I am looking for the simplest way to train a fast.ai model to classify relations between two entities in a text sentence.
I have already manually collected the training data which consists of sentences. And I have identified entities which I am interested in and classified their relations manually.
Now, I would like to train a fast.ai model on my training data.
I will use pre-trained word embeddings from the Word2Vec Google News which is suitable for my problem. I probably have to mask my entities somehow because I don’t want to the model to memorise the entity names; I want the context to be the feature, not the entity name itself.
What is the simplest way to build and train a fast.ai model for my problem?