Background
Tagged makes social discovery products that enable anyone to meet and socialize with new people online. They’re building toward a vision where anyone can use a device to instantly connect with interesting new people anytime, anywhere.
Founded in 2004 and profitable since 2008, Tagged is a market leader in social discovery with over 300 million registered members in 220 countries, 11 million unique monthly users who make over 100 million new social connections every month and upload hundreds of thousands of images per day. Tagged is based in San Francisco USA.
The challenge
It is against Tagged’s terms of service for users to post any sexually explicit material on their site. To ensure compliance with the terms of service Tagged were using a manual review process together with an internal hashing system to moderate all image uploads. If an image matched an existing signature hash in their database then it was uploaded, if it didn’t find a match then it was uploaded and a copy queued for human moderation. All images were posted, moderated and then removed if required; meaning that users could be exposed to inappropriate content before it was reported, moderated and removed.
Because of the success of the brand the volumes of uploads were increasing rapidly and the moderation queue was lengthening. Faced with hundreds of thousands of image uploads per day the company needed to source technology to help resolve the growing issue.
The solution
Tagged researched the sector for ‘real time’ software technologies capable of quickly and accurately detecting sexually explicit content within images and videos. They then contacted the market leaders Image Analyzer. After extensive accuracy and performance testing the respective technical teams worked together to integrate the Image Analyzer technology into the moderation process. The objective of the project was to utilize Image Analyzer to reduce the volume of images requiring manual review by 50%.
The integration of Image Analyzer allowed Tagged to implement a more efficient process flow. The new system routes all ‘image’ uploads through the existing internal hashing system and the Image Analyzer software, if the content is deemed acceptable by these systems then it is immediately uploaded, if it is flagged as suspect then it is passed to the manual moderation team for review prior to be uploaded to the site.
The new system is underpinned by a community reporting function which deals with any bad content which slips through.
The results
This change in process facilitated by Image Analyzer has resulted in a far more efficient system with a drastically reduced risk of users interacting with bad content and an improved user experience.
Tagged chose to configure aggressive settings for the Image Analyzer software to ensure a very low false negative rate. These settings still achieve the goal of reducing the load on the manual moderation team by 50%. Tagged acknowledge that if these were reduced, although the risk of bad content slipping through would be marginally increased, it would be possible to decrease the requirement for manual view even further
The Future
As Jill Eisenhart Manager of Partner Services & Customer Advocacy at Tagged observed;
‘We see the technology partnership Image Analyzer as invaluable to our moderation efforts moving forward. They are a great partner being both highly responsive and quick to offer innovative solutions. In terms of content the fastest growing area we are seeing is the proliferation of video, the tools which Image Analyzer has available to manage both static state and streaming are very exciting’
