Dear Prodigy Support,
Is it possible to do both NER and text classification annotation? Or I should annotated it twice, first using ner.manual and then using textcar.manual? I have text dataset that will be used for both of NER and Text classification model development.
Thank you
Hi @sigitpurnomo ,
To make the annotation task as efficient as possible we recommend doing your annotation in two passes: NER and textcat separately.
Prodigy train and data-to-spacy commands will take care of merging the annotations for the purpose of training a spaCy pipeline with these two components.
Annotation is easier and less error prone if annotators can focus on task at a time.
That said it is definitely possible to setup Prodigy to collect NER and textcat annotations at the same time via custom recipe. You can combine different UIs using blocks- as long as your input file contains all the necessary info to fill in ner_manual and classification view_ids, Prodigy will be able to render such block.
See here for more info and examples.
Thank you, @magdaaniol, for your suggestion.
I have create a custom recipe like this:
import prodigy
from prodigy.components.preprocess import add_tokens
from prodigy.components.stream import get_stream
import spacy
@prodigy.recipe("peer_review")
def peer_review_ner_cat(dataset, lang, file_path):
# We can use the blocks to override certain config and content, and set
# "text": None for the choice interface so it doesn't also render the text
blocks = [
{"view_id": "ner_manual"},
{"view_id": "choice", "text": None}
]
options = [
{"id": 3, "text": "General Comments"},
{"id": 2, "text": "Detect Localization"},
{"id": 1, "text": "Detect Suggestion"},
{"id": 0, "text": "Detect Problem"}
]
nlp = spacy.blank(lang) # blank spaCy pipeline for tokenization
stream = get_stream(file_path, loader="jsonl") # set up the stream
stream.apply(add_tokens, nlp=nlp, stream=stream) # tokenize the stream for ner_manual
return {
"dataset": dataset, # the dataset to save annotations to
"view_id": "blocks", # set the view_id to "blocks"
"stream": stream, # the stream of incoming examples
"config": {
"choice_style": "multiple",
"labels": ["Problem", "Suggestion", "Localization","General"], # the labels for the manual NER interface
"blocks": blocks # add the blocks to the config
}
}
But when I run the prodigy with this recipe using this command:
prodigy peer_review ner_cat_peer_review id dataset/peer-review-masdig.jsonl -F prodigy/recipe.py
there is an error in the web interface like shown in this image:
Can you help me to find what caused the error?
Thank you
I have solved this problem based on this information https://support.prodi.gy/t/combine-ner-and-doc-classification-in-annotation-process/3314/3
Thank you
Great! Thanks for sharing the post with the solution!
