“CurenZee” A Mobile Based System for Real Time Sri Lankan Currency Recognition using Object Detection: An Accessible Solution for Visually Impaired

Abstract

In the current era, there are countless methods of digital communication. However, one of the oldest but still in use mode of digital communication is Email. Emails have been used as methods of communication between businessmen, students, and industrial professionals for a long time. One of the major drawbacks of this communication method is Email overload. Due to large numbers of received emails, students and industry professionals often find themselves unable keep track of important emails and sometimes ends up missing vital information.

A system that takes in user preferences in the form of keywords and uses them to assign priority values to important emails and summarize them for the user to view is a novel solution to this problem. The email dataset that was used for this system is the Enron Email dataset and the it was extensively preprocessed for this purpose. The summarization component functions with the use of BERT model and the keyword extraction and similarity comparison is done through Jaccard Similarity.

The findings of the tests carried out on the system showed significant test results that validates the existence of the research project. Rouge 1 and Rouge L scores were used for the purpose of evaluating the prototype of the system and the scores generated are well withing the expected range of performance.