rel.manual with pre-labelled spans displays message "No Tasks Available"

, ,
1

Hi all, I am trying to use rel.manual with pre-labelled spans from an external model. I have coerced the model output to the Prodigy JSON format and added it as a dataset:

prodigy db-in data1 data.json

✔ Created dataset 'dhai' in database SQLite
✔ Imported 10498 annotations to 'dhai' (session 2022-04-06_20-59-25) in
database SQLite

Then, I attempt to load the data with the rel.manual command:

prodigy rel.manual data1 blank:en dataset:data1 --label AND,OR,Has_Context,Has_Value,Has_Temporal --span-label Demographics,Medical_Condition,Interevention,Lab_Test,Trial_Terms,Value,Timeframe,Context --wrap

⚠ Skipped 1 span(s) that were already present in the input data because
the tokenization didn't match.
⚠ Skipped 1 span(s) that were already present in the input data because
the tokenization didn't match.
⚠ Skipped 1 span(s) that were already present in the input data because
the tokenization didn't match.
⚠ Skipped 1 span(s) that were already present in the input data because
the tokenization didn't match.
⚠ Skipped 1 span(s) that were already present in the input data because
the tokenization didn't match.
⚠ Skipped 1 span(s) that were already present in the input data because
the tokenization didn't match.
⚠ Skipped 1 span(s) that were already present in the input data because
the tokenization didn't match.
⚠ Skipped 1 span(s) that were already present in the input data because
the tokenization didn't match.
⚠ Skipped 1 span(s) that were already present in the input data because
the tokenization didn't match.

When I navigate to the webpage, I am told that there are no tasks available, however. Understanding why this is happening would be helpful! Thanks

2

Hi! In order to annotate, you don't necessarily have to upload the data to a dataset – you can just point it to the JSON(L) file.

Prodigy will automatically skip all examples that are already present in the current dataset you're saving to, so if you're loading in data1 and saving to data1, you won't be asked about any examples that are already in the set. So this explains why you see "no tasks available".

This could also be relevant if you're annotating data with pre-defined spans: it indicates that the spans don't map to valid token boundaries. So it might be worth looking into that to see if there are any common patterns, e.g. leading/trailing whitespace included in the spans, or an off-by-one error in the character offsets.