Cyberbullying Detection using Recursive Neural Network through Offline Repository

The objective of this paper is to predict the user behaviour based on his posts on social networking sites specifically Twitter. The data sets are captured through secured API's exposed by these social network sites and stored in data lakes or NoSQL databases. Tensorflow API's have been used to do predictive analysis of this stored data through recursive networks. This paper is to demonstrate the identification of specific text from the data which is available in various forms - structured and unstructured and coming from various online sources in real time, posted by users worldwide. The online sources referred to in this paper are social networking sites, twitter etc. where multiple users collaborate with each other and post contents. To demonstrate the approach, the Information is captured from these sites through API exposed by these vendors which is captured in NoSQL databases and then NLP is applied to break the text which is then applied against text corpus to identify analogous data. From programming perspective data strictures are used to store the data at run-time. Specific WordNet API is being leveraged for their capabilities to find synonyms.