Content recommendation system using a neural network language model
The present disclosure relates to applying techniques similar to those used in neural network language modeling systems to a content recommendation system. For example, by associating consumed media content to words of a language model, the system may provide content predictions based on an ordering. Thus, the systems and techniques described herein may produce enhanced prediction results for recommending content (e.g. word) in a given sequence of consumed content. In addition, the system may account for additional user actions by representing particular actions as punctuation in the language model.
1. A computer-implemented method of providing recommendations, comprising:
obtaining a user history for a user, the user history identifying a plurality of items, the plurality of items comprising items representing one or more media items presented to the user and items representing one or more actions performed by the user;
generating a sequence of tokens that includes a respective token associated with each of the one or more media items presented to the user and a respective token associated with each of the one or more actions performed by the user;
providing each token in the sequence of tokens as an input to a recurrent neural network that is configured to process each of the tokens and, after processing a last token in the sequence of tokens, predict a next token subsequent to the last token in the sequence of tokens; and
providing a recommendation to the user based on an item associated with the predicted next token.
2. The computer-implemented method of claim 1 , wherein the one or more actions performed by the user include one of selecting an advertisement, performing a search, visiting a webpage, navigating a webpage, rating a media item, sharing a media item, and interrupting a played media item.
3. The computer-implemented method of claim 1 , wherein the one or more actions performed by the user include ending a session.
4. The computer-implemented method of claim 1 , wherein the one or more media items presented to the user include videos, music, documents, or applications.
5. The computer-implemented method of claim 1 , wherein the recurrent neural network is configured to associate each of a plurality of candidate tokens with a respective probability.
6. The computer-implemented method of claim 5 , wherein the recommendation to the user is provided as a list identifying items associated with the plurality of candidate tokens for the next token.
7. The computer-implemented method of claim 6 , wherein the items associated with the plurality of candidate tokens for the next token are ordered in the list based on probabilities associated with the plurality of candidate tokens.
8. A system for providing recommendations, comprising:
a processor, the processor configured to:
obtain a user history for a user, the user history identifying a plurality of items, the plurality of items comprising items representing one or more media items presented to the user and items representing one or more actions performed by the user;
generate a sequence of tokens that includes a respective token associated with each of one or more media items presented to the user and a respective token associated with each of the one or more actions performed by the user;
provide each token in the sequence of tokens as an input to a recurrent neural network that is configured to process each of the tokens and, after processing a last token in the sequence of tokens, predict a next token subsequent to the last token in the sequence of tokens; and
provide a recommendation to the user based on an item associated with the predicted next token.
9. The system of claim 8 , wherein the one or more actions performed by the user include one of selecting an advertisement, performing a search, visiting a webpage, navigating a webpage, rating a media item, sharing a media item, and interrupting a played media item.
10. The system of claim 8 , wherein the one or more actions performed by the user include ending a session.
11. The system of claim 8 , wherein the one or more media items presented to the user include videos, music, documents, or applications.
12. The system of claim 8 , wherein the recurrent neural network is configured to associate each of a plurality of candidate tokens with a respective probability.
13. The system of claim 12 , wherein the recommendation to the user is provided as a list identifying items associated with the plurality of candidate tokens for the next token.
14. The system of claim 13 , wherein the items associated with the plurality of candidate tokens for the next token are ordered in the list based on probabilities associated with the plurality of candidate tokens.