Personalized recommendation of a volatile item

Patent metadata

Field	Value
Publication number	US-2016239551-A1
Application number	US-201615140878-A
Country	US
Kind code	A1
Filing date	Apr 28, 2016
Priority date	Jan 19, 2010
Publication date	Aug 18, 2016
Grant date	—

Abstract

A system, method, and machine readable medium for creating a personalized recommendation of an item by creating a topic vector based on a plurality of search queries, at least one of a plurality of users associated with the search queries or a plurality of items associated with the search queries; interring a topical preference for a user based on a search query by the user; and recommending at least one item based on the topical preference and the topic vector.

First claim

1 . (canceled)

2 . A method of creating a personalized recommendation of a product, the method comprising: using one or more processors:
creating a first plurality of probability distributions based upon a dataset that includes a plurality of products and a list of search queries input into a search engine of an electronic marketplace that led to those products over a predetermined period of time, each of the first plurality of probability distributions corresponding to a respective one of a plurality of latent topics and describing, for each respective search query in the list of search queries, a probability that the respective latent topic is related to the respective search query; creating a user-based probability distribution for a user based upon the first plurality of probability distributions and a second list of search queries previously entered by the user, the user-based probability distribution describing a probability that the user is interested in each of the plurality of latent topics; selecting a latent topic from the user-based probability distribution that has a probability above a threshold probability; selecting a search query from one of the first plurality of probability distributions that has a probability above a threshold probability of relating to the selected latent topic; submitting the search query to the search engine of the electronic marketplace; and recommending to the user, at least one search result received from the search engine of the electronic marketplace in response to the submitted search query, the at least one search result describing a volatile product that is part of a product inventory on a temporary basis.

3 . The method of claim 2 , wherein the user is one of a plurality of users that submitted search queries in the dataset.

4 . The method of claim 2 , wherein creating the first plurality of probability distributions comprises using at least one of: a variational expectation-maximization algorithm, an expectation propagation algorithm, or Gibbs sampling.

5 . The method of claim 2 , wherein recommending at least one search result comprises recommending a top search result.

6 . The method of claim 2 , further comprising retrieving the list of search queries from a database of an online commerce site.

7 . The method of claim 2 , wherein the user-based probability distribution is a multinomial probability distribution.

8 . The method of claim 2 , wherein the search query is a search query that the user has not previously issued.

9 . A system for creating a personalized recommendation of a product, the system comprising:
one or more processors; a memory including instructions, which when executed by the one or more processors, cause the one or more processors to perform operations of:
creating a first plurality of probability distributions based upon a dataset that includes a plurality of products and a list of search queries input into a search engine of an electronic marketplace that led to those products over a predetermined period of time, each of the first plurality of probability distributions corresponding to a respective one of a plurality of latent topics and describing, for each respective search query in the list of search queries, a probability that the respective latent topic is related to the respective search query;
creating a user-based probability distribution for a user based upon the first plurality of probability distributions and a second list of search queries previously entered by the user, the user-based probability distribution describing a probability that the user is interested in each of the plurality of latent topics;
selecting a latent topic from the user-based probability distribution that has a probability that is above a threshold probability;
selecting a search query from one of the first plurality of probability distributions that has a probability above a threshold probability of relating to the selected latent topic;
submitting the search query to the search engine of the electronic marketplace; and
recommending to the user, at least one search result received from the search engine of the electronic marketplace in response to the submitted search query, the at least one search result describing a volatile product that is part of a product inventory on a temporary basis.

10 . The system of claim 9 , wherein the user is one of a plurality of users that submitted search queries in the dataset.

11 . The system of claim 9 , wherein the operations of creating the first plurality of probability distributions comprises the operations of using at least one of: a variational expectation-maximization algorithm, an expectation propagation algorithm, or Gibbs sampling.

12 . The system of claim 9 , wherein the operations of recommending at least one search result comprises the operations of recommending a top search result.

13 . The system of claim 9 , wherein the operations further comprise retrieving the list of search queries from a database of an online commerce site.

14 . The system of claim 9 , wherein the user-based probability distribution is a multinomial probability distribution.

15 . The system of claim 9 , wherein the high-ranking search query is a search query that the user has not previously issued.

16 . A non-transitory machine readable medium, including instructions, which when executed by the machine, cause the machine to perform operations of:
creating a first plurality of probability distributions based upon a dataset that includes a plurality of products and a list of search queries input into a search engine of an electronic marketplace that led to those products over a predetermined period of time, each of the first plurality of probability distributions corresponding to a respective one of a plurality of latent topics and describing, for each respective search query in the list of search queries, a probability that the respective latent topic is related to the respective search query; creating a user-based probability distribution for a user based upon the first plurality of probability distributions and a second list of search queries previously entered by the user, the user-based probability distribution describing a probability that the user is interested in each of the plurality of latent topics; selecting a latent topic from the user-based probability distribution that has a probability above a threshold probability; selecting a search query from one of the first plurality of probability distributions that has a probability above a threshold probability of relating to the selected latent topic; submitting the search query to the search engine of the electronic marketplace; and recommending to the user, at least one search result received from the search engine of the electronic marketplace in response to the submitted search query, the at least one search result describing a volatile product that is part of a product inventory on a temporary basis.

17 . The machine-readable medium of claim 16 , wherein the user is one of a plurality of users that submitted search queries in the dataset.

18 . The machine-readable medium of claim 16 , wherein the operations of creating the first plurality of probability distributions comprises the operations of using at least one of: a variational expectation-maximization algorithm, an expectation propagation algorithm, or Gibbs sampling.

19 . The machine-readable medium of claim 16 , wherein the operations of recommending at least one search result comprises the operations of

Assignees

• Ebay Inc

Inventors

• Ovsjanikovs Maksims
• Chen Ye

Classifications

• G06F17/30539 Primary

  Physics · mapped topic

• G06N5/04

  Inference or reasoning models · CPC title

• G06F17/30572

  Physics · mapped topic

• G06F16/26

  Visual data mining; Browsing structured data · CPC title

• G06F16/951

  Indexing; Web crawling techniques · CPC title