Natural Language Classifcation of Job Acceptance Emails
Primary Investigator (PI) Name
Md Abdullah Al Hafiz Khan
Department
CCSE – Computer Science
Abstract
As individuals attempt to enter the job market, the importance of proper electronic mail (e-mail) management rises. The ability to respond to and sort through hundreds or even thousands of messages becomes a necessary task that can take otherwise valuable time. These conditions call for an all encompassing system that can help automate the process of job email acceptance or denial while also handling other use cases for normal email usage.This study compares a developed DeBERTa language model implementation against a developed Support Vector Machine (SVM) implementation using the Naive Bayes algorithm for text classification in terms of efficiency and accuracy on a dataset of various types of emails. The research involves determining accuracy of classification and the time taken to find the proper results of email classifications. Two separate text classification implementations will be directly compared to each other based on accuracy and number of positive classifications of email types. These implementations include the DeBERTa language model and SVM using Naive Bayes method for classifying emails. Emails are separated into specific types including job acceptance, job rejection, interview emails, pending application emails, and normal non-job application emails. The data will be trained using multiple datasets that include job application emails and non-job application emails. The results of this experiment have not been concluded, but we hypothesize that the SVM implementation will outperform the DeBERTa language model implementation. The DeBERTa language model is meant for very large datasets while also needing various adjustments in order to work properly. Further results will be posted by the time of the symposium.
Disciplines
Artificial Intelligence and Robotics
Natural Language Classifcation of Job Acceptance Emails
As individuals attempt to enter the job market, the importance of proper electronic mail (e-mail) management rises. The ability to respond to and sort through hundreds or even thousands of messages becomes a necessary task that can take otherwise valuable time. These conditions call for an all encompassing system that can help automate the process of job email acceptance or denial while also handling other use cases for normal email usage.This study compares a developed DeBERTa language model implementation against a developed Support Vector Machine (SVM) implementation using the Naive Bayes algorithm for text classification in terms of efficiency and accuracy on a dataset of various types of emails. The research involves determining accuracy of classification and the time taken to find the proper results of email classifications. Two separate text classification implementations will be directly compared to each other based on accuracy and number of positive classifications of email types. These implementations include the DeBERTa language model and SVM using Naive Bayes method for classifying emails. Emails are separated into specific types including job acceptance, job rejection, interview emails, pending application emails, and normal non-job application emails. The data will be trained using multiple datasets that include job application emails and non-job application emails. The results of this experiment have not been concluded, but we hypothesize that the SVM implementation will outperform the DeBERTa language model implementation. The DeBERTa language model is meant for very large datasets while also needing various adjustments in order to work properly. Further results will be posted by the time of the symposium.