TWO STAGE JOB TITLE IDENTIFICATION SYSTEM FOR ONLINE JOB ADVERTISEMENT

Abstract

The rapid growth of online job portals and recruitment platforms has resulted in an overwhelming volume of job advertisements, making it challenging to accurately identify and classify job titles from unstructured textual data. This project proposes a Two-Stage Job Title Identification System designed to improve the precision and efficiency of extracting job titles from online job advertisements. In the first stage, the system performs text preprocessing and feature extraction, where raw job descriptions are cleaned, tokenized, and transformed into structured representations using techniques such as TF-IDF and word embeddings. This stage also includes filtering irrelevant information and identifying key linguistic patterns associated with job titles. In the second stage, advanced machine learning and deep learning models, such as Support Vector Machines (SVM), Random Forest, and Bidirectional LSTM (BiLSTM), are employed to classify and refine the extracted job titles. The two-stage approach enhances accuracy by combining rule-based filtering with intelligent classification, reducing ambiguity and improving consistency across diverse job postings. Additionally, the system can handle variations in job title formats, abbreviations, and domain-specific terminology. Experimental results demonstrate that the proposed system achieves higher accuracy and better generalization compared to single-stage models. This solution is highly beneficial for recruitment platforms, job recommendation systems, and labor market analytics, enabling efficient job categorization and improved user experience. Overall, the proposed system provides a scalable and intelligent framework for automated job title identification in large-scale online environments.