If I have a set of documents, and information on which documents a particular user "liked" (doesn't matter what that means, it's a number), how can I use Spacy to determine whether a new document will be liked by that user?
I could get the vector, then calculate the average, and see how close that was to the average vector for other stories which were liked, but that loses all the detail in the individual vectors.
If we were talking about two dimensions, I guess I want this:
Where the blurry black dots are coordinates where things multiple things have been liked. The green dot will be liked, because it's near a black dot. The red dot probably won't be liked, because it's far from any black dot.
How can I do that?
Comments
Post a Comment