Design and implementation of an NLP application for removing the signature part of an email for a selected company

This diploma thesis deals with the design and implementation of an application that can recognize an email signature from its contents. This application will later be used in the pre-processing of input data, i.e., incoming emails, within a model that classifies the content of the email itself, in order to simplify its classification and forwarding to a specific department. The structure and language used in emails are often diametrically different, as the email can be written formally or informally, it can be a personal, corporate, advertising, official, or another type of email. For this reason, the task of classifying the text of an email needs to be seen as an NLP task, in other words, the task of natural language processing. Within this diploma, the current situation of email sorting in the selected company, where this application will be included, is first described and explained. There is also research for academic theses that deal with a similar topic of email classification and finding their signature sections. Then, the various methods used in NLP applications are discussed, both in the preparation of input data, the selection of a suitable algorithm and libraries, and for the appropriate measures of validation of already trained models. The practical part of the diploma thesis describes the implementation of the application itself with examples of programming code. Finally, this thesis provides recommendations on how to best use the trained model and what would be appropriate to change or modify for the next training of the model.