Sentiment indicator for e-mails

Abstract

People receive large amount of e-mails from friends, relatives, companies, institutions and known and unknown people. It is not possible the user to read all the received emails during a busy day. Normally, people avoid reading most of the received e-mails by giving priority to the important e-mails. E-mails may contain happy news or sad news or defamatory contents or obscene contents. If there is a way for a user to get an idea of what kind of news would be there in just before he or she opens e-mails, then he or she can choose which e-mails should be opened first and find out which e-mails would make him happy. If the user can find out any obscene contents in an e-mail just before he opens it, then he or she can avoid the embarrassment of opening the email in front of a stranger and then the user will be able to delete it without simply opening it. A proper e-mail categorization is required and this attempt is to categorize e-mails according to the sentiment and the ingredient of the e-mail content, by indicating emoticons on the subject of e-mails. Text Preprocessing, Feature Extraction, Sentiment Classification methods are used to identify sentiments and proper emoticons are used to indicate sentiments on the subjects of e-mails. Machine learning and lexicon based approaches are used to predict the sentiment of emails. A better accuracy level is expected from machine learning in the process of sentiment extraction.