search

abtExp / doccano_to_bilou

Convert Annotation Output (JSONL) From Doccano To Spacy Training Ready BILOU Format.
MIT License
5 stars 2 forks source link

Doccano to bilou for RE #3

Closed Aj-232425 closed 2 years ago

Aj-232425 commented 2 years ago

Hi Anubhav, I have gone throught the script. It converts doccano jsonl output of NER to bilou format. I am working for relation extraction. How can I convert jsonl of relation extraction to bilou format. Does jsonl itself be supported by Bert /BioBERT. If not so, could you please guide. Any help will be appreciated.

abtExp commented 2 years ago

Hi Ajit, I wrote that script for NER usecase and it doesn't support converting relation info jsonl to bilou.

As i don't readily work on NLP problems, I'm not sure about the training data preparation steps for bioBERT and it's architecture, so unfortunately i won't be of much help to you.

One thing you can do is research on the training pipeline of your usecase and write a custom script to convert the jsonl to the required format.

I hope you find the solution. All the best for your project.

Cheers