Emo-فہم Emotional State Analysis from Text

Abstract

Words portray different emotions when used in different textual context. In distant communications, a message might not be conveyed across in its real sense, depicting an unwanted impression on the receiving end. Many people have the problem of not being able to express themselves clearly in communications and this problem multiplies in case of distant communications using various electronic media. A person might not be able to maintain a desired tone and emotion during a communication despite wanting to do so, owing to the use of words with dual meanings. Emo-فہم is a research-based self-contained product to be deployed as a website. It is to aid people in the analysis of the emotional tone depicted in sentences, as well as can be used for professional business and psychiatry purposes. The solution encompasses the domain of machine learning with natural language processing techniques. The user inputs the sentence either in form of direct textual input or .txt/.docx file. The solution also facilitates the user with audio input using microphone. The input speech is converted to text using Web speech API. The resultant text is sent to server for processing, which is followed by the prediction of emotion by the model and communication of results back to the client end.