APPLICANT COMPETENCY ASSESSMENT MODEL BASED ON NATURAL LANGUAGE PROCESSING TO SUPPORT INTERVIEW SUMMARIZATION SYSTEM

• Main Author: Rafif Lutvie, Nizamixavier
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/76892
• Institution: Institut Teknologi Bandung

The interview process is a common thing that is done in job recruitment by a company. Currently, this process is still carried out manually by human workers which causes the need for a lot of resources and the subjectivity of the evaluator. Therefore, there needs to be a system that can assess a person's competency automatically based on that person’s interview results and the competency dictionary owned by a company. Based on research results, the text similarity method with top-N pooling modification is more suitable than the text similarity method without modification because of its ability to only process the most relevant sentences in a text into a similarity score by using the XLM-RoBERTA model to get the best accuracy due to the size of the model as well as bigger data training and the Cohere model to get the best computing time because API services have been provided with very high computing power. Based on the results of testing alternative text similarity methods and models, the best solution is to use the text similarity method without modification with the Cohere model which has the best performance with an accuracy of 79% with an average execution time of 4.5 seconds for each interview transcript. Competency assessment systems can be made using text similarity with the text similarity method without modification giving better performance because it has better accuracy and less computation time. The Cohere model is more suitable for use because of its smaller computation time with the same accuracy.