As we are dealing with large volume, velocity, and variety of data, the veracity, and the value we derive from it has been of prime importance. False information is easy to spread and can have a significant impact on genuine people or customers. Like other sectors, the job market has also been impacted by it.
One of the approches to tackle this problem is to detect the fake jobs posting early and accurately and thereby flagging them or removing from the system.
As it involves the text data this is a Natural Language Processing Problem. It involves us identifying the job posting as being fraudulent or not and hence it is a classification problem.
We can use historical data to classify the new job postings as fake or not rather than depending upon the human to classify them. We are going to use the latest NLP related practices to classify them and try to evaluate the performance of our models.
We are going to use the state of art NLP technique developed by Google known as Bidirectional Encoder Representations from Transformers (BERT). As of late 2019, 1 out of 10 google searches used BERT for their search queries.
BERT helps better understand the nuances and context of words in searches and better match those queries with more relevant results. In short, BERT can help computers understand the language a bit more as humans do.