We need labeling of emojis

[pomkos]

I want to start expanding algo3_5.py. Right now it can count words/emojis/both per person. I would like to expand it to count "sentiments" per person.

Good way to start doing this is by labeling the emoticons that Twitch uses. They have definitions but these definitions need to become more categories. Maybe 6 categories (fear, anger, joy, sadness, surprise and disgust). These follow the "basic emotions" theory, there was a recent meta-analysis of this concept and also here is a wiki.

Anyways, 6 categories for the emoticons doable? Whoever knows Twitch culture should take the lead on this.

I added Matt and Steve cuz I know they do Twitch a lot, but feel free to assign yourselves!

[pomkos]

Commit d503437 includes the data/emoticons.csv, scraped from Twitch link.

[pomkos]

@gatesyp and I talked. We might ask users to label their channel's most frequently used custom emoticons

Default twitch emoticons: label them as they come up, create a table in csv. @gatesyp @Geczy will label them.

[gatesyp]

Quick note to expand on the labelling tool for users -- thats a nice way to scale the categorization of custom emotes. Not needed now, but if this approach yields good results then we can consider building the labelling tool for users.

[pomkos]

We now have the ability to extract all unique emoticon ids and names, along with the number of times they were used in a particular dataset. script here

Output here in csv form.

[pomkos]

Merged into unit_testing branch, cuz thats where all my deving is happening.

[Geczy]

amazon has a labeling tool for training data sets

https://aws.amazon.com/getting-started/hands-on/build-training-datasets-amazon-sagemaker-ground-truth/

[pomkos]

We need more labeled emoticons @gatesyp @Geczy