Personalized deep models for smart suggestions ranking
In an example, a deep learning network is used to calculate a similarity score between a first query in a social networking service and each of one or more suggestable entities in the social networking service. The suggestable entities are determined via a first machine learned model. The deep learning network takes as input the suggestable entities as well as a history of interactions with a graphical user interface of a social networking service by a first member of the social networking service, a history of queries performed via the graphical user interface by the first member, and the first query itself.
1. A system comprising:
one or more processors;
a computer-readable medium having instructions stored thereon, which, when executed by the one or more processors, cause the system to:
obtain a history of interactions with a graphical user interface of a social networking service by a first member of the social networking service;
obtain a history of queries performed via the graphical user interface by the first member;
receive a first query initiated by the first member, the first query identifying a first set of one or more entities in the social networking service;
form a query facet entities vector from the first set of the one or more entities;
submit the query facet entities vector to a first machine learned model to identity a set of one or more suggestable entities, each of the one or more suggestable entities being an entity similar to, but not identical to, an entity in the first set of the one or more entities, based on embeddings learned by the first machine learned model;
form a set of suggestable entities vectors from the set of the one or more suggestable entities;
form a set of history of interactions vectors from the history of interactions;
form a set of history of queries vectors from the history of queries;
pass the query facet entities vector, the set of suggestable entities vectors, the set of history of interactions vectors, and the set of history of queries vectors into a deep learning network to calculate a similarity score between the first query and each suggestable entity in the set of the one or more suggestable entities;
rank the set of the one or more suggestable entities based on the similarity scores; and
cause display of one or more of the suggestable entities based on the ranking.
2. The system of claim 1 , wherein the deep learning network includes an aggregation layer that aggregates the history of interactions vectors and the history of queries vectors separately.
3. The system of claim 2 , wherein the aggregation layer includes one or more long short term memories (LSTMs).
4. The system of claim 2 , wherein the aggregation layer includes one or more gated recurrent units.
5. The system of claim 1 , wherein the first machine learned model calculates, based on co-occurrence counts reflecting a number of user profiles in a plurality of user profiles in which corresponding nodes co-occurred, a similarity score between a first node and second node by computing distance between the first node and the second node in a d-dimensional space on which a plurality of entities are mapped.
6. The system of claim 5 , wherein the similarity score is generated using a generalized linear mixed model having a global coefficient vector applied to global function pertaining to the co-occurrence counts and a first random effects coefficient vector applied to a random effects per-country function.
7. The system of claim 1 , wherein the deep learning network is a deep and wide model.
8. A computerized method comprising:
obtaining a history of interactions with a graphical user interface of a social networking service by a first member of the social networking service;
obtaining a history of queries performed via the graphical user interface by the first member;
receiving a first query initiated by the first member, the first query identifying a first set of one or more entities in the social networking service;
forming a query facet entities vector from the first set of the one or more entities,
submitting the query facet entities vector to a first machine learned model to identify a set of one or more suggestable entities, each of the one or more suggestable entities being an entity similar to, but not identical to, an entity in the first set of the one or more entities, based on embeddings learned by the first machine learned model;
forming a set of suggestable entities vectors from the set of the one or more suggestable entities,
forming a set of history of interactions vectors from the history of interactions,
forming a set of history of queries vectors from the history of queries;
passing the query facet entities vector, the set of suggestable entities vectors, the set of history of interactions vectors, and the set of history of queries vectors into a deep learning network to calculate a similarity score between the first query and each suggestable entity in the set of the one or more suggestable entities,
ranking the set of the one or more suggestable entities based on the similarity scores; and
causing display of one or more of the suggestable entities based on the ranking.
9. The method of claim 8 , wherein the deep learning network includes an aggregation layer that aggregates the history of interactions vectors and the history of queries vectors separately.
10. The method of claim 9 , wherein the aggregation layer includes one or more long short term memories (LSTMs).
11. The method of claim 9 , wherein the aggregation layer includes one or more gated recurrent units.
12. The method of claim 8 , wherein the first machine learned model calculates, based on co-occurrence counts reflecting a number of user profiles in a plurality of user profiles in which corresponding nodes co-occurred, a similarity score between a first node and second node by computing distance between the first node and the second node in a d-dimensional space on which a plurality of entities are mapped.
13. The method of claim 12 , wherein the similarity score is generated using a generalized linear mixed model having a global coefficient vector applied to global function pertaining to the co-occurrence counts and a first random effects coefficient vector applied to a random effects per-country function.
14. The method of claim 8 , wherein the deep learning network is a deep and wide model.
15. A non-transitory machine-readable storage medium comprising instructions, which when implemented by one or more machines, cause the one or more machines to perform operations comprising:
obtaining a history of interactions with a graphical user in of a social networking service by a first member of the social networking service;
obtaining a history of queries performed via the graphical user interface by the first member;
receiving a first query initiated by the first member, the first query identifying a first set of one or more entities in the social networking service;
forming a query facet entities vector from the first set of the one or more entities,
submitting the query facet entities vector to a first machine learned model to identify a set of one or more suggestable entities, each of the one or more suggestable entities being an entity similar to, but not identical to, an entity in the first set of the one or more entities, based on embeddings learned by the first machine learned model;
forming a set of suggestable entities vectors from the set of the one or more suggestable entities;
forming a set of history of interactions vectors from the history of interactions;
forming a set of history of queries vectors from the history of queries;
passing the query facet entities vector the set a suggestable entities vectors the set of history of interactions vectors, and the set of history of queries vectors into a deep learning network to calculate a similarity score between th first query and each suggestable entity in the set of the one or more suggestable entities;
ranking the set of the one or more suggestable entities based on the similarity scores; and
causing display of one or more of the suggestable entities based on the ranking.
16. The non-transitory machine-readable storage medium of claim 15 , wherein the deep learning network includes an aggregation layer that aggregates the history of interactions vectors and the history of queries vectors separately.
17. The non-transitory machine-readable storage medium of claim 16 , wherein the aggregation layer includes one or more long short term memories (LSTMs).
18. The non-transitory machine-readable storage medium of claim 16 , wherein the aggregation layer includes one or more gated recurrent units.
19. The non-transitory machine-readable storage medium of claim 15 , wherein the first machine learned model calculates, based on co-occurrence counts reflecting a number of user profiles in a plurality of user profiles in which corresponding nodes co-occurred, a similarity score between a first node and second node by computing distance between the first node and the second node in a d-dimensional space on which a plurality of entities are mapped.
20. The non-transitory machine-readable storage medium of claim 19 , wherein the similarity score is generated using a generalized linear mixed model having a global coefficient vector applied to global function pertaining to the co-occurrence counts and a first random effects coefficient vector applied to a random effects per-country function.