Method and system for identifying keywords for use in placing keyword-targeted advertisements

• US 9,489,449 B1
• Filed: 05/20/2010
• Issued: 11/08/2016
• Est. Priority Date: 08/09/2004
• Status: Active Grant

• Current Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Original Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Inventors: Scholl, Nathaniel B., DeNeui, Alexander W.
• Primary Examiner(s): Trujillo, James
• Assistant Examiner(s): Khoshnoodi, Fariborz
• Application Number:
  US12/783,819
• Time in Patent Office: 2,364 Days
• Field of Search: 707/769, 707/736, 707/750, 707/730, 707/706, 707/741, 707/713
• US Class Current: 1/1
• CPC Class Codes: G06F 16/313 Selection or weighting of terms for indexing
• CPC Class Codes: G06F 16/3325 Reformulation based on results of preceding query
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G06Q 30/0256 User search

Abstract

A method and system for identifying search terms for placing advertisements along with search results is provided. The advertisement system selects a description of an item that is to be advertised. The advertisement system then retrieves documents that match the selected description. The advertisement system generates a score for each word of the retrieved documents that indicates relatedness of the word to the item to be advertised. After generating the scores for the words, the advertisement system identifies phrases of the words within the documents that are related to the item. The advertisement system then generates search terms for the item to be advertised from the identified phrases. The advertisement system submits the search terms and an advertisement to a search engines service for placement of a paid-for advertisement for the item.

Claims

1. A computer-implemented method in a computer system for identifying phrases related to an item from a set of one or more documents, the method comprising:

under control of one or more computer systems configured with executable instructions,

receiving a description of the item;

performing a search across a corpus of documents to retrieve a search result set of a plurality of documents, wherein each document in the plurality of documents in the search result set is related to the item based on the description, and wherein each document in the plurality of documents in the search result set comprises a respective plurality of words;

determining a first frequency of each word in the plurality of documents of the search result set, wherein the first frequency corresponds to an average number of times each word appears in each document of the search result set;

determining a second frequency of each word in the plurality of documents of the search result set, wherein the second frequency corresponds to an average number of times each word appears in each document of the corpus of documents;

for each word of the plurality of words, determining a frequency score that is based on a difference between the corresponding first frequency and the corresponding second frequency, wherein the frequency score indicates a level of relevance of the word to the item;

identifying a set of highly related words to the item based on a threshold number of the plurality of words with the highest frequency scores;

selecting at least one anchor word from the set of highly related words; and

identifying at least one phrase including the at least one anchor word by searching each of the plurality of documents in the search result set for the at least one anchor word and identifying at least one word proximal to the at least one anchor word within the plurality of documents.

2. The computer-implemented method of claim 1, wherein identifying the at least one phrase comprises:

identifying a first phrase that contains two or more words;

extending the first phrase by one or more words proximal to the first phrase to yield an extended first phrase;

determining that the extended first phrase has a higher relevant score with respect to the item than the first phrase; and

identifying the extended first phrase as the at least one phrase.

3. The computer-implemented method of claim 2, further comprising:

identifying, based at least in part on relevant scores for the one or more words proximal to the first phrase, the one or more words proximal to the first phrase.

4. The computer-implemented method of claim 1, further comprising:

excluding one or more words from the at least one phrase that have a frequency in the corpus of documents exceeding a frequency threshold.

5. The computer-implemented method of claim 1, further comprising:

excluding duplicate documents from the search result set.

6. The computer-implemented method of claim 1, wherein the set of highly related words is a first set of highly related words to the item and is based on a first threshold number of the plurality of words with the highest frequency scores, wherein the method further comprises:

identifying a second subset of related words to the item based on a second threshold number of words with the highest frequency scores, wherein the second threshold number of words is larger than the first threshold number of words, and

wherein identifying at least one phrase including the at least one anchor word further comprises;

searching each of the plurality of documents in the search result set for the at least one anchor word; and

identifying at least one word from the second subset of related words that is proximal to the at least one anchor word within the plurality of documents.

7. The computer-implemented method of claim 1, wherein the threshold number of the plurality of words with the highest frequency scores includes a predetermined percentage of the plurality of words.

18. The computer-implemented method of claim 1, further comprising:

submitting information about the item to a search engine; and

receiving the search result set from the search engine.

19. The computer-implemented method of claim 1, further comprising:

generating one or more search terms for the item based on the at least one phrase; and

submitting a request to place an advertisement for the item to a search engine service, the request including the one or more search terms.

