Generating training data for a computer-implemented ranker

Patent metadata

Field	Value
Publication number	US-11645692-B2
Application number	US-201916690072-A
Country	US
Kind code	B2
Filing date	Nov 20, 2019
Priority date	Nov 20, 2019
Publication date	May 9, 2023
Grant date	May 9, 2023

Abstract

Technologies relating to generation of training data for use in training a computer-implemented ranker are described. A product name extracted from a URL in a log entry of a search log of a search engine is mapped to a catalog product name from a catalog of a website, wherein the catalog product name identifies a product that is available for acquisition by way of a webpage of a website. A query assigned to the URL in the log entry is assigned to the catalog product name due to the product name being mapped to the catalog product name. The computer-implemented ranker is trained based upon the query being assigned to the catalog product name.

First claim

What is claimed is:

1. A computing system that is configured to train a computer-implemented ranker, wherein the computing system comprises:
a processor; and
memory storing instructions that, when executed by the processor, cause the processor to perform acts comprising:
identifying a log entry in a search engine log of a search engine, wherein the log entry comprises:
a URL of a first webpage that belongs to a first website; and
a query submitted to the search engine, wherein the search engine returned the first webpage in search results upon receipt of the query;

extracting, from the URL of the first webpage, a first product name, wherein the first product name identifies a first product that is offered for acquisition by way of the first webpage;
subsequent to extracting the first product name from the URL of the first webpage, assigning the query in the log entry to a second product name, wherein the second product name identifies a second product that is offered for acquisition by way of a second webpage that belongs to a second website, wherein the first website is different from the second website; and
training the computer-implemented ranker based upon pairs of product names and queries, wherein the pairs of product names and queries include the second product name and the query, and further wherein the computer-implemented ranker is configured to rank product names based upon queries received by the second website, wherein the product names identify products that are offered for acquisition by way of the second website.

2. The computing system of claim 1 , wherein identifying the log entry in the search engine log of the search engine comprises:
searching the search engine log for log entries that include URLs of webpages that belong to the first website, wherein the log entry is identified due to the URL of the first webpage including a domain name of the first website.

3. The computing system of claim 2 , wherein extracting the first product name from the URL of the first webpage comprises:
identifying the first product name in the URL of the first webpage based upon a template that defines a structure of URLs that include the domain name.

4. The computing system of claim 1 , the acts further comprising:
identifying a second log entry in the search engine log of the search engine, wherein the second log entry comprises:
a second URL of a third webpage that belongs to a third website, wherein the third website is different from the first website and the second website; and
a second query submitted to the search engine, wherein the search engine returned the third webpage in second search results upon receipt of the second query;

extracting, from the second URL of the third webpage, a third product name, wherein the third product name identifies a third product that is offered for acquisition by way of the third webpage;
subsequent to extracting the third product name from the second URL of the third webpage, assigning the second query in the second log entry to the second product name, wherein the pairs of product names and queries include the second product name and the second query.

5. The computing system of claim 1 , the acts further comprising:
prior to assigning the query in the log entry to the second product name, encoding the first product name into a first vector of a predefined length;
encoding the second product name into a second vector of the predefined length; and
computing a distance between the first vector and the second vector, wherein the query in the log entry is assigned to the second product name based upon the distance between the first vector and the second vector.

6. The computing system of claim 5 , the acts further comprising:
comparing the distance between the first vector and the second vector to a predefined threshold, wherein the query in the log entry is assigned to the second product name when the distance between the first vector and the second vector is beneath the threshold.

7. The computing system of claim 1 , the acts further comprising:
subsequent to identifying the log entry in the search engine log of the search engine, retrieving the first webpage through use of the URL of the first webpage in the search engine log;
extracting metadata from the first webpage, wherein the metadata comprises a value of an attribute of the first product identified by the first product name; and
assigning the metadata to the second product name in a searchable index for the second website, such that when a second query that includes the value of the attribute is received by way of the second website, the second product name is included in search results that are returned based upon the second query.

8. The computing system of claim 1 , the acts further comprising:
receiving the query from a server computing system that hosts the second website, wherein the server computing system received the query from a client computing device that had loaded a third webpage that belongs to the second website;
generating search results based upon the query, wherein the search results include the second webpage;
ranking, by the computer-implemented ranker, the search results to generate a ranked list of search results, wherein the second webpage is positioned in the ranked list of search results based upon the query being assigned to the second product name.

9. The computing system of claim 1 , wherein the computer-implemented ranker is customized for the second website.

10. The computing system of claim 9 , wherein the computer-implemented ranker is a re-ranker that is configured to re-rank ranked lists of search results output by a general-purpose ranker, and further wherein the general-purpose ranker is configured to rank search results for multiple different websites.

11. A method for training a computer-implemented ranker, the method comprising:
identifying a log entry in a search engine log of a search engine, wherein the log entry comprises:
a URL of a first webpage that belongs to a first website; and
a query submitted to the search engine, wherein the search engine returned the first webpage in search results upon receipt of the query;

extracting, from the URL of the first webpage, a first product name, wherein the first product name identifies a first product that is offered for acquisition by way of the first webpage;
subsequent to extracting the first product name from the URL of the first webpage, assigning the query in the log entry to a second product name, wherein the second product name identifies a second product that is offered for acquisition by way of a second webpage that belongs to a second website, wherein the first website is different from the second website; and
training the computer-implemented ranker based upon pairs of product names and queries, wherein the pairs of product names and queries include the second product name and the query, and further wherein the computer-implemented ranker is configured to rank product names based upon queries received by the second website, wherein the product names identify products that are offered for acquisition by way of the second website.

12. The method of claim 11 , wherein identifying the log entry in the search engine log of the search engine comprises:
searching the search engine log for log entries that include URLs of webpages that belong to the first website, wherein the log entry is identified due to the URL of the first webpage including a domain name of the first website.

13. The method of claim 12 , wherein extracting the first product name from the URL of the first webpage comprises:
identifying the first product

Assignees

• Microsoft Technology Licensing Llc

Inventors

• Xia Xiaozhen
• Pan Jun
• Lu Yumao
• Sureshchandra Bhimani Kalpesh

Classifications

• G06Q30/0629 Primary

  by pre-processing results, e.g. ranking or ordering results · CPC title

• G06F16/9538

  Presentation of query results · CPC title

• G06F16/951 Primary

  Indexing; Web crawling techniques · CPC title

• G06F16/9532

  Query formulation · CPC title