Venue link detection for social media messages

A computer system for inferring linkage between social media messages and venues, comprising:

• one or more processors;
• memory;
• and
• one or more programs stored in the memory configured for execution by the one or more processors, the one or more programs comprising instructions for:
• accessing a social network graph, stored in a database on the computer system, comprising user nodes representing social media users, message nodes representing social media messages generated by the social media users, and venue nodes, wherein the venue nodes are distinct and separate from the nodes representing social media users and the nodes representing social media messages, wherein the venue nodes include:
• a plurality of nodes representing a plurality of physical venues;
• and
• a no-venue node that does not represent a physical venue;
• wherein a link in the social network graph between a respective message node and the no-venue node indicates that the social media message does not correspond to a physical venue;
• constructing a plurality of training feature vectors, wherein each training feature vector includes a respective plurality of features that use paths through the social network graph to measure connectedness between a respective social media message and a respective venue;
• using the training feature vectors to train a classifier to estimate probabilities that social media messages are associated with venues;
• receiving a new social media message from a user;
• constructing a feature vector for the new social media message, wherein each feature vector includes a plurality of features that use paths through the social network graph to measure connectedness between the new social media message and the no-venue node;
• executing the trained classifier using the feature vector as input to compute a probability that the new social media message is associated with the no-venue node;
• when the computed probability is greater than a predefined threshold value, determining that the new social media message does not correspond to a physical venue of the plurality of physical venues;
• and
• when the computed probability is less than or equal to the predefined threshold value, determining that the new social media message corresponds to a respective physical venue of the plurality of physical venues.

A non-transitory computer readable storage medium storing one or more programs configured for execution by a computer system having one or more processors and memory, the one or more programs comprising instructions for:

• accessing a social network graph, stored in a database on the computer system, comprising user nodes representing social media users, message nodes representing social media messages generated by the social media users, and venue nodes, wherein the venue nodes are distinct and separate from the nodes representing social media users and the nodes representing social media messages, wherein the venue nodes include:
• a plurality of nodes representing a plurality of physical venues;
• and
• a no-venue node that does not represent a physical venue;
• wherein a link in the social network graph between a respective message node and the no-venue node indicates that the social media message does not correspond to a physical venue;
• constructing a plurality of training feature vectors, wherein each training feature vector includes a respective plurality of features that use paths through the social network graph to measure connectedness between a respective social media message and a respective venue;
• using the training feature vectors to train a classifier to estimate probabilities that social media messages are associated with venues;
• receiving a new social media message from a user;
• constructing a feature vector for the new social media message, wherein each feature vector includes a plurality of features that use paths through the social network graph to measure connectedness between the new social media message and the no-venue node;
• executing the trained classifier using the feature vector as input to compute a probability that the new social media message is associated with the no-venue node;
• when the computed probability is greater than a predefined threshold value, determining that the new social media message does not correspond to a physical venue of the plurality of physical venues;
• and
• when the computed probability is less than or equal to the predefined threshold value, determining that the new social media message corresponds to a respective physical venue of the plurality of physical venues.