8. A non-transitory computer-readable storage medium having stored thereon instructions identifying phrases related to an item from a set of one or more information sources, the instructions upon being executed by one or more processors causing one or more computer systems to:

receive a description of the item;

perform a search across a corpus of documents to retrieve a search result set of a plurality of documents, wherein each document in the plurality of documents is related to the item based on the description, and wherein each document in the plurality of documents in the search result set comprises a respective plurality of words;

determine a first frequency of each word in the plurality of documents of the search result set, wherein the first frequency corresponds to an average number of times the word appears in each document of the search result set;

determine a second frequency of each word in the plurality of documents in the search result set, wherein the second frequency corresponds to an average number of times the word appears in each document of the corpus of documents;

for each word of the plurality of words, determine a frequency score that is based on a difference between the corresponding first frequency and the corresponding second frequency, wherein the frequency score indicates a level of relevance of the word to the item;

identify a predetermined number of highly related words to the item based on the predetermined number of words with the highest frequency scores;

select at least one anchor word from the predetermined number of highly related words; and

identify at least one phrase including the at least one anchor word by searching each of the plurality of documents in the search result set for the at least one anchor word and identifying at least one word proximal to the at least one anchor word within the plurality of documents.

9. The non-transitory computer-readable storage medium of claim 8, wherein the instructions that, upon being executed, cause the one or more computer systems to identify the at least one phrase comprises causing the one or more computer systems to:

identify a first phrase that contains two or more words;

identify, based at least in part on relevant scores for one or more words proximal to the first phrase, the one or more words proximal to the first phrase;

extend the first phrase by the one or more words proximal to the first phrase to yield an extended first phrase;

determine that the extended first phrase has a higher relevant score with respect to the item than the first phrase; and

identify the extended first phrase as the at least one phrase.

10. The computer-readable storage medium of claim 9, wherein the instructions upon being executed further cause the one or more computer systems to:

exclude one or more words synonymous to the at least one word in the first phrase.

11. The computer-readable storage medium of claim 8, wherein the instructions upon being executed further cause the one or more computer systems to:

exclude noise words from the at least one phrase.

12. The computer-readable storage medium of claim 8, wherein the instructions upon being executed further cause the one or more computer systems to:

exclude duplicate documents from the search result set.

20. The non-transitory computer-readable storage medium of claim 8, wherein the instructions upon being executed further cause the one or more computer systems to:

generate one or more search terms for the item based on the at least one phrase; and

submit a request to place an advertisement for the item to a search engine service, the request including the one or more search terms.

13. A computer system for identifying phrases related to an item from a set of one or more documents, comprising:

at least one processor;

memory including instructions that, upon being executed by the at least one processor, cause the computer system to;

receive a description of the item;

perform a search across a corpus of documents and retrieve a search result set of a plurality of documents, wherein each document in the plurality of documents is related to the item based on the description, and wherein each document in the plurality of documents in the search result set comprises a respective plurality of words;

determine a first frequency of each word in the plurality of documents in the search result set, wherein the first frequency corresponds to an average number of times the word appears in each document of the search result set;

determine a second frequency of each word in the plurality of documents in the search result set, wherein the second frequency corresponds to an average number of times the word appears in each document of the corpus of documents;

determine a frequency score that is based on a difference between the corresponding first frequency and the corresponding second frequency of each word, wherein the frequency score indicates a level of relevance of the word to the item;

identify a predetermined number of highly related words to the item based on the predetermined number of words with the highest frequency scores;

select at least one anchor word from the predetermined number of highly related words; and

identify at least one phrase including the at least one anchor word by searching each of the plurality of documents in the search result set for the at least one anchor word and identifying at least one word proximal to the at least one anchor word within the plurality of documents.

14. The computer system of claim 13, wherein the instructions that, upon being executed, cause the computer system to identify the phrase includes causing the computer system to:

identify a first phrase that contains two or more words;

extend the first phrase by one or more words proximal to the first phrase to yield an extended first phrase;

determine that the extended first phrase has a higher relevant score with respect to the item than the first phrase; and

identify the extended first phrase as the at least one phrase.

15. The computer system of claim 14, wherein the instructions upon being executed further cause the computer system to:

identify, based at least in part on relevant scores for the one or more words proximal to the first phrase, the one or more words proximal to the first phrase.

16. The computer system of claim 13, wherein the instructions upon being executed further cause the computer system to:

exclude noise words from the at least one phrase.

17. The computer system of claim 13, wherein the instructions upon being executed further cause the computer system to:

exclude duplicate documents from the search result set.

21. The computer system of claim 13, wherein the instructions upon being executed further cause the computer system to:

generate one or more search terms for the item based on the at least one phrase; and

submit a request to place an advertisement for the item to a search engine service, the request including the one or more search terms.