Automated question answering for CVs of candidates
Especially in large companies, the number of incoming applications is very high. Since not all applicants are suitable, it can be very time-consuming to search through the most suitable candidates through the recruiting database. This results in slow response times for candidates that match the companies' criteria.
A machine learning model is needed that is able to automatically answer predefined questions based on resume documents or attachments. These can be, for example, questions about the degree, field of study, or work experience. The major challenge for such a model is to abstract the questions and information from the resume and identify semantic relationships and dependencies to find the appropriate candidates. This requires a training dataset consisting of a large number of applications in which the relevant information has been marked ('labeled').
Potential solution approaches
The automatic evaluation of incoming candidate resumes requires a powerful ML model. For example, a model based on the BERT (Bidirectional Encoder Representations from Transformers) method developed by Google. The main advantage of a BERT model is that it is able to recognize semantically similar words and expressions, taking into account their context. For example, it knows that 'apprenticeship' and 'training' mean basically the same thing in the context of a job application. The pre-trained BERT base model, which has a basic understanding of the language in question (e.g. English), requires adaptation to the specific problem. For this purpose, the model is trained on the training dataset with the labelled applications.