I want to annotate some passages in the SQuAD format. After some research, I found that the Haystack Annotation Tool by deepset.ai is the best option for this purpose.

Most of the old tutorials suggest opening an account on their annotation website direcly. However, they no longer host it on their website; instead, they provide the code on GitHub and instruct users to use it locally. In this page and in a video tutorial, they outline three steps for users: download Docker Desktop, clone the git repo from this link, and use the command docker-compose up.

The problem arose when I installed and opened Docker Desktop on a VM with 32GB of RAM; it displayed an error: "Docker Desktop - Unexpected WSL error." After some investigation, I discovered that I need to make changes in the BIOS to fix it. However, my network engineer informed me that they cannot make any changes in the BIOS.

Subsequently, I installed Docker on my PC with only 8GB of RAM. Unfortunately, it consumes more than half of my RAM and CPU, making it challenging to use properly. Furthermore, in the documentation for the Haystack annotation tool, they suggest changing the database information. Unfortunately, I have no idea which database they are referring to.


DB_HOSTNAME: "db" DB_NAME: "databasename" DB_USERNAME: "somesafeuser" DB_PASSWORD: "somesafepassword"

POSTGRES_USER: "somesafeuser" POSTGRES_PASSWORD: "somesafepassword" POSTGRES_DB: "databasename"

COOKIE_KEYS: "somesafecookiekeys" JWT_SECRET: "somesafesecret"

I don't think generating Question-Answer datasets in Machine Learning, like SQuAD, should be that difficult to use. Why is there no online/offline, easy-to-use, user-friendly UI for it? Is the HayStack Annotation tool by deepset.ai the only tool that can create SQuAD data easily? Why isn't anyone else making it easier to use?


1 Answer 1


You can use Prodigy, an online annotation tool by the same people that created and maintains Spacy. Here you have a QA recipe and here a related support question.

If you prefer open source solutions, you can try with Doccano or Brat, although I'm not sure if they already support QA annotation or not (see this).


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.