Not compatible with spacy 3.x

[ayeshah]

When running the sample code I get the following error:

---------------------------------------------------------------------------
ModuleNotFoundError                       Traceback (most recent call last)
<ipython-input-82-777c977a1d80> in <module>
      1 import doccano_transformer
      2 
----> 3 from doccano_transformer.datasets import NERDataset
      4 from doccano_transformer.utils import read_jsonl
      5 

~/miniconda3/envs/base/lib/python3.7/site-packages/doccano_transformer/datasets.py in <module>
      3 from typing import Any, Callable, Iterable, Iterator, List, Optional, TextIO
      4 
----> 5 from doccano_transformer.examples import Example, NERExample
      6 
      7 

~/miniconda3/envs/base/lib/python3.7/site-packages/doccano_transformer/examples.py in <module>
      2 from typing import Callable, Iterator, List, Optional
      3 
----> 4 from spacy.gold import biluo_tags_from_offsets
      5 
      6 from doccano_transformer import utils

ModuleNotFoundError: No module named 'spacy.gold'

Seems like this is removed from spacy v3.x: https://github.com/explosion/spaCy/releases

[agademic]

Just change the line

from space.gold import biluo_tags_from_offsets

to

from spacy.training import offsets_to_biluo_tags

and the corresponding function call in the doccano_transformer.examples file and it should be fine.

[EDIT] It works only if you want to use the dataset.to_conll2003 method.

For dataset.to_spacyit still throws an error, since the token object (created in utils) doesn't seem to be spacy compatible.

[sonvx]

• If you don't have other constraints, downgrade spacy should work:

pip install spacy==2.3.2

Or

conda install -c conda-forge spacy==2.3.2

[mirfan899]

Having the same issue.

[Matt-Payne]

How do you let a bug like this just sit? Unusable after spacy 3

[sonvx]

@mirfan899 @Matt-Payne you should clearly explain if my approach doesn't work: if you followed, what errors did you get? I clearly mentioned that there might be other constraints :).

[mirfan899]

What does the issue say? He is not asking for constraints you came up with. It clearly mentions that doccano-ransformer not compatible with spacy 3.

[abtExp]

I've written a script to convert the doccano output jsonl to bilou format json that can be directly converted and used for spacy training. Check it out here : https://github.com/abtExp/doccano_to_bilou

My script doesn't rely on spacy, thus no compatibility issues.