Turkish Job Posting Generation with Retrieval-Augmented Generation Model
Zeliha Erim, Burcu Yılmaz · 2025
With the rapid development of technology, the communication gap between human resources and field experts who have difficulty in following the developments, especially in the software field, has caused incomplete or incorrect job postings to be written. The motivation for this study was the faster and more accurate publication of software job postings, the creation of attractive job posts and the lack of literature studies in the Turkish language in this field. It was revealed that the texts obtained as a result of Hybrid Retrieval augmented generation (RAG), question multiplication and reranking produced useful results. It was observed that with the applied RAG system, texts that were %0.0257 more realistic and meaningful could be produced than the large language model used. It was aimed to contribute to the literature by using advanced natural language processing techniques and large language models with Turkish software job postings obtained from external online sources.