Recommendations based on branding

1. A method, comprising:

maintaining an index that specifies a plurality of brand relationships among a plurality of brands, wherein the maintaining includes:

accessing a corpus specifying a plurality of user queries; and

for a first brand relationship between first and second brands of the plurality of brands, determining a recommendation score for the first brand relationship based on a semantic similarity of terms used in a subset of queries, from the plurality of user queries, that are associated with the first and second brands, wherein determining the recommendation score includes identifying user queries of the plurality of user queries that contain a disjunction of brand terms associated with the first and second brands;

receiving an indication of a user activity performed on a computing device by a user;

based on the user activity, identifying a brand preference, of the user, for the first brand;

determining a recommendation to provide to the user based on the brand preference, including by:

querying the index based on the first brand; and

selecting the second brand to include in the recommendation based on the recommendation score; and

providing the recommendation to the user.

2. The method of claim 1 , wherein the user activity comprises at least one of navigating through category menus, submitting a query, submitting a bid, purchasing an item, subscribing to a status change alert items for sale.

3. The method of claim 1 , wherein brand preference is identified based on at least one of one or more brands included in a query and a brand of an item purchased or bid on by the user.

4. The method of claim 1 , further comprising:

expanding, using one or more hardware processors, a seed set of brands corresponding to a category in the index, including by:

accessing the corpus specifying the plurality of user queries;

evaluating user queries of the plurality of user queries that contain a disjunction of brand terms; and

identifying a new brand to add to the seed set of brands based on the evaluating.

5. The method of claim 4 , further comprising determining one or more measures of semantic similarity between user queries of the plurality of user queries, wherein the one or more measures of semantic similarity are based on common terms in the user queries.

6. The method of claim 1 , further comprising identifying a category from the user activity, wherein the recommendation provided to the user comprises the brand preference in a different category than the category identified from the user activity.

7. A system comprising:

a non-transitory memory; and

one or more hardware processors coupled to the non-transitory memory and configured to read instructions from the non-transitory memory to cause the system to perform operations comprising:

maintaining an index that specifies a plurality of brand relationships among a plurality of brands, wherein the maintaining includes:

accessing a corpus specifying a plurality of user queries; and

for a first brand relationship between first and second brands of the plurality of brands, determining a recommendation score for the first brand relationship based on a semantic similarity of terms used in a subset of queries, from the plurality of user queries, that are associated with the first and second brands, wherein determining the recommendation score includes identifying user queries of the plurality of user queries that contain a disjunction of brand terms associated with the first and second brands;

receiving an indication of a user activity performed on a computing device by a user;

based on the user activity, identifying a brand preference, of the user, for the first brand;

determining a recommendation to provide to the user based on the brand preference, including by:

querying the index based on the first brand; and

selecting the second brand to include in the recommendation based on the recommendation score; and

providing the recommendation to the user.

8. The system of claim 7 , wherein the user activity comprises at least one of navigating through category menus, submitting a query, submitting a bid, purchasing an item, subscribing to a status change alert items for sale.

9. The system of claim 7 , wherein brand preference is identified based on at least one of one or more brands included in a query and a brand of an item purchased or bid on by the user.

10. The system of claim 7 , wherein the operations further comprise:

expanding a seed set of brands corresponding to a category in the index, including by:

accessing the corpus specifying the plurality of user queries;

evaluating user queries of the plurality of user queries that contain a disjunction of brand terms; and

identifying a new brand to add to the seed set of brands based on the evaluating.

11. The system of claim 10 , wherein the operations further comprise determining one or more measures of semantic similarity between user queries of the plurality of user queries, wherein the one or more measures of semantic similarity are based on common terms in the user queries.

12. The system of claim 7 , wherein the operations further comprise identifying a category from the user activity, wherein the recommendation provided to the user comprises the brand preference in a different category than the category identified from the user activity.

13. A non-transitory machine-readable medium having stored thereon machine-readable instructions executable to cause a machine to perform operations comprising:

maintaining an index that specifies a plurality of brand relationships among a plurality of brands, wherein the maintaining includes:

accessing a corpus specifying a plurality of user queries; and

for a first brand relationship between first and second brands of the plurality of brands, determining a recommendation score for the first brand relationship based on a semantic similarity of terms used in a subset of queries, from the plurality of user queries, that are associated with the first and second brands, wherein determining the recommendation score includes identifying user queries of the plurality of user queries that contain a disjunction of brand terms associated with the first and second brands;

receiving an indication of a user activity performed on a computing device by a user;

based on the user activity, identifying a brand preference, of the user, for the first brand;

determining a recommendation to provide to the user based on the brand preference, including by:

querying the index based on the first brand; and

selecting the second brand to include in the recommendation based on the recommendation score; and

providing the recommendation to the user.

14. The non-transitory machine-readable medium of claim 13 , wherein the user activity comprises at least one of navigating through category menus, submitting a query, submitting a bid, purchasing an item, subscribing to a status change alert items for sale.

15. The non-transitory machine-readable medium of claim 13 , wherein brand preference is identified based on at least one of one or more brands included in a query and a brand of an item purchased or bid on by the user.

16. The non-transitory machine-readable medium of claim 13 , wherein the operations further comprise:

expanding, using one or more hardware processors, a seed set of brands corresponding to a category in the index, including by:

accessing the corpus specifying the plurality of user queries;

evaluating user queries of the plurality of user queries that contain a disjunction of brand terms; and

identifying a new brand to add to the seed set of brands based on the evaluating.

17. The non-transitory machine-readable medium of claim 16 , wherein the operations further comprise determining one or more measures of semantic similarity between user queries of the plurality of user queries, wherein the one or more measures of semantic similarity are based on common terms in the user queries.