Ranking products by mining comparison sentiment

• US 8,731,995 B2
• Filed: 05/12/2008
• Issued: 05/20/2014
• Est. Priority Date: 05/12/2008
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Sun, Jian-Tao, Hu, Jian, Ponath, Christoph, Chen, Zheng, Bao, Ling, Wang, Qionghua, Wu, Jianmin
• Primary Examiner(s): Pats, Justin M
• Assistant Examiner(s): Goldberg, Ivan R
• Application Number:
  US12/118,764
• Publication Number:
  US 20090281870A1
• Time in Patent Office: 2,199 Days
• Field of Search: 705/7, 705/10, 705/7.32, 705/7.29, 705/345, 704/9
• US Class Current: 705/7.32
• CPC Class Codes: G06F 40/263 Language identification
• CPC Class Codes: G06F 40/289 Phrasal analysis, e.g. finite state techniques or chunking
• CPC Class Codes: G06F 40/30 Semantic analysis
• CPC Class Codes: G06Q 10/10 Office automation; Time management
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G06Q 30/0201 Market modelling; Market analysis; Collecting market data
• CPC Class Codes: G06Q 30/0203 Market surveys; Market polls
• CPC Class Codes: G06Q 30/0282 Rating or review of business operators or products
• CPC Class Codes: G06Q 99/00 Subject matter not provided for in other groups of this subclass

Abstract

A method of ranking a plurality of products includes obtaining a numerical user score for each of the plurality of products, calculating an opinion score for each of the plurality of products for which a written comparison sentiment applies, determining a final score by combining the numerical user rating and the opinion score for each of the plurality of products, and ranking the plurality of products based on the final score. The opinion score is derived from one or more written comparison sentiments containing a first product name and a second product name. In another aspect, a computer readable medium has instructions stored thereon which, when executed, cause a computer to perform a method of ranking a plurality of products including at least a first product and a second product.

Claims

1. A computer-implemented method of ranking a plurality of products comprising at least a first product and a second product, the method comprising:

obtaining a numerical user score for each of the plurality of products;

obtaining written user reviews pertaining to one or more of the products;

identifying at least one superlative statement from the written user reviews that is determined to include, in terms of product names, only an individual product name and that does not indicate a relative preference between products;

identifying written comparison sentiments from the written user reviews by excluding the identified at least one superlative statement from the written user reviews, each written comparison sentiment containing a direct product comparison between first and second products, the direct product comparison including a first product name and a second product name and describing a subjective user preference of the first product relative to the second product;

calculating an opinion score for each of the plurality of products, based at least in part on the written comparison sentiments;

determining a final score using a computer by combining the numerical user score and the opinion score for each of the plurality of products, wherein a tradeoff parameter is utilized to give a greater weight to the opinion score than the numerical user score in determining the final score; and

ranking the plurality of products based on the final score.

2. The method of claim 1, wherein calculating an opinion score comprises classifying the comparison sentiment into one of three categories, indicating either that the first product is preferred over the second product, the second product is preferred over the first product, or that there is no preference sentiment between the first product and the second product.

3. The method of claim 2, wherein classifying the comparison sentiment comprises determining relative position information comprising whether a selected word of the written comparison sentiment occupies one of three positions:

either between the first product name and the second product name, before both the first product name and the second product name, or after both the first product name and the second product name.

4. The method of claim 2, wherein classifying the comparison sentiment comprises determining description relationship information comprising whether a selected word of the written comparison sentiment applies to only the first product, only the second product, both the first product and the second product, or neither the first product nor the second product.

5. The method of claim 2, wherein classifying the comparison sentiment comprises:

determining relative position information comprising whether a selected word of the written comparison sentiment occupies one of three positions;

either between the first product name and the second product name, before both the first product name and the second product name, or after both the first product name and the second product name; and

determining description relationship information comprising whether a selected word of the written comparison sentiment applies to only the first product, only the second product, both the first product and the second product, or neither the first product nor the second product.

6. The method of claim 1, wherein obtaining a numerical user score comprises using a Bayesian weighted averaging method.

7. The method of claim 1, and further comprising:

setting the trade-off parameter based on a distance between the opinion scores and the numerical user scores of the plurality of products.

