MAGIC BRICKS
Collaboration to develop an advanced AI-based automated moderation software to effectively segment and appraise data
Client’s Profile:
Magicbricks.com is India’s number 1 online property site. It is a website that provides a common platform for property buyers & sellers to locate properties of interest in India, and source information about all property related issues. Their platform helps buyers & sellers to locate properties of interest & source information on real estate that can be of use to all.
Description of Project:
The requirement was to design an automatic moderation solution for the client. We majorly employed artificial intelligence to help them keep and even exceed their high-quality standards. We efficiently deployed machine learning as well to so as to gradually increase their efficacy in evaluating and moderating the data uploaded on their site.
Challenge:
- Whenever data is uploaded upon any website or platform, it undergoes moderation.
- Data is fully checked, for example, whether an image concerned has been uploaded at the right spot or not or whether the information uploaded contains anything confidential.
- After qualifying about 45-70 such checklists, the data is finally uploaded and then only the post is open for viewer engagement.
- As the moderation process was done manually and moderation teams are expensive, it took a lot of time and money of the company. As their business scaled up and consumer-base grew, more moderators needed to be hired.
- They recognized that it was impossible to maintain the cost of employing a huge staff of moderators and that their existing process would have to change.
- That’s when they approached Infotonics Media to fully automate the moderation process, so that their turn-around time would reduce to a good extent and their costs would be saved.
Solution:
- We designed an artificial Intelligence based moderation solution for them wherein it took only 3 to 10 seconds to check the whole image and finalize it.
- In very rare cases it would take 10 seconds, otherwise whole process got completed within 1 to 3 seconds only.
- It would not only check the image, but also generate comprehensive report on it. If the image needed to be rejected, it rejects it then and there, otherwise would upload it.
- We also designed a workflow for them in which the images could be classified into sub-sections as kitchen, bathrooms and bedrooms. The module ensured that in case we happened to upload a kitchen image into the bedroom section, it would automatically reject it, suggest us to upload it into the kitchen section and then would put it under the kitchen section.
- Along with this, we also made an API which can hold about 10,000 records moderating per hour.