Content moderation on social media using Machine Learning

Overview

Forum discussions are a popular method for obtaining consumer feedback for any business. Our client had an eCommerce platform, selling a wide range of FMCG consumer items. Their platform enabled their customers to conduct discussions about the products listed on their platform. It gives the company the customer perspective of their products and their services. Such feedback provides an opportunity and input for the business to improve and grow its base. Our client wanted us to build an automatic content moderation system to maintain the language etiquette on their e-commerce platform.

Challenge

There was a need to moderate these huge discussion forums to maintain the tone and nature of the language used. The team performing manual content moderation was overwhelmed and stretched with the volume and repetitive nature of the discussion. We needed to come up with a solution that will handle large amount of feedback data and classify the tone of the discussion. To achieve this, we were required to study and detect the nature of the language used via text recognition and redact unhealthy language for achieving moderation.

Solution

The problem involved identifying foul language from the discussion boards and redacting them automatically. We developed a Natural Language Processing model to identify text sections that constitute unhealthy language. We then redacted the content from their discussion board using simple text management tools.

Impact

The automation of the repetitive task of content moderation proved to be beneficial for their team. The staff were able to concentrate on the relevant feedback and improve their contribution to the company. We observed an increase in their quality of work. The business was cost effective and work productive in addition to the increase in sales due to valuable feedback

Technology stack