Hello Eylem,
regarding your question how to choose metadata: As the researcher (and expert for your experiment) you should consider everything that is relevant to comprehend and reproduce your scientific work, which is not reflected in the recorded data. This highly depends on your experiment, but might include:
- Time and date: When was the experiment conducted?
- Environmental conditions like temperature and humidity
- Who conducted the experiment? (But mind personal data here)
- Which measurement equipment has been used? Which tolerance does it have?
- Which software, hardware, or even operating system was used? Which software version?
- …
Specific repositories might require additional metadata: Check first, because collecting metadata once your experiment is finished can be hard.
Depending on your type of research and especially research method, one of NFDI4Ing’s archetypical researchers (the “archetypes”, https://nfdi4ing.de/archetypes/) might give you a more specific answer, e.g. metadata schema from Doris for High Performance Computing (HPC). A colleague of mine developed the SensOr Interacing Language (SOIL) to make sensor data FAIR by modelling with additional metadata (rf. https://doi.org/10.5281/zenodo.7757249).
Regarding metadata schemes: There are various out there, generic as well as specific. Generic typically have higher compatibility, while specific metadata schemes are more accurate in their description. So: It depends on your use case, your domain etc. which one to pick.
The AIMS project (Applying Interoperable Metadata Standards, https://www.aims-projekt.de/) developed the Metadata Profile Service [https://profiles.nfdi4ing.de] where you can build individual metadata profiles by specifying existing profiles. This modular and hierarchical approach makes metadata profiles specific while being interoperable.
