Extraction and summarization of sentiment information

• US 7,792,841 B2
• Filed: 05/30/2006
• Issued: 09/07/2010
• Est. Priority Date: 05/30/2006
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Hanks, Steven J., McAllister, Ian A., Bao, Ling, Ponath, Christoph R.
• Primary Examiner(s): Vo; Tim T.
• Assistant Examiner(s): Tran; Anh Tai V
• Application Number:
  US11/442,909
• Publication Number:
  US 20070282867A1
• Time in Patent Office: 1,561 Days
• Field of Search: None
• US Class Current: 707/749
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising

Abstract

Methods and systems for extraction and summarization of sentiment information related to a particular research subject are disclosed. A method includes accessing sources of information that contain sentiment information that is related to the research subject and extracting the sentiment information from the sources of information as opinions related to the research subject. Opinion categories related to features of the research subject are identified. From this information a summarization of the sentiment information that is related to the particular research subject that includes the identified opinion categories is generated. Subsequently, access is provided to the summarization for graphical presentation.

Claims

1. A computer-implemented method for extraction and summarization of sentiment information contained in sources of information that is related to a particular research subject, comprising:

using a processor to access said sources of information that contain said sentiment information that is related to a plurality of research subject, said sentiment information comprises textually recorded information that characterize a user'"'"'s opinion related to said research subject;

simultaneously extracting said sentiment information from said sources of information as opinions related to the plurality of research subjects said research subject based on model parameters and text extracting rules;

identifying opinion categories related to features of said particular research subject based upon a determined magnitude of the number of opinions obtained that are related to said opinion categories;

generating a summarization of said sentiment information that comprises said opinion categories;

providing access to said summarization of said sentiment information for graphical presentation;

updating said model parameters and said text extracting rules based on an offline training process and feedback from users, comprising;

identifying sentimental information within tagged training sets and untagged training sets;

identifying polarity characterizations of the sentimental information within the tagged training set;

updating opinion categories based on the identified polarity characterizations and the feedback from users;

updating said model parameters and said text extracting rules based on the identified polarity characterization; and

accessing said updated model parameters and text extracting rules for use in extraction and summarization of said sentiment information.

2. The method of claim 1 wherein said offline training process comprises manual tagging, feature clustering analysis, human evaluation and adjustment and algorithm evaluation.

3. The method of claim 1 wherein said sources of information comprise customer reviews, forums, discussion groups, and blogs.

4. The method of claim 1 wherein said opinions are capable of being accessed and reviewed individually.

5. The method of claim 4 wherein said opinions that are capable of being accessed and reviewed individually are graphically presented.

18. The method of claim 1, further comprising determining whether said sources of information contain information related to said research subject.

6. A computer-readable storage medium having computer-executable instructions for performing extraction and summarization of sentiment information contained in sources of information that is related to a particular research subject comprising:

using a processor to access said sources of information that contain said sentiment information that is related to a plurality of research subject, said sentiment information comprises textually recorded information that characterize a user'"'"'s opinion related to said research subject;

simultaneously extracting said sentiment information from said sources of information as opinions related to the plurality of research subjects said research subject based on model parameters and text extracting rules;

identifying opinion categories related to features of said particular research subject based upon a determined magnitude of the number of opinions obtained that are related to said opinion categories;

generating a summarization of said sentiment information that comprises said opinion categories;

providing access to said summarization of said sentiment information for graphical presentation;

updating said model parameters and said text extracting rules based on an offline training process and feedback from users, comprising;

identifying sentimental information within tagged training sets and untagged training sets;

identifying polarity characterizations of the sentimental information within the tagged training set;

updating opinion categories based on the identified polarity characterizations and the feedback from users;

updating said model parameters and said text extracting rules based on the identified polarity characterization; and

accessing said updated model parameters and text extracting rules for use in extraction and summarization of said sentiment information.

7. The storage medium of claim 6 wherein said offline training process comprises manual tagging, feature clustering analysis, human evaluation and adjustment and algorithm evaluation.

8. The storage medium of claim 6 wherein said documents are found at various internet addresses.

9. The storage medium of claim 6 wherein said documents include customer reviews, forums, discussion groups, and blogs.

10. The storage medium of claim 6 wherein said features comprise topics related to said subject matter that are discussed most often.

11. The storage medium of claim 6 wherein said opinions are capable of being accessed and reviewed individually.

12. An apparatus comprising:

a processor for executing a method for extraction and summarization of sentiment information contained in sources of information that is related to a particular research subject, comprising;

using a processor to access said sources of information that contain said sentiment information that is related to a plurality of research subject, said sentiment information comprises textually recorded information that characterize a user'"'"'s opinion related to said research subject;

simultaneously extracting said sentiment information from said sources of information as opinions related to the plurality of research subjects said research subject based on model parameters and text extracting rules;

identifying opinion categories related to features of said particular research subject based upon a determined magnitude of the number of opinions obtained that are related to said opinion categories;

generating a summarization of said sentiment information that comprises said opinion categories;

providing access to said summarization of said sentiment information for graphical presentation;

updating said model parameters and said text extracting rules based on an offline training process and feedback from users, comprising;

identifying sentimental information within tagged training sets and untagged training sets;

identifying polarity characterizations of the sentimental information within the tagged training set;

updating opinion categories based on the identified polarity characterizations and the feedback from users;

updating said model parameters and said text extracting rules based on the identified polarity characterization; and

accessing said updated model parameters and text extracting rules for use in extraction and summarization of said sentiment information.

13. The apparatus of claim 12 wherein said offline training process comprises manual tagging, feature clustering analysis, human evaluation and adjustment and algorithm evaluation.

14. The apparatus of claim 12 wherein said documents are found at various internet addresses.

15. The apparatus of claim 12 wherein said documents include customer reviews, forums, discussion groups, and blogs.

16. The apparatus of claim 12 wherein said research subject is a product or a service.

17. The apparatus of claim 12 wherein said opinions are capable of being accessed and reviewed individually.