Datasets:
Could `common_names_to_tax.json` be transformed into a valid mapping?
Contrary to what the filename might suggest, common_names_to_tax.json
doesn't define a one-to-one relationship at the moment. It's more like a many-to-many mapping, which is a bit weird to work with. I'm seeing one-to-many relationships, e.g. the common name "deer" being mapped to 3 different taxonomies. I'm also seeing many-to-one relationships, e.g. "squirrel" and "tree squirrels" are mapped to the same taxonomy. Could this situation be improved?
I believe this issue was not present in the original source (https://lila.science/taxonomy-mapping-for-camera-trap-data-sets/), since there all the mappings were contextual, i.e. a mapping was unique in the context of a dataset. The dataset name was unfortunately dropped when the mapping was imported to HF.
Thanks,
Ștefan
Thanks for raising this! I can see how this would be weird to work with, and I'm more than happy to implement a different solution. I'll reach out to the maintainer of the original LILA (https://lila.science) to see if they have recommendations on how to consolidate the taxonomy + common_name mapping to avoid this awkward many-to-many situation, and I'll loop back here as soon as I have an idea for how to do it. If you have any preferences for what the final mapping (and an interface to it) could look like, let me know!