Sentiment classification using out of domain data

• US 8,942,470 B2
• Filed: 12/06/2013
• Issued: 01/27/2015
• Est. Priority Date: 12/16/2008
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Ni, Xiaochuan, Chen, Zheng, Sun, Jian-Tao, Wang, Gang
• Primary Examiner(s): Alavi, Amir
• Application Number:
  US14/099,566
• Publication Number:
  US 20140101081A1
• Time in Patent Office: 417 Days
• Field of Search: 382/100, 382/159, 382/190, 382218-220, 382/224, 382/225, 704/5, 704/9, 706/12, 706/54
• US Class Current: 382/159
• CPC Class Codes: G06F 16/353 into predefined classes
• CPC Class Codes: G06F 18/24 Classification techniques
• CPC Class Codes: G06F 40/30 Semantic analysis
• CPC Class Codes: G06N 20/00 Machine learning
• 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

Abstract

Providing sentiment classification of out of domain data are disclosed herein. In some aspects, a source domain having a trained classifier is matched to a target domain having a target classifier. The trained classifier may include identifiers that may be used to predict the sentiment of opinion data for the source domain. The target classifier may use the identifiers of the trained classifier to determine the sentiment of opinion data for the target domain.

Claims

1. A method, comprising:

collecting textual data of a target domain;

categorizing, by a target classifier associated with the target domain, the textual data according to sentiment using identifiers previously imported to the target classifier from a trained classifier associated with a source domain; and

sending data for display in a user interface to present a summary of results that are based at least in part on the categorizing the textual data.

2. The method of claim 1, wherein the sending the data for display in the user interface to present the summary of results further comprises presenting, in a first section of a plurality of sections of the user interface, a total number of instances of the textual data and a percentage of the total number of instances of the textual data that fall into one or more sentiment categories according to the categorizing the textual data.

3. The method of claim 2, wherein the sending the data for display in the user interface to present the summary of results further comprises presenting, in a second section of the plurality of sections of the user interface, a break out of the total number of instances of the textual data across multiple sub-domains of the target domain, the break out indicating a percentage of the total number of instances of the textual data within individual ones of the multiple sub-domains that fall into the one or more sentiment categories.

4. The method of claim 1, wherein the textual data comprises opinion data that is posted on the Internet.

5. The method of claim 4, wherein the opinion data includes at least one of customer reviews, blog entries, or forum entries.

6. The method of claim 1, further comprising, prior to the categorizing the textual data:

selecting the trained classifier to train the target classifier, the trained classifier comprising the identifiers used by the trained classifier to predict a sentiment of textual data of the source domain; and

populating the target classifier with at least one of the identifiers.

7. The method of claim 1, further comprising:

sorting the textual data of the target domain into groups based at least in part on an occurrence of a particular identifier in the textual data of the target domain, wherein the particular identifier is labeled with a known sentiment;

creating an additional identifier based at least in part on an occurrence of a term that is used a predefined number of times in one of the groups; and

assigning the known sentiment to the additional identifier.

8. The method of claim 7, further comprising assigning a first weight to the identifiers and a second weight to the additional identifier.

9. The method of claim 8, wherein the first weight is greater than the second weight.

10. The method of claim 6, wherein the selecting the trained classifier to train the target classifier includes selecting the source domain from multiple domains, and wherein the selecting the source domain is based at least in part on an identification of a common attribute between the source domain and the target domain.

11. A system comprising:

one or more processors;

memory storing computer-executable instructions that, when executed on the one or more processors, perform acts comprising;

collecting textual data of a target domain;

categorizing, by a target classifier associated with the target domain, the textual data according to sentiment using identifiers previously imported to the target classifier from a trained classifier associated with a source domain; and

sending data for display in a user interface to present a summary of results that are based at least in part on the categorizing the textual data.

12. The system of claim 11, wherein the sending the data for display in the user interface to present the summary of results further comprises presenting, in a first section of a plurality of sections of the user interface, a total number of instances of the textual data and a percentage of the total number of instances of the textual data that fall into one or more sentiment categories according to the categorizing the textual data.

13. The system of claim 12, wherein the sending the data for display in the user interface to present the summary of results further comprises presenting, in a second section of the plurality of sections of the user interface, a break out of the total number of instances of the textual data across multiple sub-domains of the target domain, the break out indicating a percentage of the total number of instances of the textual data within individual ones of the multiple sub-domains that fall into the one or more sentiment categories.

14. The system of claim 11, wherein the textual data comprises opinion data that is posted on the Internet.

15. The system of claim 11, the acts further comprising, prior to the categorizing the textual data:

matching the trained classifier to the target classifier based at least in part on identifying a common attribute between the source domain and the target domain, the trained classifier comprising the identifiers used by the trained classifier to predict a sentiment of textual data of the source domain; and

associating at least one of the identifiers with the target classifier.

16. A method, comprising:

collecting opinion data of a target domain;

categorizing, by a target classifier associated with the target domain, the opinion data according to sentiment using identifiers previously imported to the target classifier from a trained classifier associated with a source domain; and

sending data for display in a user interface to present a summary of results that are based at least in part on the categorizing the opinion data, the summary of results comprising a total number of instances of the opinion data, and a percentage of the total number of instances of the opinion data that fall into one or more sentiment categories according to the categorizing the opinion data.

17. The method of claim 16, wherein the percentage of the total number of instances of the opinion data that fall into the one or more sentiment categories are provided on the user interface as a break out across multiple sub-domains of the target domain.

18. The method of claim 16, further comprising, prior to the categorizing the opinion data:

selecting the trained classifier to train the target classifier, the trained classifier comprising the identifiers used by the trained classifier to predict a sentiment of opinion data of the source domain; and

populating the target classifier with at least one of the identifiers.

19. The method of claim 18, further comprising, prior to the categorizing the opinion data:

selecting an additional trained classifier from an additional source domain, the additional trained classifier comprising additional identifiers used by the additional trained classifier to predict a sentiment of opinion data of the additional source domain; and

populating the target classifier with at least one of the additional identifiers, wherein the additional identifiers are used for the categorizing the opinion data of the target domain.

20. The method of claim 16, further comprising:

evaluating an accuracy of the identifiers by using the identifiers to categorize the opinion data of the target domain having a known sentiment.