Can I use a pre-trained model for sentiment analysis/text classification of unlabelled data?

Question

I'm planning on working on a project where I'll have a large collection of tweets about coronavirus vaccines. None of the tweets will have a label (e.g. positive, neutral, negative). Therefore I won't be able to train a model based on the labels.

I have a vague understanding of pre-trained models like BERT or VADER. I don't know however if I can use a model trained on other (text) data (like the ones mentioned above) and use it to run a sentiment analysis for the tweets I have.

Is it possible to do this? Or would it require labeled data in order to train the model with that specific data relating to the vaccine tweets?

Answer 1

You need at least a few labelled vaccine tweets (positive, neutral, negative) to train a BERT model so that it starts to understand the domain.

For VADER you don't need any labelled data.

However, when we compare the accuracies, the BERT model always performs better.