[Issue# 1189] Hugging Face - Question Answering Notebook

[muhtalhakhan]

// Adding Question Answering notebook using Hugging Face Transformers.

Question Answering models can retrieve the answer to a question from a given text, which is useful for searching for an answer in a document. Some question answering models can generate answers without context!

Loading the pipeline
Import Pipeline from Transformer after installing the transformers and tensorflow.

Input
The input would be the text from any source.

Output
For the output a new window is opened at the link provided above by Gradio, where you can enter your context and get your answer delivered.

[jravenel]

@muhtalhakhan can you create Issue and attach it to PR (see message from @Dr0p42) + add it to the Community roadmap: https://github.com/orgs/jupyter-naas/projects/4?fullscreen=true

This is how it's supposed to be done so we can track your progress.
Thanks

[muhtalhakhan]

Hey @Dr0p42 , and @jravenel I have created an issue which is a 1189 and added it to the review column in the community roadmap.

[FlorentLvr]

Hi @muhtalhakhan for your contribution, I just reviewed your notebook and changed it to pass the control:

• I assigned an issue to your PR => you can't create a solution (PR) if there is no problem (Issue). The issue was created, you just had to link it to your PR.
• IMO framework => the section "## Model" was not in your notebook, I had it and rework a bit your code to make it easier to understand. Feel free to make so change if you don't find it relevant. The goal is to have the same framework for all notebooks to make it easy for anybody to use your templates catalog.

After passing the control, I tried to run your notebook but it throws me several errors.
Could you have a look and fix it? Let me know if I can help :)

[muhtalhakhan]

Hey @fravenel, I am glad that you took the time and fixed the errors popping out. I apologise for the late response, was busy with my Final year project phase 1, and its done. I am into phase 2 now.

I have also checked the committed notebook and the code, I will follow the same procedure for the upcoming templates as well.

I would love to take your guidance alongside.

Thank you.

[muhtalhakhan]

Alright, you guys can review and submit it to be there in the templates library ^_^

@jravenel @fravenel

please check.

[FlorentLvr]

Alright, you guys can review and submit it to be there in the templates library ^_^

@jravenel @fravenel

please check.

@muhtalhakhan, I don't see any change, I still have the same error as I mentioned above.
Did you push your changes to your branch?

[muhtalhakhan]

Hey @fravenel, I had checked and checked it by now. Every cell is working properly.

Attaching the screenshots.
Yeah, I pushed over to the branch.

If you still find anything missing then do let me know as I would be needing help then.

[muhtalhakhan]

Reviewed the changes to the commit you did way before @jravenel

Thank you!

[jravenel]

@muhtalhakhan can you please check in a Naas cloud environment? I see you are in vs code but as explained in README it's critical we are all in a same env for testing. This is what Naas cloud is for. Thanks a lot 🙏

[jravenel]

Reviewed the changes to the commit you did way before @jravenel

Thank you!

Not sure I get that 😅

[muhtalhakhan]

@muhtalhakhan can you please check in a Naas cloud environment? I see you are in vs code but as explained in README it's critical we are all in a same env for testing. This is what Naas cloud is for. Thanks a lot 🙏

Okay, I got your point @jravenel, now I will test it there at Naas Cloud Env.

Sorry for the hassle.

[muhtalhakhan]

Reviewed the changes to the commit you did way before @jravenel
Thank you!

Not sure I get that 😅

😳

[muhtalhakhan]

Hey @jravenel ,
I have worked upon the errors and removed the gradio for now and would add it after sometime when I find out about it's use in Jupyter Notebooks.

cc: @fravenel

[muhtalhakhan]

fixed issue #1189

[jravenel]

@muhtalhakhan I just tried the template. I managed to install the module and run the notebook but here are my feedback :

1/ the h2 is redundant with the title, can you remove it?

2/ the description is not enough to understand the use case and misleading, you talk about searching in a document but there is no input document provided.
Either we add an input document like the bitcoin white paper pdf example (at it can be retrieve online) or either we remove this part, don't you think?

3/ try expect to install transformer and then in the cell below another time seems redundant also here

4/ a few test I made where giving irrelevant reponses like when I ask the number of countries and the answer is "countries"

5/ you say in the description that you can run without context but the code says "cannot be empty"

Overall, it needs a bit of rework, I think you can increase value by passing a document, otherwise it's not really obvious what this template can bring. I hope this feedback is useful?

[jravenel]

removing Community Roadmap from Projects section, it should not be in PR, cf README

[muhtalhakhan]

@muhtalhakhan I just tried the template. I managed to install the module and run the notebook but here are my feedback :

1/ the h2 is redundant with the title, can you remove it?

2/ the description is not enough to understand the use case and misleading, you talk about searching in a document but there is no input document provided. Either we add an input document like the bitcoin white paper pdf example (at it can be retrieve online) or either we remove this part, don't you think?

3/ try expect to install transformer and then in the cell below another time seems redundant also here

4/ a few test I made where giving irrelevant reponses like when I ask the number of countries and the answer is "countries"

5/ you say in the description that you can run without context but the code says "cannot be empty"

Overall, it needs a bit of rework, I think you can increase value by passing a document, otherwise it's not really obvious what this template can bring. I hope this feedback is useful?

