BIAS QUOTIENT MEASUREMENT AND DEBIASING FOR RECOMMENDATION ENGINES
Systems and methods for debiasing a recommendation engine are disclosed herein. A search query associated with a user profile is received at a recommendation engine. Control circuitry generates a result set of items of content based on the search query and generates a bias score for a content attribute based on the result set. The control circuitry also generates a time-averaged bias score for the content attribute based on a plurality of search queries associated with the user profile. Based on the bias score and the time-averaged bias score, the control circuitry determines whether a bias is signaled for the content attribute. Finally, the control circuitry outputs, for display via a computing device, the result set or a debiased result set based on a result of the determination of whether the bias is signaled.
1 - 50 . (canceled)
51 . A method comprising:
storing, in a memory, a persona rule indicating a type of user profile persona and a content attribute applicable to the type of user profile persona;
accessing a search query and search query results associated with a user profile;
determining whether the user profile corresponds to the stored type of user profile persona;
determining that a bias is signaled for a first content attribute of the search query results, wherein the signaled bias is correlated with the user profile; and
in response to determining that the signaled bias is correlated with the user profile:
adding to the search query results a content item that is applicable to the user profile to generate debiased results, wherein the content item is uncorrelated, orthogonal or complementary to the signaled bias; and
outputting, for display via the computing device, the debiased results.
52 . The method of claim 51 further comprising:
storing, in a memory, an attribute rule indicating uncorrelated, orthogonal or complementary relationships; and
searching a database, based on the attribute rule, to identify content that is uncorrelated, orthogonal or complementary to the signaled bias.
53 . The method of claim 52 further comprising determining, based on the persona rule, that the content orthogonal or complementary to the signaled bias is applicable to the user profile.
54 . The method of claim 53 wherein determining, based on the persona rule, that the content orthogonal or complementary to the signaled bias is applicable to the user profile is based on a second attribute.
55 . The method of claim 54 wherein determining, based on the persona rule, that the content orthogonal or complementary to the signaled bias is applicable to the user profile is based on the amount of content in a content source that share the first attribute and second attribute.
56 . The method of claim 51 , wherein determining that a bias is signaled for the first content attribute comprises computing a quotient based on a number of content items in the search query results having the first content attribute and a total number of content items in the search query results.
57 . The method of claim 56 , wherein determining whether the bias is signaled for the first content attribute comprises determining whether the quotient exceeds a threshold quotient stored in memory.
58 . The method of claim 51 wherein the content attribute is one of genre, language, character, and release date.
59 . The method of claim 51 further comprising retrieving from the user profile the persona type.
60 . A system comprising:
memory;
control circuitry configured to:
store, in the memory, a persona rule indicating a type of user profile persona and a content attribute applicable to the type of user profile persona;
access a search query and resulting search query results of the search query associated with a user profile;
determine whether the user profile corresponds to the stored type of user profile persona;
determine that a bias is signaled for a first content attribute of the search query results wherein the signaled bias is correlated with the user profile; and
in response to determining that the signaled bias is correlated with the user profile:
add to the search query results a content item that is applicable to the user profile to generate debiased results, wherein the content item is uncorrelated, orthogonal or complementary to the signaled bias; and
output, for display via the computing device, the debiased results.
61 . The system of claim 60 , the control circuitry further configured to:
store, in the memory, an attribute rule indicating uncorrelated, orthogonal or complementary relationships; and
search a database, based on the attribute rule, to identify content that is uncorrelated, orthogonal or complementary to the signaled bias.
62 . The system of claim 61 the control circuitry further configured to determine, based on the persona rule, that the content orthogonal or complementary to the signaled bias is applicable to the user profile.
63 . The system of claim 62 wherein determining, based on the persona rule, that the content orthogonal or complementary to the signaled bias is applicable to the user profile is based on a second attribute.
64 . The system of claim 63 wherein determining, based on the persona rule, that the content orthogonal or complementary to the signaled bias is applicable to the user profile is based on the amount of content in a content source that share the first attribute and second attribute.
65 . The system of claim 60 , wherein the control circuitry further configured to determine that a bias is signaled for the first content attribute by computing a quotient based on a number of content items in the search query results having the first content attribute and a total number of content items in the search query results.
66 . The system of claim 65 , wherein the control circuitry is further configured to determine whether the bias is signaled for the first content attribute by determining whether the quotient exceeds a threshold quotient stored in memory.
67 . The system of claim 60 wherein the content attribute is one of genre, language, character, and release date.
68 . The system of claim 60 the control circuitry further configured to retrieve from the user profile the persona type.
69 . A non-transitory, computer-readable memory having instructions stored thereon that, when executed by a processor, cause the performance of a set of actions comprising:
storing, in a memory, a persona rule indicating a type of user profile persona and a content attribute applicable to the type of user profile persona;
accessing a search query and resulting search query results of the search query associated with a user profile;
determining whether the user profile corresponds to the stored type of user profile persona;
determining that a bias is signaled for a first content attribute of the search query results wherein the signaled bias is correlated with the user profile; and
in response to determining that the signaled bias is correlated with the user profile:
adding to the search query results a content item that applicable to the user profile to generate debiased results, wherein the content item is uncorrelated, orthogonal or complementary to the signaled bias; and
outputting, for display via the computing device, the debiased results.
70 . The non-transitory, computer-readable memory of claim 69 , further causing the actions of:
storing, in a memory, an attribute rule indicating uncorrelated, orthogonal or complementary relationships; and
searching a database, based on the attribute rule, to identify content that is uncorrelated, orthogonal or complementary to the signaled bias.