ABSTRACT
Current methods for the development, adaptation and maintenance of multilingual job-matching engines (software systems that automatically match candidates with job descriptions) are too expensive (manual labor) and take too much time. Typical systems are built around a semantic core that understands unstructured job-related input (CVs, job descriptions) and uses the semantic information to match jobs and applicants. When transforming such engines for a new domain (e.g., Automotive Industry) or a new language, a great deal of semantics-related (expensive) human work is involved in order to make the system recognize and process the new input.

Additionally, context information could greatly improve the matching process: in the same way that Amazon recommends store items to customers, based on previous purchases and the purchases of similar customers, it should be possible to obtain better job-matching results, based on the knowledge of what individual and similar candidates read, explore, etc. on content-rich web sites. Such contextual information is currently available as large collections on job seeking behavior. However, the potential of this data is currently not tapped. ASIMOV will explore methods for automating the development process of semantic engines combined with the leveraging of behavioral data to increase the matching performance of existing systems.