This repository corresponds to the paper, A Review of the Unified Corpus and a Methodology for Improvement on Generalizable Emotion Detection. Methods within the repository are labelled and/or are self explanatory as to their purpose, however the file structure here may not be as simple. Files in this repository are labelled such that the first name is the train set, the second name is the test set, and a mode of single or multi is given if applicable. In the case of similarity score comparisons, the files are given in the same form of first and second, but no mode is applicable. In the jupyter notebook files, you will find test cases related to verifying the validity of the code as well as saved outputs for many of them. Other outputs can be found through old commit history files. While the files are very large, if you wish to run a classifier, it is advised that you download the pretrained classifier from the following google drive. I would have uploaded these files to the repo, but they are far too large for github.
https://drive.google.com/drive/folders/1vu8s0qqxYQcaEB7l22nYiXgyCSWA1X2I?usp=sharing
The primary file to focus on is main.ipynb. The file main-Copy2.ipynb also has some useful saved outputs, as this file was created for the purpose of running another jupyter notebook in parallel with the first on a seperate kernel for corpus similarity trials. There are likely small differences in the code between these two files, but for the purpose of the assignment, you can focus on main.ipynb, and refer to main-Copy2.ipynb as simply a reference for outputs on corpus similarity.
Some sample code is shown in analyzeResults.ipynb, however, I was unable to produce any worthwhile visualizations of data before the deadline, and I decided to scrap those ideas in favor of focusing on my paper. With very little modification, the results can still be viewed as a table in this file however.
As you can see, there is also a file called untitled.ipynb
This file was uploaded near the beginning of the project, but due to a glitch with how I am running the github project, I'm unable to remove the file from the repository. As such, you can completely ignore this file.
Another thing to ignore would be the folder of cited papers, as I was originally saving these for personal use when I uploaded them to github. It is not up to date with the final paper, and so, if you wish to view the papers cited, it would be best to simply check the references of my paper, where most of the papers have hyperlinks to online free pdfs.