Method for learning a latent interest taxonomy from multimedia metadata
Techniques are disclosed herein for learning latent interests based on metadata of one or more images. An analysis tool associates one or more attributes with each of the objects based on a time and a location described in the metadata of that object. Each of the attributes describes one of a plurality of locations or an event scheduled to occur at one or more of the plurality of locations. The analysis tool identifies one or more concepts from a distribution of the one or more attributes to each of the objects. Each of the one or more concepts includes at least a first attribute in the distribution that co-occurs with a second attribute in the distribution.
1. A method for identifying latent relationships between interests based on metadata of a plurality of digital multimedia objects, the method comprising:
evaluating metadata associated with each digital multimedia object against a knowledge graph, wherein the knowledge graph is built from data comprising information external to each of the digital multimedia objects and wherein the knowledge graph provides a plurality of attributes, each attribute describing information relating to an event scheduled at one of a plurality of locations;
associating, based on the evaluation, one or more of the plurality of attributes with each of the digital multimedia objects, the one or more attributes correlating with a time and a location described in the metadata of that object;
identifying one or more concepts from a distribution of the associated attributes to each of the objects; and
mapping each of the plurality of attributes to the one or more concepts based on the distribution.
2. The method of claim 1 , wherein mapping each of the plurality of attributes to the one or more concepts comprises:
determining a membership score of each of the one or more attributes to each of the one or more concepts, wherein the membership score is a measure indicating a strength of correlation of a given attribute to a given concept.
3. The method of claim 2 , wherein each of the attributes is associated with at least one of the one or more concepts based on the membership score.
4. The method of claim 1 , further comprising:
determining, from the one or more concepts, a hierarchical relationship between a first concept and at least a second concept.
5. The method of claim 1 , wherein the one or more concepts are further identified based on latent Dirichlet allocation.
6. The method of claim 1 , wherein the attributes are imputed to each of the one or more digital multimedia objects from the knowledge graph, wherein the attributes further describe the plurality of locations.
7. The method of claim 1 , wherein each of the digital multimedia objects is one of an image or a video.
8. A non-transitory computer-readable storage medium storing instructions, which, when executed on a processor, performs an operation for identifying latent relationships between interests based on metadata of a plurality of digital multimedia objects, the operation comprising:
evaluating metadata associated with each digital multimedia object against a knowledge graph, wherein the knowledge graph is built from data comprising information external to each of the digital multimedia objects and wherein the knowledge graph provides a plurality of attributes, each attribute describing information relating to an event scheduled at one of a plurality of locations;
associating, based on the evaluation, one or more of the plurality of attributes with each of the digital multimedia objects, the one or more attributes correlating with a time and a location described in the metadata of that object;
identifying one or more concepts from a distribution of the associated attributes to each of the objects; and
mapping each of the plurality of attributes to the one or more concepts based on the distribution.
9. The computer-readable storage medium of claim 8 , wherein mapping each of the plurality of attributes to the one or more concepts comprises:
determining a membership score of each of the one or more attributes to each of the one or more concepts, wherein the membership score is a measure indicating a strength of correlation of a given attribute to a given concept.
10. The computer-readable storage medium of claim 9 , wherein each of the attributes is associated with at least one of the one or more concepts based on the membership score.
11. The computer-readable storage medium of claim 8 , wherein the operation further comprises:
determining, from the one or more concepts, a hierarchical relationship between a first concept and at least a second concept.
12. The computer-readable storage medium of claim 8 , wherein the one or more concepts are further identified based on latent Dirichlet allocation.
13. The computer-readable storage medium of claim 8 , wherein the attributes are imputed to each of the one or more digital multimedia objects from the knowledge graph, wherein the attributes further describe the plurality of locations.
14. The computer-readable storage medium of claim 8 , wherein each of the digital multimedia objects is one of an image or a video.
15. A system, comprising:
a processor; and
a memory storing one or more application programs configured to perform an operation for identifying latent relationships between interests based on metadata of a plurality of digital multimedia objects, the operation comprising:
evaluating metadata associated with each digital multimedia object against a knowledge graph, wherein the knowledge graph is built from data comprising information external to each of the digital multimedia objects and wherein the knowledge graph provides a plurality of attributes, each attribute describing information relating to an event scheduled at one of a plurality of locations,
associating, based on the evaluation, one or more of the plurality of attributes with each of the digital multimedia objects, the one or more attributes correlating with a time and a location described in the metadata of that object,
identifying one or more concepts from a distribution of the associated attributes to each of the objects, and
mapping each of the plurality of attributes to the one or more concepts based on the distribution.
16. The system of claim 15 , wherein mapping each of the plurality of attributes to the one or more concepts comprises:
determining a membership score of each of the one or more attributes to each of the one or more concepts, wherein the membership score is a measure indicating a strength of correlation of a given attribute to a given concept.
17. The system of claim 16 , wherein each of the attributes is associated with at least one of the one or more concepts based on the membership score.
18. The system of claim 15 , wherein the operation further comprises:
determining, from the one or more concepts, a hierarchical relationship between a first concept and at least a second concept.
19. The system of claim 15 , wherein the one or more concepts are further identified based on latent Dirichlet allocation.
20. The system of claim 15 , wherein each of the digital multimedia objects is one of an image or a video and wherein the attributes are imputed to each of the one or more digital multimedia objects from the knowledge graph, wherein the attributes further describe the plurality of locations.