Op 18 januari presenteert Textkernel Match! bij CLIN (Computational Linguistics in the Netherlands) dat bij ATAK in Enschede zal plaatsvinden.
Remko Bonnema, technisch directeur van Textkernel, demonstreert het matchen van cv’s op vacatures en welke stappen er zijn ondernomen om de match-resultaten te verbeteren. Daarna volgt een discussie over wat er nog nodig is om een optimale match te vinden tussen vacatures en cv’s.
Lees hieronder het gehele Engelse uittreksel.
Matching CVs to job vacancies: semantic search in action
Remko Bonnema, Yves Peirsman, Gerard Goossen, Mihai Rotaru, Lena Bayeva, Chao Li, Florence Berbain, Carsten Lygteskov Hansen – Textkernel
Document understanding systems have proven very successful within many different domains by transforming unstructured documents into a structured and searchable information source. However, a major problem still remains for users: finding the desired information requires both knowledge of the domain and of the query language. A successful solution should guide and inspire the user’ search without enforcing unwanted assumptions. One example of such assumption is to expect users have a full grasp of the data model used for structuring the CVs when searching for suitable candidates in a database of analyzed CVs.
Our solution to the above problem is to automatically transform a job vacancy into a query. This query is then used to search for matching CVs and it is fully compliant with the data model used in the CVs database. Such queries allow for optimal results even for inexperienced users. The whole process employs a wide variety of NLP techniques which are required to bridge the lexical and semantic gaps between the job vacancy and the CV domains.
The talk consists of a demonstration of the system and a presentation of the different features available. We also present an overview of the steps taken to improve the matching results and a discussion of what is still needed to find an optimal match between job vacancies and CVs.