--- license: mit --- # isy-thl/bge-reranker-base-course-skill-tuned ## Overview This model is a finetuning of BAAI/bge-reranker-base on a German dataset containing positive and negative skill labels and learning outcomes of courses as the query. The model is trained to perform well on calculating relevance scores for learning outcome and esco skill pairs in German language. ## Using FlagEmbedding ``` pip install -U FlagEmbedding ``` Get relevance scores (higher scores indicate more relevance): ```python from FlagEmbedding import FlagReranker reranker = FlagReranker('isy-thl/bge-reranker-base-course-skill-tuned', use_fp16=True) # Setting use_fp16 to True speeds up computation with a slight performance degradation scores = reranker.compute_score([['Einführung in die Arbeitsweise von WordPress', 'WordPress'], ['Einführung in die Arbeitsweise von WordPress', 'Software für Content-Management-Systeme nutzen'], ['Einführung in die Arbeitsweise von WordPress', 'Website-Sichtbarkeit erhöhen']]) print(scores) ``` The resulting scores can be normalized using a sigmoid function ```python score = 1 / (1 + math.exp(-score)) ``` ## Performance ![Eval Results comparing intfloat/multilingual-e5-base, isy-thl/multilingual-e5-base-course-skill-tuned and also a version reranked with isy-thl/bge-reranker-base-course-skill-tuned](https://cdn-uploads.huggingface.co/production/uploads/64481ef1e6161a1f32e60d96/x5xqyU-_raRyVOGqGVpq-.png) ## Acknowledgments Special thanks to the contributors from the **Institut für Interaktive Systeme**, **Kursportal Schleswig-Holstein**, **Weiterbildung Hessen eV**, **MyEduLife**, and **Trainspot** for their invaluable support and contributions to the dataset and finetuning process. **Funding:** This project was funded by the **Federal Ministry of Education and Research**.