Identifying Entities Using a Deep-Learning Model

Patent metadata

Field	Value
Publication number	US-2017193390-A1
Application number	US-201514984956-A
Country	US
Kind code	A1
Filing date	Dec 30, 2015
Priority date	Dec 30, 2015
Publication date	Jul 6, 2017
Grant date	—

Abstract

In one embodiment, a method includes accessing a first set of entities, with which a user has interacted, and a second set of entities in a social-networking system. A first set of vector representations of the first set of entities are determined using a deep-learning model. A target entity is selected from the first set of entities, and the vector representation of the target entity is removed from the first set. The remaining vector representations in the first set are combined to determine a vector representation of the user. A second set of vector representations of the second set of entities are determined using the deep-learning model. Similarity scores are computed between the user and each of the target entity and the entities in the second set of entities. Vector representations of entities in the second set of entities are updated based on the similarity scores using the deep-learning model.

First claim

1 . A method comprising:
accessing, by one or more computing devices:
a first set of entities that a user of a social-networking system has interacted with in the social-networking system, and
a second set of entities in the social-networking system;
determining, by one or more computing devices, a first set of vector representations of the first set of entities using a deep-learning model; selecting, by one or more computing devices, a target entity from the first set of entities; removing, by one or more computing devices, from the first set of vector representations the vector representation of the target entity; combining, by one or more computing devices, the remaining vector representations in the first set of vector representations to determine a vector representation of the user; determining, by one or more computing devices, a second set of vector representations of the second set of entities using the deep-learning model; computing, by one or more computing devices:
a similarity score between the target entity and the user by comparing the vector representation of the user with the vector representation of the target entity, and
similarity scores between the user and the entities in the second set of entities by comparing the vector representation of the user with the vector representations of the entities in the second set of entities; and
updating, by one or more computing devices, the vector representations of one or more entities in the second set of entities based on the similarity scores using the deep-learning model.

2 . The method of claim 1 , further comprising:
determining, by one or more computing devices, an embedding for the user, each entity of the first set of entities, and each entity of the second set of entities, wherein:
each embedding corresponds to a point in a multi-dimensional embedding space,
the embedding space comprises a plurality of points corresponding to a plurality of entities; and
each embedding is based on a respective vector representation determined using the deep-learning model.

3 . The method of claim 1 , further comprising assigning, by one or more computing devices, a respective ranking to the target entity and to each entity of the second set of entities based on the similarity scores, and wherein updating the vector representations of the one or more entities of the second set of entities is further based on the rankings.

4 . The method of claim 3 , wherein each of the one or more entities of the second set of entities whose vector representations are updated have a similarity score that is greater than the similarity score of the target entity.

5 . The method of claim 1 , further comprising identifying one or more entities plurality of entities as relevant to the user by applying a search algorithm to the embedding space, wherein points corresponding to one or more embeddings of the identified entities are within a threshold distance of the point corresponding to the embedding of the user in the embedding space.

6 . The method of claim 5 , further comprising sending, to a client system of the user, the one or more identified entities for display to the user.

7 . The method of claim 1 , wherein target entity is selected randomly.

8 . The method of claim 1 , wherein the user has not interacted with the second set of entities in the social-networking system.

9 . The method of claim 1 , wherein the social-networking system comprises a social graph comprising a plurality of nodes and a plurality of edges connecting the nodes, each of the edges between two of the nodes representing a single degree of separation between them, the nodes comprising:
a first node corresponding to the user; and a plurality of second nodes that each correspond to a respective entity.

10 . The method of claim 9 , wherein the user has interacted with the first set of entities by a social-networking action of the user, wherein the social-networking action is taken with respect to the first node and a respective second node corresponding to a respective entity of the first set of entities.

11 . The method of claim 10 , wherein the social-networking action represents an expression of affinity for an entity.

12 . The method of claim 1 , wherein at least one of the entities comprises a page hosted by the social-networking system.

13 . The method of claim 1 , wherein the vector representations comprise d-dimensional intensity vectors.

14 . One or more computer-readable non-transitory storage media embodying software that is operable when executed to:
access:
a first set of entities that a user of a social-networking system has interacted with in the social-networking system, and
a second set of entities in the social-networking system;
determine a first set of vector representations of the first set of entities using a deep-learning model; select a target entity from the first set of entities; remove from the first set of vector representations the vector representation of the target entity; combine the remaining vector representations in the first set of vector representations to determine a vector representation of the user; determine a second set of vector representations of the second set of entities using the deep-learning model; compute:
a similarity score between the target entity and the user by comparing the vector representation of the user with the vector representation of the target entity, and
similarity scores between the user and the entities in the second set of entities by comparing the vector representation of the user with the vector representations of the entities in the second set of entities; and
update the vector representations of one or more entities in the second set of entities based on the similarity scores using the deep-learning model.

15 . The media of claim 14 , wherein the software is further operable when executed to:
determine an embedding for the user, each entity of the first set of entities, and each entity of the second set of entities, wherein:
each embedding corresponds to a point in a multi-dimensional embedding space,
the embedding space comprises a plurality of points corresponding to a plurality of entities; and
each embedding is based on a respective vector representation determined using the deep-learning model.

16 . The media of claim 14 , wherein the software is further operable when executed to assign a respective ranking to the target entity and to each entity of the second set of entities based on the similarity scores, and wherein updating the vector representations of the one or more entities of the second set of entities is further based on the rankings.

17 . A system comprising: one or more processors; and a memory coupled to the processors comprising instructions executable by the processors, the processors operable when executing the instructions to:
access:
a first set of entities that a user of a social-networking system has interacted with in the social-networking system, and
a second set of entities in the social-networking system;
determine a first set of vector representations of the first set of entities using a deep-learning model; select a target entity from the first set of entities; remove from the first set of vector representations the vector representation of the target entity; combine the remaining vector representations in the first set of vector representations to determine a vector representation of the user; determine a second set of vector representations of the second set of entities using the deep-learn

Assignees

• Facebook Inc

Inventors

• Weston Jason E
• Adams Keith
• Chopra Sumit

Classifications

• G06Q10/40

  Business processes related to social networking or social networking services · CPC title

• G06N20/00 Primary

  Machine learning · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N99/005 Primary

  Physics · mapped topic