Open ironllamagirl opened 2 years ago
Since this is a classification problem rather than a semantic search problem, you could just use a pretrained bert model as a bi-encoder.
You can use the CrossEncoder also for individual texts. So you pass a list with individual sentences instead of sentence pairs
@nreimers Oh I did not know that. I will try it out Thanks!!
I'll note that the examples and documentation are not clear that one can use just a single input sentence. (Also, i'm getting a tokenizer/padding error when using the CE + softmax evaluator, but that's a seperate issue)
I have a text classification problem where I need to classify text into one of 4 categories. I would like to use sbert but read that crossencoder only takes pair input.
How do I go about doing this? is there an example code? Thank you so much