Text classification using ULMFiT giving very low accuracy

(Seemant) #1

I was working on text classification and thought of using Fastai for that, but I can only get an accuracy of about 25% on the dataset. I think I did everything right but then also I am not getting good accuracy.

The dataset can be found here. We have to predict the category of news articles.

It would be helpful If some of you guys can try this dataset with Fastai to figure out what is not working here and share their experiences.

0 Likes

#2

I am also doing this dataset and getting the same issue. Here is my code if you want to check out: https://colab.research.google.com/drive/1NR_AL2HuClWCFbgCrwKpFUh40FzjCQnp (ive only run up to the first fit_one_cycle)

My model predicts Politics for everything as it is most prevalent.

0 Likes