here after the user uploads the CSV file, we still get to see the red error message.

Do you expect me to see this?

Tested in OSX 10.15.2

We have a BERT-based binary(?) classifier and a dev or test set; we want to return the predictions the model has made. Should return a list of predictions, one for each item in the posted dev/test set.

Sending an email to the user when their framing classification results is ready that contains the link to their results on Google Drive (bu has unlimited Google Drive).

Give error to the user if they don't enter an email address or if they enter something that is not in accordance with regex of the email.

• add random article sampler that randomly samples k articles and creates a spreadsheet for labeling
• add intake for labeled sample and full dataset (assumes the original dataset contains the articles in the labeled sample)
• do naive bayes to create a labeled dataset using the small labeled sample
• create a spreadsheet that can be uploaded to the classifier endpoint

After we select the 'policy issue' endpoint, we need an endpoint to use it in Ajax.

Current incomplete code for processing it is:

	$("button[type='submit']").on('click', function (){
		let radioValue = $("input[name='gunviolence']:checked").val();
		alert(radioValue)
		console.log(radioValue)
		debugger;
	})

if this is a duplicate issue, please link it to the other issue/pull.

The current version is:

Given a training set file (I'm going to try to support csv, tsv, and Excel) with a content column and a label column, we fine-tune and save a BERT-based classifier.

In the event that multiple people want to use the BERT training at one time, we'll need to figure out queueing. David suggested Python's multiprocessing library, which is what I think I'll be using.

Users should enter the following if they choose to sign-up and create an account:

• user Full Name
• username
• user email
• password

Users should enter the following if they choose to sign-in and login to their account:

• username or email (that they used for creating the account)
• password
• an option for forgetting the password and resetting the password
• an option for forgetting the username/password

The page should also have the following at the very minimum:

• logo of OpenFraming
• Sign-in (if the user already has an account with us)
• Sign-up (if the user already doesn't have an account with us)

A 2 minute video to send along with EMNLP 2020 submission to show how to use the system, and what are the features.

Given a validation set and a fine-tuned BERT-based (binary) classifier, return the following:
{
"macro_f1": float,
"macro_precision": float,
"macro_recall": float,
"accuracy": float
}

May need to do this for each category if we're doing one binary classifier per category.

• the annotated samples shouldn't be too short or too long (for example check a minimum of 500-1000 annotated samples)

the fix is an amazing one. Thanks @edward-edberg

Could you now try to fix the following which was already correct? Previously when we were clicking on 'create an account' after we click on 'I'm already a member' took us back to 'sign up' section (in your case, page). Now, it doesn't.

Upload a training file to a classifier and start training.

For now, we'll support only CSV files, but we can (and should) always add formats later(especially Excel).

*We need to create a tutorial for the user to learn how to use the Website quickly

*Mona creates the video

*Alyssa annotates the video with voice

*Embed the video on the webpage

As per Margrit's suggestion to make the UI more user-friendly

We need to have the email address of the user so that when the results are available we can send them the result.

*We should also use methods such as email address validation to make sure the user is entering a correctly formatted email address.

do form validation using jQuery to make sure the frames entered by the user are comma-separated

The line is linked below. It looks like a bug to me(dir_name is not defined at that point), but I just wanted to confirm with you @lentil-soup .

https://github.com/davidatbu/openFraming/blob/30b506a6eb223057a110de8e8f1424ae6631766f/backend/modeling/classifier.py#L47

currently, the password is in SECRET_KEY = in db.py

*Also showing the default values of frames for gun violence policy issue.

Suggestions --> please take a look on how clicking on 'create an account' in sign in section takes us to sign up section and follow same method of thoughts.

*Am waiting on Derry and Lei to give me the paragraph to visualize it.

This is very important so the user get an idea what they get as result and what the task actually is about.

currently has a problem

currently we are not entering the index.html after we sign-in to the account

We should probably discuss this in more depth - but are there any specific considerations I should have in mind when designing the architecture for training multiple binary classifiers? We /are/ doing multiple binary classifiers, right? (not a multiclass classifier)? In that case, will we need to modify the dev set accuracy to include scoring for all classes?

current screenshot of the problem