Hello,
I am new at prodigy and i would like to ask some information. Is there a way to create a custom recipe with textcat in order to annotate Named Entities?? Any Suggestions??
Thanks a lot
Hi! Could you explain a bit more about what you’re trying to do? What do you want to extract, and do you want to train a text classification model or a named entity recognizer (or both)?
Hi ! I would like to train a named entity recognizer . I would like to do this by creating my own custom recipe.
I would like to do something like this in the image but with my own data
Thanks a lot
Thanks!
The built-in ner.manual recipe should do exactly what you need. You can load in your own data and a set of labels, and you’ll be able to annotate your data. The spaCy model is only used for tokenization, i.e. to split the words so you can annotate faster and don’t have to highlight exact words.
prodigy ner.manual your_dataset en_core_web_sm your_data.jsonl --label "PERSON, ORG, GPE, LOC"
The dataset your_dataset will then contain your annotations in JSON format – one example for each text, with a "spans" property containing the entities. You can then use the data to train a model – either using Prodigy and/or spaCy, or any other library and solution of your choice.
Just be aware that if you’re only labelling data manually and want to train a new model you’ll need a lot of annotations. If you haven’t seen it already, you might also want to check out this video on training a new entity type with Prodigy:
It also shows some other tips and tricks that you can use to avoid labelling every single entity type by hand. For example, you can create patterns that help you select examples that you can say yes or no to. In some cases, manual labelling is the best option – but we’d always recommend to automate as much as possible. You might have to experiment with different approaches until you’ve found the one that works best for your model – and the faster you can do this, the better 
Thanks a lot your information was very helpful