Hi,
I am wondering if it is possible with the textcat-manual
recipe to reduce the list of labels to pre-selected/pre-annotated labels only, plus "none-of-the-above". I get an error if I don't point to a file of labels or include the labels in the command line after --label
.
Command Line:
PRODIGY_ALLOWED_SESSIONS=jane,john prodigy textcat.manual dataset_name /path_to_file/filename.jsonl --label /path_to_file/labels.txt
.jsonl file sample:
{"id":"1234","deployment":"xxxx","date":"2021-11-01","text":"random text goes here","label":"label1","orig_accept":["label1"],"score":"0.5","accept":["label1","label2","label3","none-of-the-above"]}
Additionally, the pre-annotated labels are ranked 1-3. Is there any way to include that ranking in the UI?
For instance, the command line content and .jsonl content below: I was able to get the top three labels in the UI, with the top (pre-annotated) label chosen pre-checked. However, I had to hard code the label names for the individual document's results in the command line arguments. I would like to just have the top three labels in my .jsonl file appear in the UI. This differs for each document.
.jsonl file:
{"id":"1234","deployment":"xxxx","date":"2021-11-02","text":"random text","label":"label1","orig_accept":["label1"],"score":"0.5","accept":["label1"]}
command line arguments:
PRODIGY_ALLOWED_SESSIONS=jane,john prodigy textcat.manual dataset_name /path_to_file/filename.jsonl --label label1,label2,label3,none-of-the-above
Thanks!
Cheyanne