Where do we store the text files that we load in for annotation.

First and foremost, apologies for the basic question I promise I’ve googled high and low.

Currently on windows invoking prodigy from via a Jupiter Notebook.

After creating a data set my_set and loading in the vocabulary

I run !pythom -m prodigy ner.teach my_set en_core_web_lg news_headlines.jsonl

but I’m quite sure my latter path is not correct because it just continues to run and I never see the annotation tool.

Interestingly, when doing prodigy stats -ls it shows 6 sessions which are my six attempts.

Any help or assistance would be much appreciated thank you so much and again sorry for the low end question.

Hi! If you pass in the path to a file and it’s incorrect (doesn’t exist, bad file etc.), Prodigy should raise an error and tell you. Did you see any output or error message from that cell when you ran the command?

Another thing to consider: In order to serve you the web app, Prodigy starts a web server when you run one of its prodigy commands. Notebooks are super cool for running Python, but they’re not always the best environment for a command-line tool like this. You’re essentially starting a web server from within an app served via a web server, which can easily add some unwanted complexity. So at least for debugging, you might want to try running the commands directly in your terminal.

(Btw, this is not official yet but I can reveal it here: We’re currently working on a Prodigy extension for JupyterLab, which will open the annotation interface in a widget and just make the whole integration much smoother. Maybe even a GUI for putting together the commands. We’ve teamed up with a Jupyter core dev for this, so I’m suuuper exctied :blush:)

1 Like

Thank you very much for taking the time to respond. Doing it outside of the Notebook did the trick and I am on my way with regards to annotating. With respect to future integration of Prodigy and the Notebook, sounds very exciting, can’t wait! Thanks for all the work you and the Spacy team do!

1 Like