Hey @jravenel, I will work over your review and make more sense in the description.

1/ Yeah, I will change this.

2/ Yeah, we can add document there already but for that I will be needing some help to embed it there while accessing the template as it is using a pipeline which needs a context as an input and context here can be the text from the data or any document's data and afterwards a question would be asked which would be starting with question words.

3/ Yeah, I feel that I overlooked it. Will remove it for sure.

4/yeah, it will fail that countries one as you have not provided an input stating the countries name or some text describing the context for which you are asking the question.

5/I will double check my description as with an empty context we cannot run and ask question.

Yeah the feedback is valuable and detailed, If you find sometime tomorrow @ Friday then I want to lend some time with you related to this "PR".

Thank you

Spidey

[muhtalhakhan]

removing Community Roadmap from Projects section, it should not be in PR, cf README

could not understand the "cf README".

[jravenel]

cf = refer to README.

1/ good, let's be more specific on this template being able to answer questions from a PDF document
2/ I was thinking just adding in the Input section a variable : DOCUMENT_PATH = "https://bitcoin.org/bitcoin.pdf"
3/ good
4/ the question is really easy to anticipate the need from a user perspective, documentation of the notebook need to be explicit with example and I think focusing on bitcoin use case will help,
5/ good

All clear?

[muhtalhakhan]

cf = refer to README.

1/ good, let's be more specific on this template being able to answer questions from a PDF document 2/ I was thinking just adding in the Input section a variable : DOCUMENT_PATH = "https://bitcoin.org/bitcoin.pdf" 3/ good 4/ the question is really easy to anticipate the need from a user perspective, documentation of the notebook need to be explicit with example and I think focusing on bitcoin use case will help, 5/ good

All clear?

1/ already working on it.

2/Nice idea, I can add a thing to fetch the text from pdf.

3/thank you!

4/Exactly but it requires some input to be answered, and yeah will be adding that bitcoin use case.

5/on it.

Yeah.. Will get back to you.

[jravenel]

For fetching the text from pdf @muhtalhakhan check out what @MinuraPunchihewa is working on in the product roadmap he has done a template for that!

[muhtalhakhan]

For fetching the text from pdf @muhtalhakhan check out what @MinuraPunchihewa is working on in the product roadmap he has done a template for that!

Yeah, I saw that but I'll be using a library called as PyPDF.

[muhtalhakhan]

Hey @jravenel made the changes accordingly, please check and let me know if anything further needs my attention.

Thank you!

[muhtalhakhan]

Hey @jravenel , your review is required.

Please check kindly.

Do let me know!

[jravenel]

Hey @jravenel , your review is required.

Please check kindly.

Do let me know!

@muhtalhakhan did you test the notebook on your end because I'm not able to get anything from the notebook: no answers.
I ask the simple question "What is bitcoin?" where I would expect the first occurrence to be replied "A Peer-to-Peer Electronic Cash System" but nothing comes out, check the GIF here to see.
If you have a question that actually works please share, I cannot validate as it is as there is nothing to show.
Thanks for your feedback

[muhtalhakhan]

I will be checking it all again and if I couldn't do anything useful I'll close that PR with a comment mentioning the issue there. But- I need a day or two for sitting against the laptop.

…

On Tue, 1 Nov 2022, 05:49 Jeremy Ravenel, ***@***.***> wrote: Hey @jravenel <https://github.com/jravenel> , your review is required. Please check kindly. Do let me know! @muhtalhakhan <https://github.com/muhtalhakhan> did you test the notebook on your end because I'm not able to get anything from the notebook: no answers. I ask the simple question "What is bitcoin?" where I would expect the first occurrence to be replied "A Peer-to-Peer Electronic Cash System" but nothing comes out, check the GIF here to see. If you have a question that actually works please share, I cannot validate as it is as there is nothing to show. Thanks for your feedback [image: Nov-01-2022 01-45-19] <https://user-images.githubusercontent.com/21052349/199135147-7cab5c8c-7ce2-4f9c-96f0-fa1eda2c235b.gif> — Reply to this email directly, view it on GitHub <#1188 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/APA3FDA7SWYXLUAXELVV24LWGBSJZANCNFSM6AAAAAAQYIPNEY> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[muhtalhakhan]

Hey @jravenel , I have attached the video and screenshot of the notebook that how is it working and checked it again.

Looking forward.

Naas.mp4

[muhtalhakhan]

[jravenel]

Hey @muhtalhakhan it worked for me now.
Maybe was doing it too late 😇😅

I asked
What is it bitcoin → An electronic ...
Who created bitcoin → Satoshi Nakamoto
...

I guess that's good enough: I'm happy that it's working.
Other than that, the IMO framework was not optimal so please check the diff for next time, I basically the whole structure of the sections to make it more readable.

Thanks for the this contribution, hope to see you do more :)

[muhtalhakhan]

It made my day 😎
Thank you @jravenel 🥂