8. A computer-implemented method of ranking a plurality of products comprising at least a first product and a second product, the method comprising:

obtaining a numerical user score for each of the plurality of products;

obtaining written user reviews pertaining to one or more of the products;

identifying at least one superlative statement in the written user reviews that is determined to include, in terms of product names, only an individual product name and does not indicate a relative preference between products;

identifying one or more written comparison sentiments from the written user reviews by excluding the identified at least one superlative statement from the written user reviews, each written comparison sentiment containing a direct product comparison between the first product and the second product, the direct product comparison including a first product name and a second product name and describing a subjective user preference of the first product relative to the second product;

calculating an opinion score for each of the plurality of products, wherein the opinion scores for the first and second products are calculated by;

for each of the one or more written comparison sentiments,

selecting a keyword within the written comparison sentiment;

determining relative position information for the keyword relative to the first and second product names; and

determining description relationship information of the keyword relative to the first and second product names; and

using both the relative position information and the description relationship information to calculate the opinion score;

determining a final score for each product of the plurality of products using a computer by combining the numerical user score and the opinion score for the product; and

ranking the plurality of products based on the final score determined for each of the plurality of products.

9. The method of claim 8, wherein a tradeoff parameter is utilized to weight the numerical user score and the opinion score differently in determining the final score.

10. The method of claim 9, wherein the tradeoff parameter gives greater weight to the opinion score than to the numerical user score.

11. The method of claim 8, and further comprising setting a trade-off parameter based on a distance between the opinion scores and the numerical user scores of the plurality of products and determining the final score by weighting the numerical user score and the opinion score using the trade-off parameter for each of the plurality of products.

12. The method of claim 8, wherein determining relative position information comprises:

determining whether the selected keyword occupies one of three positions;

either between the first product name and the second product name, before both the first product name and the second product name, or after both the first product name and the second product name.

13. The method of claim 8, wherein determining description relationship information comprises:

determining whether the selected keyword of the written comparison sentiment applies to only the first product, only the second product, both the first product and the second product, or neither the first product nor the second product.

14. A hardware computer readable storage medium having instructions stored thereon which, when executed, cause a computer to perform a method of ranking a plurality of products, the method comprising:

obtaining a numerical user score for each product of the plurality of products;

calculating an opinion score for each product of the plurality of products from one or more written comparison sentiments containing a direct product comparison between a first product and a second product, the direct product comparison including a first product name and a second product name and indicating a subjective user preference of the first product relative to the second product,

for each product of the plurality of products, determining a numerical distance between the opinion score and the numerical user score for the product;

determining an overall numerical distance by summing the numerical distances determined for each of the products;

setting a trade-off parameter based on the overall numerical distance;

determining a final score for each product of the plurality of products using the computer by combining the numerical user score and the opinion score for the product,

wherein the numerical user score and the opinion score are weighted using the trade-off parameter; and

ranking the plurality of products based on the final score determined for each product of the plurality of products.

15. The hardware computer readable medium of claim 14, wherein determining whether the first product is preferred over the second product comprises:

selecting a keyword in the written comparison sentiment;

determining relative position information comprising whether the selected word of the written comparison sentiment occupies one of three positions;

either between the first product name and the second product name, before both the first product name and the second product name, or after both the first product name and the second product name; and

determining whether the selected keyword applies to only the first product, only the second product, both the first product and the second product, or neither the first product nor the second product.

16. The hardware computer readable medium of claim 14, wherein obtaining a numerical user score comprises using a Bayesian weighted averaging method.

17. The hardware computer readable medium of claim 14, and further comprising:

obtaining written user reviews pertaining to at least one of the first and second products;

identifying at least one superlative statement from the written user reviews that is determined to include, in terms of product names, only an individual product name and the superlative statement does not indicate a relative preference between a plurality of products; and

identifying the one or more written comparison sentiments from the written user reviews by excluding the identified at least one superlative statement from the written user reviews, each written comparison sentiment containing a direct product comparison between the first product and the second product, the direct product comparison including a first product name and a second product name and describing a subjective user preference of the first product relative to the second product.

18. The hardware computer readable medium of claim 14, wherein the tradeoff parameter is utilized to weight the numerical user score and the opinion score differently in determining the final score.

19. The hardware computer readable medium of claim 18, wherein the tradeoff parameter gives greater weight to the opinion score than to the numerical user score.