Large Language Models for Named Entity Recognition (NER) of Skills in Job Postings in German

This paper addresses the application of Named Entity Recognition (NER) techniques for the extraction of skills from job postings written in German. While traditional methods have been surpassed by deep learning models such as those used in GPT-3, the costs and resource requirements of these models may present significant challenges. To address these issues, the study explores various variants of large language models, which are accessible through APIs and differ in size and cost. These models have not yet been sufficiently explored for applications such as identifying skills in job postings. Specifically, this paper aims to identify the most suitable large language model variant for an operational system and to optimize the query process for skill extraction from job postings.