Customized Keyword Query Suggestions on Online Social Networks

Abstract

In one embodiment, a method includes receiving an indication that a user of an online social network (OSN) is interacting with a search interface of the OSN, identifying one or more entities of the OSN associated with the user based at least in part on a search history of the user, identifying, for each identified entity, one or more content objects associated with the entity, wherein each content object is accessible by the user based on privacy settings defining a visibility of the content object as being visible to at least the user, generating, for each identified entity, a plurality of candidate keywords associated with the entity by extracting one or more n-grams from each identified content object, and sending, to the user, one or more suggested queries associated with the entity, each suggested query comprising one or more of the candidate keywords associated with the entity.

1 . A method comprising:
 receiving, from a client system, an indication that a user of an online social network is interacting with a search interface of the online social network;   identifying one or more entities of the online social network associated with the user based at least in part on a search history of the user;   identifying, for each identified entity, one or more content objects associated with the entity, wherein each content object is accessible by the user based on privacy settings associated with the content object, and wherein the privacy settings define a visibility of the content object as being visible to at least the user;   generating, for each identified entity, a plurality of candidate keywords associated with the entity by extracting one or more n-grams from each identified content object associated with the entity; and   sending, to the client system, instructions for presenting one or more suggested queries associated with the entity, each suggested query comprising one or more of the candidate keywords associated with the entity.   
     
     
         2 . The method of  claim 1 , wherein, for one or more of the identified content objects, the privacy settings define the visibility of the content object as being visible to the user based at least in part on the user being associated with one or more groups of entities with respect to the entity associated with the content object. 
     
     
         3 . The method of  claim 2 , wherein a relationship status between one or more of the groups of entities with respect to the identified entity comprise one or more of friends, friends of friends, and public. 
     
     
         4 . The method of  claim 1 , further comprising:
 accessing 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 plurality of entity nodes corresponding to a plurality of entities of the online social network, respectively, wherein the one or more identified entities correspond to a respective one or more first entity nodes of the plurality of entity nodes, and wherein the user corresponds to a second entity node of the plurality of entity nodes; and 
 a plurality of content nodes corresponding to a respective plurality of posts of the online social network, wherein each content node is connected to one or more of the plurality of entity nodes by one or more edges, and wherein each of the identified content objects corresponds to a respective first content node of the plurality of content nodes. 
   
     
     
         5 . The method of  claim 4 , wherein the user is determined to be interacting with a search interface associated with one of the first entity nodes. 
     
     
         6 . The method of  claim 4 , wherein the search history of the user comprises a plurality of queries, wherein at least one of the queries of the search history is associated with one of the first entity nodes. 
     
     
         7 . The method of  claim 4 , further comprising:
 receiving, from the client system, an unstructured text query;   parsing the unstructured text query to identify one or more n-grams; and   determining that at least one of the identified n-grams corresponds to one of the first entity nodes.   
     
     
         8 . The method of  claim 1 , wherein generating, for each identified entity, the plurality of candidate keywords comprises:
 calculating a relevance score for each extracted n-gram based on a TF-IDF analysis; and   generating a candidate keyword corresponding to each n-gram having a calculated relevance score greater than a threshold relevance score.   
     
     
         9 . The method of  claim 1 , wherein generating, for each identified entity, the plurality of candidate keywords comprises:
 accessing one or more images associated with one or more of the identified content objects associated with the entity;   identifying one or more n-grams corresponding to one or more features of the accessed images; and   generating a candidate keyword corresponding to each identified n-gram.   
     
     
         10 . The method of  claim 1 , wherein generating, for each identified entity, the plurality of candidate keywords comprises:
 calculating a rank for each candidate keyword based at least in part on a number of social signals associated with the content object from which the n-gram corresponding to the candidate keyword was extracted; and   storing one or more candidate keywords in association with the entity, each stored candidate keyword having a calculated rank higher than a threshold rank.   
     
     
         11 . The method of  claim 10 , wherein the social signals associated with each content object comprise one or more of:
 an accessing or viewing of the content object;   a comment on the content object;   a sharing of the content object; or   a liking of the content object, the comment on the content object, or the sharing of the content object.   
     
     
         12 . The method of  claim 10 , wherein calculating the rank for each candidate keyword is further based on an analysis of the candidate keyword according to a language model. 
     
     
         13 . The method of  claim 10 , wherein calculating the rank for each candidate keyword is further based on a time of creation associated with the content object from which the n-gram corresponding to the candidate keyword was extracted. 
     
     
         14 . The method of  claim 10 , wherein calculating the rank for each candidate keyword comprises:
 determining that a first candidate keyword appears in a list of trending-topic keywords; and   up-ranking the first candidate keyword based on the determination.   
     
     
         15 . The method of  claim 10 , wherein, for one or more first candidate keywords of the plurality of candidate keywords, calculating the rank for each first candidate keyword comprises:
 determining that the first candidate keyword appears in more than one content objects associated with the entity; and   up-ranking the first candidate keyword based on a number of the content objects associated with the entity comprising the first candidate keyword.   
     
     
         16 . The method of  claim 10 , wherein, for one or more first candidate keywords of the plurality of candidate keywords, calculating the rank for each first candidate keyword comprises:
 determining that one or more n-grams corresponding to a particular number of second candidate keywords, respectively, were extracted from the content object from which the n-gram corresponding to the first candidate keyword was extracted; and   down-ranking the first candidate keyword based on a determination that the particular number of second candidate keywords each has a higher rank than the first candidate keyword.   
     
     
         17 . The method of  claim 10 , wherein storing, for each identified entity, the one or more candidate keywords in association with the entity comprises storing the candidate keywords in one or more of:
 a local cache of the client system; or   one or more data stores associated with the online social network.   
     
     
         18 . The method of  claim 1 , wherein, for each identified entity, one or more of the content objects associated with the entity comprise one or more respective posts authored by the entity on the online social network. 
     
     
         19 . One or more computer-readable non-transitory storage media embodying software that is operable when executed to:
 receive, from a client system, an indication that a user of an online social network is interacting with a search interface of the online social network;   identify one or more entities of the online social network associated with the user based at least in part on a search history of the user;   identify, for each identified entity, one or more content objects associated with the entity, wherein each content object is accessible by the user based on privacy settings associated with the content object, and wherein the privacy settings define a visibility of the content object as being visible to at least the user;   generate, for each identified entity, a plurality of candidate keywords associated with the entity by extracting one or more n-grams from each identified content object associated with the entity; and   send, to the client system, instructions for presenting one or more suggested queries associated with the entity, each suggested query comprising one or more of the candidate keywords associated with the entity.   
     
     
         20 . A system comprising: one or more processors; and a non-transitory memory coupled to the processors comprising instructions executable by the processors, the processors operable when executing the instructions to:
 receive, from a client system, an indication that a user of an online social network is interacting with a search interface of the online social network;   identify one or more entities of the online social network associated with the user based at least in part on a search history of the user;   identify, for each identified entity, one or more content objects associated with the entity, wherein each content object is accessible by the user based on privacy settings associated with the content object, and wherein the privacy settings define a visibility of the content object as being visible to at least the user;   generate, for each identified entity, a plurality of candidate keywords associated with the entity by extracting one or more n-grams from each identified content object associated with the entity; and   send, to the client system, instructions for presenting one or more suggested queries associated with the entity, each suggested query comprising one or more of the candidate keywords associated with the entity.