Classification of official letters using TF-IDF method

Abstract

Abstract This study developed an application for official letter management in Sukasada 1 Public High School. The development of the application aimed to contribute to government programs in improving public services, especially in correspondence. In the letter classification process, the researcher implemented the TF-IDF method whose working procedures included conversion, Optical Character Recognition (OCR), filtering, tokenizing, and classification. Conversion is to convert official mail to digital form (jpg). OCR is to get the text of the letter then carries out the official letter in the form of an image and then filtering and tokenizing are carried out. Classification is the process of grouping a letter into its category by calculating the cosine similarity value between the letter being tested and the letter in the system. To test the accuracy of the classification results, a confusion matrix is used. The results of the study are in the form of web-based and mobile-based applications. Operators and admins used the web application to manage official letters at SMAN 1 Sukasada. The mobile application served to facilitate the principal in accessing mail data from a smartphone. The system was able to group letters with 78% accuracy (good), precision 77% and recall 77%.