Complete: Swaroop: Tasks313-316 DialogRE

[matthew-huff]

No description provided.

[swarooprm]

@eshaanpathak, wondering if you can help review these tasks?

[eshaanpathak]

@swarooprm Yeah, I can review this. I'm guessing you made this research project an assignment for a class you're TAing for haha? Can you also update the src/test_all.py with the latest version that we have in the AI2 repo? Can you also link the AI2 repo in the main README.md to give everybody here more ideas on how they should format their tasks?

@matthew-huff Hey Matthew. Thanks for the submission!

1. In tasks/README.md, switch out the apostrophe ('' character) with the grave accent (`` character that is usually the same key as the ~~ character on your keyboard). This way it will highlight and look nice in the markdown file in tasks/README.md.
2. In tasks/README.md, make sure all characters in the filename are lowercased and that the .json file extension ending is not present.
3. Instead of two spaces, you should have one space followed by a newline character before each speaker in the dialog. For example, "Speaker 1: Hello! \nSpeaker 2: Hi! \nSpeaker 1:". This will make each speaker be on a new line.
4. You should add a third positive example to your tasks.
5. You should also make your filenames more descriptive of the task, though not too long.

Please let me know when you have made these changes and if you have any questions!

[matthew-huff]

@eshaanpathak Hi Eshaan, thanks for the pointers. Hopefully these files are all fixed now!

[eshaanpathak]

Hey @matthew-huff, here are some more edits I recommend:

1. Make sure to stay consistent throughout the examples on adding the space and new line characters. Some only have a new line or just a space still.
2. Make sure there also new lines throughout the instances as well and not just in the examples.
3. I think the last task should be categorized as "Answer Generation".

[swarooprm]

@matthew-huff, can you address reviewer comments so that I can evaluate your HW?

[matthew-huff]

@matthew-huff, can you address reviewer comments so that I can evaluate your HW?

Hi @swarooprm , I must've missed the additional notes from Eshaan. I'll fix those up and tag you in a comment here once I'm finished, thanks for the heads up!

[matthew-huff]

Hi @swarooprm , I fixed those newline errors and committed. Thanks again.

[eshaanpathak]

For tasks 313 and 316, can you actually save the JSON file with an indent=4 so it's easy to read? Like how you did for tasks 314 and 315.

Some more advice:

1. You don't need to add newline characters before the very first speaker in the dialog.
2. And make sure you add newline characters to your positive and negative examples too.

[eshaanpathak]

Thanks for the changes! Here are some more suggestions:

1. I think you may have missed adding newline characters to some of the speakers in the examples.
2. You also don't need to add two spaces before the newline character, just one space is fine.

[eshaanpathak]

One last thing: can you save your JSON files in utf-8 encoding? There are some examples and instances that have unicode encodings such as "\u2019" when they should be apostrophes, quotations, etc.

[swarooprm]

Your HW2 is complete (assuming that you fix the utf-8 encoding issue)

Would you like to submit the same tasks to our big project repo /~https://github.com/allenai/natural-instructions-expansion/pulls? You just need to fork that repo, rename tasks starting from the latest number in the repo and create PR there. This way, you can be part of our research on "learning from instructions instead of learning from examples (conventional machine learning)". This is purely optional and your HW2 grade does not depend on this @matthew-huff

[matthew-huff]

@swarooprm I just fixed that UTF-8 issue. Sure, I'll fork that repo and add the tasks there.

[swarooprm]

@matthew-huff, Just a reminder that, if you plan to submit these tasks to the main Github repo and join our research project, Monday (27th Sept) is the deadline. Again, this is optional and independent of your HW2 grade.