Is one data effective enough for the model to not get it wrong next time?

Question

Context So my boss wants me to make a model that has following properties.

1. NLP-sentiment classification
2. We need a model for each domai