Does the paper reach some conclusion on an optimal embedding size? I was thinking about that the other day. It's interesting from a linguistics perspective. I

Does the paper reach some conclusion on an optimal embedding size? I was thinking about that the other day. It's interesting from a linguistics perspective. I wonder if each dimension in an optimal-size vector could be given a human-comprehensible label.