Data Set Management in Annif

[mo-fu]

When extending Annif with more hyperparamter optimization functionality or training via API it may be useful to have data set management.

Possible functionalities:

• adding a data set annif data add $DATA_SET_NAME $PATH
• splitting a data set into folds annif data split $DATE_SET_NAME 0.7:train 0.2:test 0.1:validate, could be adressed using annif train ${DATA_SET_NAME}#train
• removing data sets annif data remove $DATA_SET_NAME

[osma]

Can you provide an example how this could look from the user perspective? For example CLI commands or REST API calls?

[mo-fu]

I added some examples on CLI usage.

[osma]

Ah, now I understand what you mean by this, thanks!

How would this be implemented? Where would the managed data sets be stored? Somewhere under the data directory? Would these be copies of the originals or something else?

This would expand the scope of Annif quite a lot. I'm not sure it would be worth the additional complexity. But it's an interesting idea.