Sentiment extraction from consumer reviews for providing product recommendations

Patent metadata

Field	Value
Publication number	US-9646078-B2
Application number	US-11946508-A
Country	US
Kind code	B2
Filing date	May 12, 2008
Priority date	May 12, 2008
Publication date	May 9, 2017
Grant date	May 9, 2017

Abstract

A system and method for recommending a product to a user in response to a query for a product with a feature wherein the recommendation is accompanied by a quotation expressing a sentiment about the feature or the product.

First claim

We claim:

1. A method for extracting quotations related to a product from a document, comprising:
building a span query relevant to a feature associated with the product and a sentiment about the feature by:
building syntactic templates from a lexicography relevant to the feature and the sentiment about the feature, wherein each syntactic template corresponds to multiple phrasings of the same meaning, and
determining a first semantic template relevant to the syntactic templates;

receiving a document;
applying the span query to the document to generate a span query result, wherein the span query result includes a quotation from the document relevant to the feature and the sentiment about the feature, the quotation satisfying the first semantic template and including words within relative positions as constrained by the span query; and
storing the span query result.

2. The method of claim 1 further comprising applying a heuristic model to the document.

3. The method of claim 1 wherein determining a first semantic template comprises resolving a conflict between a second and a third semantic template.

4. The method of claim 3 wherein resolving the conflict comprises applying defeasible logic programming.

5. The method of claim 1 wherein the document is a web page.

6. A non-transitory computer-readable medium having computer-readable program instructions stored therein, the computer-readable program instructions configured to, when executed by a computer, cause the computer to:
build a span query relevant to a feature and a sentiment about the feature by:
building syntactic templates from a lexicography relevant to the feature and the sentiment about the feature, wherein each syntactic template corresponds to multiple phrasings of the same meaning; and
determining a first semantic template relevant to the syntactic templates;

receive a document;
apply the span query to the document to generate a span query result, wherein the span query result includes a quotation from the document relevant to the feature and the sentiment about the feature, the quotation satisfying the first semantic template and including words within relative positions as constrained by the span query; and
store the span query result.

7. The non-transitory computer-readable medium of claim 6 wherein the computer-readable program instructions are further configured to, when executed by the computer, cause the computer to apply a heuristic model to the document.

8. The non-transitory computer-readable medium of claim 6 wherein determining a first semantic template comprises resolving a conflict between a second and a third semantic template.

9. The non-transitory computer-readable medium of claim 8 wherein resolving the conflict comprises applying defeasible logic programming.

10. The method of claim 1 wherein determining a first semantic template relevant to the syntactic template comprises:
building at least one atomic semantic template from the syntactic template; and
building a complex semantic template from the at least one atomic semantic template wherein the complex semantic template comprises at least one prohibitive clause, the at least one prohibitive clause comprising an expression relevant to the feature that does not convey sentiment.

11. The method of claim 10 wherein:
building at least one atomic semantic template comprises building a first and a second atomic semantic template; and
building the complex semantic template comprises resolving a conflict between the first and the second atomic semantic templates.

12. The non-transitory computer-readable medium of claim 6 wherein determining a first semantic template relevant to the syntactic template comprises:
building at least one atomic semantic template from the syntactic template; and
building a complex semantic template from the at least one atomic semantic template wherein the complex semantic template comprises at least one prohibitive clause, the at least one prohibitive clause comprising an expression relevant to the feature that does not convey sentiment.

13. The non-transitory computer-readable medium of claim 12 wherein:
building at least one atomic semantic template comprises building a first and a second atomic semantic template; and
building the complex semantic template comprises resolving a conflict between the first and the second atomic semantic templates.

14. The method of claim 1 wherein the feature is an abstract characteristic.

15. The non-transitory computer-readable medium of claim 6 wherein the feature is an abstract characteristic.

16. The method of claim 1 , further comprising:
building one or more additional span queries;
applying the one or more additional span queries to the document; and
storing additional span query results from applying the one or more additional span queries.

17. The non-transitory computer-readable medium of claim 6 , wherein the computer-readable program instructions are further configured to, when executed by the computer, cause the computer to:
build one or more additional span queries;
apply the one or more additional span queries to the document; and
store additional span query results from applying the one or more additional span queries.

18. An apparatus, comprising:
a processor configured to:
build a span query relevant to a feature associated with a product and a sentiment about the feature by:
building syntactic templates from a lexicography relevant to the feature and the sentiment about the feature, wherein each syntactic template corresponds to multiple phrasings of the same meaning, and
determining a first semantic template relevant to the syntactic templates;

receiving a document;
apply the span query to the document to generate a span query result, wherein the span query result includes a quotation from the document relevant to the feature and the sentiment about the feature, the quotation satisfying the first semantic template and including words within relative positions as constrained by the span query.

19. The apparatus of claim 18 , wherein the processor configured to determine the first semantic template comprises the processor being configured to resolve a conflict between a second and a third semantic template by applying defeasible logic programming.

20. The apparatus of claim 18 , wherein the processor configured to determine a first semantic template relevant to the syntactic template includes the processor being configured to:
build at least one atomic semantic template from the syntactic template; and
build a complex semantic template from the at least one atomic semantic template.

21. The apparatus of claim 18 , wherein the complex semantic template comprises at least one prohibitive clause, the at least one prohibitive clause comprising an expression relevant to the feature that does not convey sentiment.

22. The apparatus of claim 21 , wherein:
the processor configured to build the at least one atomic semantic template includes the processor being configured to build a first and a second atomic semantic template; and
the processor configured to build the complex semantic template includes the processor being configured to resolve a conflict between the first and the second atomic semantic templates.

Assignees

• Galitsky Boris
• Mckenna Eugene William
• Groupon Inc

Inventors

• Galitsky Boris
• Mckenna Eugene William

Classifications

• G06F16/33

  Querying · CPC title

• G06Q30/0631 Primary

  Recommending goods or services · CPC title

• G06F40/30

  Semantic analysis · CPC title

• G06F16/334

  Query execution (filtering based on additional data G06F16/335) · CPC title

• G06F40/279

  Recognition of textual entities · CPC title