Transformer Update
This week, I actually managed to make a small amount of progress on my transformer! A bit of better time management enabled me to actually spend valuable time writing code and doing research for my emotion classification transformer. Overall, I complete two main tasks: the first is that I was able to successfully begin managing my data for the transformer and the second is that figured out how I am going to go about creating my NN.
Before I began processing the data, I copied down a simple function from a Google tutorial that I could use in order to strip unnecessary characters from the text entries that would be the data input for the transformer. This is important because unknown characters will make it impossible to find word vectors for the associated words in the text. Next, I used a simple Pandas script to extract the Emobank data, sanitize the data with the above function, and split the data into test and train datasets.
After writing this code, it was time to start writing the transformer. However, upon further research, I realized that I simply don’t have the knowledge to write a transformer from scratch. As such, I scrapped my original plan and made the choice to use the HuggingFace Transformers library.
Currently, I don’t have the actual transformer written, so hopefully next time I will have a bit more to update on that front.
