Detecting Fraudulent Job Advertisements Using DistilBERT-Based Natural Language Processing

Resumen

The emergence of online job portals has taken over as the leading job recruitment environment though the problem of fraudulent job opportunities is a serious threat to job seekers. False job ads usually take advantage of the job seekers by asking them to provide personal details or procuring them money. It is hard to flag such fraud posts manually as there are a lot of job posts on the Internet. In this paper, a system that employs machine learning to identify fake job postings is suggested through DistilBERT which is a transformer-based model of natural language processing. The proposed method will process job description involving text preprocessing and tokenization and then pass it to a fine-tuned DistilBERT model to do the classification. This model describes the contextual relationship in textual information and predicting whether a job vacancy is authentic or fake. Streamlit is used to provide a web-based application in which users can input job descriptions and obtain predictions in real-time. As proved by the results of the experiment, the given system has an accuracy of 98.88, which proves the efficiency of transformer-based models in fraudulent job post identifications.