Article Preview
TopIntroduction
As stated by the United States (US) Department of Labor, the rate of unemployment is 11.1% in the Bureau of Labor Statistics US Department of Labor, Employment Situation of US as of June 2020., Even though a lot of factors exist behind the present unemployment rate, several people are there in the US as well as in other parts of the world, those who look forward to getting new jobs because of the job loss and some other financial crisis. In recent days, almost all companies have enabled posting in job boards directly or pulling job data from the job aggregators, since most of the job postings have been performed online. Nevertheless, it is unsure that every job postings are real, since some of them are fraudulent that may be intended to obtain data or other confidential details from desperate job seekers . It has been posted in USC Career Center. Avoid Fradulent Job Postings. Accessed July 2020.
According to Alghamdi, B., & Alharby, F. (2019), recently in the area of Online Recruitment Frauds (ORF), an employment scam becomes one of the crucial problems. Even though it possesses the benefits (i.e. quick and easy access for the job seekers) and preferred by most organizations, yet performing a job posting on social media has considered being a double-edged sword. The reason behind all of these scams created by the fraudsters might be the cheating of money from the job-seekers by offering employment to them. Hence, it is mandatory to make sure about the job advertisement whether it has been posted on the real site or the fake one.
According to Ibrishimova, M. D., & Li, K. F. (2019) and Zhou, X., Zafarani, R., Shu, K., & Liu, H. It is the biggest challenge and highly impossible to identify these kinds of frauds. For conquering these challenges, the frameworks belong to Artificial Intelligence (AI) and Natural Language Processing (NLP) can predominantly get utilized as they facilitate the researchers to develop the classifier that makes automatic detection of fake news possible . It has considered being a crucial process at this time since numerous job-seekers apply online for jobs because of the present unemployment scenario. Besides, it is mandatory to escaping the scam victims from fake job postings to recover the economy. As described by Khan, J.Y., Khondaker, M., Islam, T., Iqbal, A. and Afroz, S., (2019).Next to NLP, The Machine Learning (ML) method has vitally utilized, since it applies various classification algorithms to identify the fake posts. During that, the fake job posts have segregated from a huge set of job ads with the help of the classification system and warn the user.
In previous studies, According to Ahmed, H., Traore, I., & Saad, S. (2017),Kaliyar, R. K. (2018) various conventional ML and Neural Networks (NNs) approaches have been exploited for detecting fake news . Nevertheless,As described by Singhania, S., Fernandez, N., & Rao, S. (2017) those methods have solely concentrated on social and political news . Therefore, the frameworks were also developed according to their area of interest, since the features of the models have been designed for specific datasets. Consequently, they get trapped into dataset bias and deliver a poor performance over the news of some other topic. Nonetheless, it does not apply many of the enhanced machine learning approaches, such as neural network-based approaches, those which have ascertained to be the optimal solution in several text classification issues. Moreover, the previous studies significantly include the flaw that they have implemented over a specific kind of dataset. Due to this reason, the performance evaluation of several models gets intricate.
Moreover, in these works, the difference between the topics and word embeddings displays slight or refined modification between the fake and real news. It limits the prototype’s capability to comprehend how far extend related or unrelated the reported news appears when compared to the original news. It reduces the accuracy of the system. In addition to depend completely on language, the method relies on remote n-grams, often removed from the suitable context info. Word embedding systems are mostly providing an useful way to represent the meaning of the word. At the same time existing works, a label encoding function is used for converting the text data into the numerical format. New encoding methods are required for labeling text data numerically in an understandable manner. In some kind of circumstances, sentences of diverse lengths could be signified as a tensor with altered dimensions. Traditional models cannot handle the sparse and high order topographies quite well.