Consumption history privacy
An audience selection system for the selection of an entity, based on an entity's consumption history without requiring the storage of a content descriptor for identifying content previously accessed by the entity. By directly and/or indirectly observing the usage of words used to locate content through a search engine over time for a population, a list of depersonalized keywords can be discovered, creating the ability to characterize content based on depersonalized keywords. A protected consumption history can be recorded for an entity using depersonalized keywords instead of recording a content descriptor for identifying the content. Depersonalized keywords do not uniquely identify content. Associating depersonalized keywords with an entity does not mean that the entity has used those depersonalized keywords; it only means that the entity has accessed content which has been accessed in the past by other entities in a population using the depersonalized keywords.
1. A computer-implemented method comprising:
characterizing multiple items of content accessed over a network, each item of content associated with a content identifier and characterized with a set of multiple depersonalized keywords, the set comprising words submitted to an online search engine by a population comprising multiple entities in the past to locate the content;
receiving, at a keyword mapping system, a consumption history off an individual entity, the consumption history comprising at least two different content identifiers, each content identifier associated with a respective item of content accessed over the network by the individual entity;
recording a protected consumption history of the individual entity in a data store at the keyword mapping system, each content identifier of the consumption history represented in the protected consumption history by the respective characteristic set of multiple depersonalized keywords associated with the content identifier's item of content;
discarding the content identifiers of the consumption history;
receiving a list of specified keywords;
determining a frequency of each specified keyword in the depersonalized keywords of the individual entity's protected consumption history; and
assessing the suitability of the individual entity for receiving selected content based on the frequencies.
2. The method of claim 1 wherein the step of characterizing comprises weighting the depersonalized keywords based on recency.
3. The method of claim 1 wherein the step of characterizing comprises weighting the depersonalized keywords based on frequency.
4. The method of claim 1 wherein the step of characterizing comprises weighting the depersonalized keywords based on relevance.
5. The method of claim 1 wherein assessing the suitability of the individual entity for receiving selected content comprises assessing the suitability of the individual entity for receiving an online advertisement.
6. The method of claim 1 wherein assessing the suitability of the individual entity for receiving selected content comprises assessing the suitability of the individual entity for receiving customized content over the network.
7. The method of claim 1 wherein:
content comprises aggregate content and aggregate content is characterized by aggregating depersonalized keywords associated with multiple items of content.
8. A non-transitory computer-readable storage medium storing executable computer program instructions for audience selection, the computer program instructions comprising instructions for:
characterizing multiple items of content accessed over a network, each item of content associated with a content identifier and characterized with a set of multiple depersonalized keywords, the set comprising multiple words submitted to an online search engine by a population comprising multiple entities in the past to locate the content;
receiving, at a keyword mapping system, a consumption history of an individual entity the consumption history comprising at least two different content identifiers, each content identifier associated with a respective item of content accessed over the network by the individual entity;
recording a protected consumption history of the individual entity in a data store at the keyword mapping system, each content identifier of the consumption history represented in the protected consumption history by the respective characteristic set of multiple depersonalized keywords associated with the content identifier's item of content;
discarding the content identifiers of the consumption history;
receiving a list of specified keywords;
determining a frequency of each specified keyword in the depersonalized keywords of the individual entity's protected consumption history; and
assessing the suitability of the individual entity for receiving selected content based on the frequencies.
9. The non-transitory computer-readable storage medium of claim 8 wherein the instructions further comprise:
receiving a protected behavioral model and assessing the suitability of the entity by applying the protected behavioral model to the entity's protected consumption history.
10. The non-transitory computer-readable storage medium of claim 8 wherein the protected consumption history excludes the content descriptor for identifying the content.
11. A system comprising:
a processor;
a computer readable storage medium storing processor-executable computer program instructions for:
characterizing multiple items of content accessed over a network, each item of content associated with a content identifier and characterized with a set of multiple depersonalized keywords, the set comprising words submitted to an online search engine by a population comprising multiple entities in the past to locate the content;
receiving, at a keyword mapping system, a consumption history of an individual entity, the consumption history comprising at least two different content identifiers, each content identifier associated with a respective item of content accessed over the network by the individual entity;
recording a protected consumption history of the individual entity in a data store at the keyword mapping system, each content identifier of the consumption history represented in the protected consumption history by the respective characteristic set of multiple depersonalized keywords associated with the content identifier's item of content;
discarding the content identifiers of the consumption history;
receiving a list of specified keywords;
determining a frequency of each specified keyword in the depersonalized keywords of the individual entity's protected consumption history; and
assessing the suitability of the individual entity for receiving selected content based on the frequencies.
12. A system comprising:
a processor;
a computer readable storage medium storing processor-executable computer program instructions for:
characterizing multiple items of content accessed over a network, each item of content associated with a content identifier and characterized with a set of multiple depersonalized keywords, the set comprising words submitted to an online search engine by a population comprising multiple entities in the past to locate the content;
receiving consumption histories of multiple entities, at a keyword mapping system, each respective consumption history comprising at least two different content identifiers, each content identifier associated with a respective item of content accessed over the network by an individual entity;
recording a protected consumption history of each of the multiple entities in a data store at the keyword mapping system, each content identifier in the consumption histories represented in the protected consumption histories by the characteristic set of multiple depersonalized keywords associated with the content identifier's item of content;
discarding the content identifiers of the consumption histories;
receiving a list of specified keywords;
determining a frequency of each specified keyword in the depersonalized keywords of the individual entity's protected consumption history;
identifying a training set comprising entities chosen from the storage based on the frequencies;
building a behavioral model based on the training set wherein features of the behavioral model comprise one or more depersonalized keywords;
receiving a specified entity's protected consumption history; and
assessing the suitability of the specific entity for selection by applying the behavioral model to the specified entity's protected consumption history.
13. A computer-implemented method comprising:
accessing a respective protected consumption history of each of a plurality of entities in a storage wherein each protected consumption history comprises a history of the access of multiple items of content over a network by an entity with each item of content characterized by a set of multiple depersonalized keywords, each set comprising multiple words submitted to an online search engine in the past by a population comprising multiple entities to locate each respective item of content, wherein each protected consumption history is purged of content identifiers;
identifying a training set comprising entities chosen from the storage by examining the protected consumption histories for one or more depersonalized keywords;
building a behavioral model based on the training set wherein features of the behavioral model comprise depersonalized keywords;
accessing a specified entity's protected consumption history; and
assessing the suitability of the specific entity for selection by applying the behavioral model to the specified entity's protected consumption history.
14. A computer-implemented method comprising:
accessing a respective consumption history of each of a plurality of entities in a storage wherein each consumption history comprises a history of consumption events comprising the access of content over a network by an entity, with the content associated with a content identifier;
selecting a training set comprising entities chosen from the storage, with the consumption history of each entity in the training set including a specified consumption event;
building a behavioral model based on the training set, wherein features of the behavioral model comprise content identifiers;
converting the behavioral model to a protected behavioral model by replacing the content identifiers with depersonalized keywords in the features of the behavioral model;
accessing a specified entity's protected consumption history, the protected consumption history comprising a history of accessing of multiple items of content over a network by the specified entity with each item of content characterized by a set of multiple depersonalized keywords, each set of multiple depersonalized keywords comprising multiple words submitted to an online search engine in the past by a population comprising multiple entities to locate each respective item of content; and
assessing the suitability of the specific entity for receiving selected content by applying the protected behavioral model to the specified entity's protected consumption history.