Identifying multimedia asset similarity using blended semantic and latent feature analysis
Methods and system for determining a similarity relationship between a plurality of digital assets and a target digital asset comprises creating a normalized semantic feature vector associated with a search query, discovering the target asset based on the normalized semantic feature vector, generating a normalized latent feature vector associated with the target asset, comparing the normalized semantic feature vector with semantic feature vectors for each of the digital assets to generate a semantic comparison value, comparing the normalized target latent feature vector with latent feature vectors for each of the digital assets to generate a latent comparison value, blending the semantic comparison vector value with the latent feature comparison vector value to create a target comparison value for each of the digital assets, and reporting the digital assets having the highest target comparison values to the user or group of users.
1. A method, comprising:
creating, using a content database, a target semantic feature vector for a target multimedia digital asset based on metadata by at least one of tokenizing one or more terms of the metadata, stemming the one or more terms of the metadata, or identifying synonyms of the one or more metadata terms;
generating, using the content database, a target latent feature vector for the target multimedia digital asset based on user-based information for the target multimedia digital asset;
performing, using the content database, a semantic feature comparison between the target multimedia digital asset and an other multimedia digital asset by comparing the target semantic feature vector for the target multimedia digital asset with an other semantic feature vector, wherein the other semantic feature vector is created at the content database based on metadata that describes the contents of the other multimedia digital asset by at least one of tokenizing one or more terms of the metadata, stemming the one or more terms of the metadata, or identifying synonyms of the one or more metadata terms;
calculating, based on the semantic feature comparison using the content database, a semantic feature similarity score;
generating a semantic feature vector score, using the content database, based on the semantic feature similarity score;
performing a latent feature comparison, using the content database, between the target multimedia digital asset and the other multimedia digital asset by comparing the target latent feature vector for the target multimedia digital asset with an other latent feature vector, wherein the other latent feature vector is generated based on user-based information for the other multimedia digital asset;
calculating, based on the latent feature comparison using the content database, a latent feature similarity score;
generating, using the content database, a latent feature vector score based on the latent feature similarity score;
generating a latent feature contribution factor by computing a scalar value based on a formula utilizing an amount of user-based information collected from the other multimedia asset, wherein the amount of user-based information comprises a number of user interactions and user activities associated with the other multimedia asset;
calculating, using the content database, a blended score for the other multimedia digital asset, wherein the blended score comprises a first product of the latent feature vector score weighted by the latent feature contribution factor, the first product added to a second product of the semantic feature vector score weighted by a scalar value reduced in proportion to the weight of the latent feature vector by the latent feature contribution factor;
in response to determining that the number of user interactions and user activities associated with the other multimedia asset has increased, increasing the latent feature contribution factor based on the increased number of user interactions and user activities such that:
(i) contribution of the latent feature vector score in the calculation of the blended score is increased proportional to the increased number of user interactions and user activities; and
(ii) contribution of the semantic feature vector score in the calculation of the blended score is decreased proportional to the increased number of user interactions and user activities; and
causing display, at a device that is remote from the content database, of an interface identifying similarity of the other multimedia digital asset relative to the target multimedia digital asset based on the blended score.
2. The method of claim 1 , further comprising:
receiving a search query that includes one or more terms associated with semantic features of a multimedia digital asset; and
determining the target multimedia digital asset based on the one or more terms in response to the search query.
3. The method of claim 1 , wherein the metadata for at least one of the target multimedia digital asset and the other multimedia digital asset is associated with discrete portions or segments within the corresponding multimedia digital asset.
4. The method of claim 1 , wherein the metadata for at least one of the target multimedia digital asset and the other multimedia digital asset includes one or more of: title, creation date, director, producer, writer, production studio, actors, characters, dialog, subject matter, genre, objects, settings, locations, themes, or legal clearance to third party copyrighted material associated with the corresponding multimedia digital asset.
5. The method of claim 1 , wherein creating the target semantic feature vector includes one or more of: lower-casing the one or more metadata terms, spell correcting the one or more metadata terms, creating a searchable index of the one or more metadata terms, or creating a searchable inverted index of the one or more metadata terms.
6. The method of claim 1 , further comprising receiving input indicating a selection of the target multimedia digital asset, wherein at least said displaying the interface is responsive to the input.
7. The method of claim 1 , wherein generating the target latent feature vector for the target multimedia digital asset is further based upon user-based information collected in association with one or more terms included in a search query responsive to which the interface is displayed.
