12y ago

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.

esilverberg2 12y ago

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/ 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.

webpurify 12y ago

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!

bhayden 12y ago
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?
- webpurify 12y ago
  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.
bonif 12y ago
Is it active for all languages ? Does not seem to work at all with italian