I have a scenario in which we should leverage previously asked questions (not questions pairs, single question in a column) to locate similar questions within those questions.
How can I fine-tune my model to manage out of vocabulary, as my data includes domain-specific questions (3300 questions)?.
Right now, I'm using hugging face sentence transformers, which is already pre-trained on huge data.
For example, BERT knows that gold is a metal, but, in our domain corpus, it's a platform. We have some terminologies which were not exposed openly, how can I fine-tune the model to get related sentences (handling OOV).