(copied direct form quara competion.) Some of the function for preprocesing and attention layer was used form the other public kernals and can learn a lot from public kernals of experienced Data scientists.
An existential problem for any major website today is how to handle toxic and divisive content. Quora wants to tackle this problem head-on to keep their platform a place where users can feel safe sharing their knowledge with the world.
Quora is a platform that empowers people to learn from each other. On Quora, people can ask questions and connect with others who contribute unique insights and quality answers. A key challenge is to weed out insincere questions -- those founded upon false premises, or that intend to make a statement rather than look for helpful answers.
In this competition, I will develop models that identify and flag insincere questions, more scalable methods to detect toxic and misleading content.
Functionig of attention layer can be understand from this post https://medium.com/@joealato/attention-in-nlp-734c6fa9d983
Public leaderboard score 0.70235 Private leaderboard score is 0.69729