Practical aspects for the creation of an audio dataset from field recordings with optimized labeling budget with AI-assisted strategy