In Sentiment Analysis when we say that a statement is Positive or Negative, what do we mean by these.
For example, someone says “iPhone is good device”, so this is a positive sentence from the Apple Company view side, however that is a Negative one from Samsung Company side.
Where did you get the Samsung example from? By positive and negative we mean the sentiment that is attached to an utterance. Is it positive (good) or negative (bad). Both your examples, with Apple and Samsung, should be good.
That was just an example. So, let me tell it in another way. Suppose that we are going to measure how many people have good thinking about “Apple products” and how many are against it by analyzing some tweets. Can we do it with sentiment-analysis and transformers? (we do not have labeled data)
You can find a model that has been fine-tuned one sentiment data, and use that.
Thanks Bram,
Just the last question: For these fine-tuned models, it means that, those guys have any dataset with labeled data and using BERT to tune it according to their requirements, is that right?
Yes, they finetuned an existing base model such as BERT on the downstream task of sentiment analysis using a dataset that they had available.
Thanks, Bram. Solved!