Soutenance de thèse de Albeiro De Jesus ESPINAL PULGARIN

In the context of pre-selecting candidates during recruitment, this thesis explores the modeling of job advertisements and resumes with a view to a more robust automatic matching between these documents. Despite the significant advancements brought about by artificial intelligence for automated matching, several challenges remain. These include managing unstructured data from resumes, accounting for uncertainty during the extraction of information from job advertisements, ensuring the transparency and interpretability of the methods used, and considering the opinion and organizational context of recruiters.
To address these challenges, we introduce a methodology oriented towards integrating recruiter strategies, representing uncertainty, and interpreting automatic language processing. The proposed approach combines the extraction of salient information and the semantic annotation of job advertisements through a possibilistic model based on ontologies. It also incorporates major language models, such as BERT, as well as grapholinguistic representation of text for an in-depth analysis of resumes. Implemented within the recruitment department of DSI Group, a French consulting company in Information and Communication Technologies, this methodology has demonstrated an increased ability to identify essential information from job advertisements and efficiently segment contemporary resumes. This contributes to the construction of more robust and context-sensitive automatic matching processes.
Our work highlights the contemporary challenges and opportunities of matching these documents and suggests innovative solutions integrating traditional and recent methods. These insights may be of interest to researchers, recruitmenprofessionals, and decision-makers.