You pretty much summed up the drawbacks of the embeddings approach. In my experience it's pretty hard to extract the relevant parts of text, especially when the

You pretty much summed up the drawbacks of the embeddings approach. In my experience it's pretty hard to extract the relevant parts of text, especially when the text is uniform.