Smart Document Classifier and Analyzer Using NLP

Circulars are the primary mode of communication in an established hierarchy. The Vishweshvaraya Technological University releases around 35–40 circulars every month which averages to around 450 annually. RVCE administration themselves generate up to 15–20 circulars on a monthly basis. It is tedious for an individual to keep track of the large flow of crucial information. A machine learning model thus becomes necessary to meticulously scan the documents, classify the information and present the essence in a brief manner. This paper presents a machine learning model developed in Python that is capable of classifying announcements given by the college into four categories: fees, exams, holidays, and events. The classification of the documents are carried out using the SVM algorithm. Using computer vision algorithms, crucial data such as the circular id, date of issue, authority signatures, and tabular data are retrieved. All of this information is placed into a template to produce a brief, easy-to-read notification. To automate the transmission of these templates to their appropriate stakeholders, a mail bot is created. The model achieved an accuracy of 98.2%. The results clearly reflect that the model can accurately classify announcements into the four categories with a high degree of reliability.