Hyderabad: Are you annoyed with frequent offensive suggestions while surfing on search engines on the internet? Here is a solution. A three-member team of the International Institute of Information Technology-Hyderabad (IIIT-Hyderabad) has developed a new technology which would automatically detect offensive suggestions on the search engine and either filter or block them completely.
Using the Deep Learning technique of Artificial Intelligence, Harish Yenala, an MS by research student of IIIT Hyderabad, Manoj Chinnakotla, senior applied scientist and professor, IIIT Hyderabad, and Jay Goyal, principal development manager, Microsoft, India have developed the technology which can process data and learning in the same way as a human brain and detect/clear offensive suggestions. For a point in case, while a child searching for a topic on ‘kite flying’ enters the prefix ‘ki’ in a search engine, the suggestion that comes up could be ‘killing people’. Such suggestions would be automatically removed using this technology.
“This system works on the lines of the human brain. Using this technique, offensive suggestions on the search engine can be filtered or blocked. So far, we have trained the system with more than 50,000 sentences, phrases and tested over 10,000 offensive suggestions. Whatever offensive suggestions we gave were filtered by the system,” said Dr. Manoj.
Researchers until now have used conventional solutions such as manually curating the list of patterns involving such offensive words, phrases and slangs, or Classical Machine Learning (ML) techniques which use various hand-crafted features (typically words etc.) for learning the intent classifier.
Recently, the team presented a paper titled ‘Convolutional Bi-Directional LSTM for Detecting Inappropriate Query Suggestions in Web Search’ at the Pacific-Asia Conference on Knowledge Discovery and Data Mining (PAKDD) 2017, South Korea. While 450 paper submissions were received from 50 countries across the world, the paper presented by the IIIT-Hyderabad team bagged the ‘Best Paper Award’.
Although technology has focused on detecting offensive terminology in search engines, the team is sure it will also be effective in other online platforms such as chatbots.
“Using the application programming interface based on the system and parsing the search text in social media platforms, email services, chat rooms, discussion forums, and search engines will lead to a safer and secure web,” the team said. The team is now working on introducing the technology in Bing, a search engine of the Microsoft. The project is expected to be completed in two to three months.