8. The method of claim 1 , further comprising receiving a search request for a multimedia digital asset similar to the target multimedia digital asset, the display of the interface responsive to the search request.
9. The method of claim 1 , wherein the target semantic feature vector, other semantic feature vectors, target latent feature vector, and other latent feature vectors are normalized.
10. A system, comprising:
a processor coupled to a memory, wherein the circuitry uses the processor and memory to:
create, using a content database, a target semantic feature vector for a target multimedia digital asset based on metadata by at least one of tokenizing one or more terms of the metadata, stemming the one or more terms of the metadata, or identifying synonyms of the one or more metadata terms;
generate, using the content database, a target latent feature vector for the target multimedia digital asset based on user-based information for the target multimedia digital asset;
perform, using the content database, a semantic feature comparison between the target multimedia digital asset and an other multimedia digital asset by comparing the target semantic feature vector for the target multimedia digital asset with an other semantic feature vector, wherein the other semantic feature vector comprises metadata that describes the contents of the other multimedia digital asset;
calculate, based on the semantic feature comparison using the content database, a semantic feature similarity score;
generate a semantic feature vector score, using the content database, based on the semantic feature similarity score;
perform a latent feature comparison, using the content database, between the target multimedia digital asset and the other multimedia digital asset by comparing the target latent feature vector for the target multimedia digital asset with an other latent feature vector, wherein the other latent feature vector is generated based on user-based information for the other multimedia digital asset;
calculate, based on the latent feature comparison using the content database, a latent feature similarity score;
generate, using the content database, a latent feature vector score based on the latent feature similarity score;
generate a latent feature contribution factor by computing a scalar value based on a formula utilizing an amount of user-based information collected from the other multimedia asset, wherein the amount of user-based information comprises a number of user interactions and user activities associated with the other multimedia asset;
calculate, using the content database, a blended score for the other multimedia digital asset, wherein the blended score comprises a first product of the latent feature vector score weighted by the latent feature contribution factor, the first product added to a second product of the semantic feature vector score weighted by a scalar value reduced in proportion to weight of the latent feature vector by the latent feature contribution factor,
in response to determining that the number of user interactions and user activities associated with the other multimedia asset has increased, increasing the latent feature contribution factor based on the increased number of user interactions and user activities such that:
(i) contribution of the latent feature vector score in the calculation of the blended score is increased proportional to the increased number of user interactions and user activities; and
(ii) contribution of the semantic feature vector score in the calculation of the blended score is decreased proportional to the increased number of user interactions and user activities;
and cause display, at a device that is remote from the content database, of an interface identifying similarity of the other multimedia digital asset relative to the target multimedia digital asset based on the blended score.
11. The system of claim 10 , wherein the circuitry is further configured to:
receive a search query that includes one or more terms associated with semantic features of a multimedia digital asset; and
determine the target multimedia digital asset based on the one or more terms in response to the search query.
12. The system of claim 10 , wherein the metadata for at least one of the target multimedia digital asset and the other multimedia digital asset is associated with discrete portions or segments within the corresponding multimedia digital asset.
13. The system of claim 10 , wherein the metadata for at least one of the target multimedia digital asset and the other multimedia digital asset includes one or more of: title, creation date, director, producer, writer, production studio, actors, characters, dialog, subject matter, genre, objects, settings, locations, themes, or legal clearance to third party copyrighted material associated with the corresponding multimedia digital asset.
14. The system of claim 10 , wherein creating the target semantic feature vector includes one or more of: lower-casing the one or more metadata terms, spell correcting the one or more metadata terms, creating a searchable index of the one or more metadata terms, or creating a searchable inverted index of the one or more metadata terms.
15. The system of claim 10 , wherein the circuitry is further configured to receive input indicating a selection of the target multimedia digital asset, wherein at least said displaying the interface is responsive to the input.
16. The system of claim 10 , wherein the circuitry is further configured to generate the target latent feature vector for the target multimedia digital asset based on user-based information collected in association with one or more terms included in a search query responsive to which the interface is displayed.
17. The system of claim 10 , wherein the circuitry is further configured to receive a search request for a multimedia digital asset similar to the target multimedia digital asset, the display of the interface responsive to the search request.
18. The system of claim 10 , wherein the target semantic feature vector, other semantic feature vectors, target latent feature vector, and other latent feature vectors are normalized.