Context-based sentiment analysis

1. A system for context-based sentiment analysis, comprising:

one or more processors configured to:

obtain a piece of comment data that is initially determined to be associated with a positive sentiment, a negative sentiment, a neutral sentiment, or a mixed sentiment, the piece of comment data being included in a comment;

determine a context sentiment type of an associated context related to the piece of comment data, the associated context being distinct from the comment and including text that is created at a different time than the piece of comment data, and the determination being made with reference to at least a set of context features, wherein the context sentiment type indicates that the associated context is positive sentiment-entailing, negative sentiment-entailing, or neutral, wherein the determining of the context sentiment type of the associated context related to the piece of comment data comprises to:

determine whether the context sentiment type of the associated context related to the piece of comment data is positive sentiment-entailing; and

in response to a determination that the context sentiment type is determined to be positive sentiment-entailing, classify the piece of comment data includes using a positive model-based comment classifier, comprising to:

perform an initial classification; and

in response to a determination that the initial classification is positive and a positive context model-based comment classification of the positive model-based comment classifier is not-positive, determine a final classification to be neutral; and

classify the piece of comment data, the classification being based at least in part on the piece of comment data and the context sentiment type; and

one or more memories coupled to the one or more processors, configured to provide the one or more processors with instructions.

2. The system of claim 1 , wherein the associated context is supplied by a first user, and the comment is supplied by a second user distinct from the first user.

3. The system of claim 1 , wherein the associated context includes a question or topic, and the comment is made in response to the associated context.

4. The system of claim 1 , wherein the set of context features includes a keyword or key phrase.

5. The system of claim 1 , wherein the positive model-based comment classifier is trained using training data comprising sample comment data known to be associated with positive sentiment-entailing contexts.

6. The system of claim 1 , wherein to classify the piece of comment data includes to use a negative model-based comment classifier in the event that the context sentiment type is determined to be negative sentiment-entailing.

7. The system of claim 6 , wherein the negative model-based comment classifier is trained using training data comprising sample comment data known to be associated with negative sentiment-entailing contexts.

8. A method for context-based sentiment analysis, comprising:

obtaining a piece of comment data that is initially determined to be associated with a positive sentiment, a negative sentiment, a neutral sentiment, or a mixed sentiment, the piece of comment data being included in a comment;

determining a context sentiment type of an associated context related to the piece of comment data, the associated context being distinct from the comment and including text that is created at a different time than the piece of comment data, and the determination being made with reference to at least a set of context features, wherein the context sentiment type indicates that the associated context is positive sentiment-entailing, negative sentiment-entailing, or neutral, wherein the determining of the context sentiment type of the associated context related to the piece of comment data comprises:

determining whether the context sentiment type of the associated context related to the piece of comment data is positive sentiment-entailing; and

in response to a determination that the context sentiment type is determined to be positive sentiment-entailing, classifying the piece of comment data includes using a positive model-based comment classifier, comprising to:

performing an initial classification; and

in response to a determination that the initial classification is positive and a positive context model-based comment classification of the positive model-based comment classifier is not-positive, determining a final classification to be neutral; and

classifying the piece of comment data, the classification being based at least in part on the piece of comment data and the context sentiment type.

9. The method of claim 8 , wherein the associated context is supplied by a first user, and the comment is supplied by a second user distinct from the first user.

10. The method of claim 8 , wherein the associated context includes a question or topic, and the comment is made in response to the associated context.

11. The method of claim 8 , wherein the set of context features includes a keyword or key phrase.

12. The method of claim 8 , further comprising training a positive model-based classifier using training data comprising sample comment data known to be associated with positive sentiment-entailing contexts.

13. The method of claim 8 , further comprising performing a negative model-based comment classification in the event that the context sentiment type is determined to be negative sentiment-entailing.

14. The method of claim 13 , further comprising training a negative model-based classifier using training data comprising sample comment data known to be associated with negative sentiment-entailing contexts.

15. A computer program product for context-based sentiment analysis, the computer program product being embodied in a tangible computer readable storage medium and comprising computer instructions for:

obtaining a piece of comment data that is initially determined to be associated with a positive sentiment, a negative sentiment, a neutral sentiment, or a mixed sentiment, the piece of comment data being included in a comment;

determining a context sentiment type of an associated context related to the piece of comment data, the associated context being distinct from the comment and including text that is created at a different time than the piece of comment data, and the determination being made with reference to at least a set of context features, wherein the context sentiment type indicates that the associated context is positive sentiment-entailing, negative sentiment-entailing, or neutral, wherein the determining of the context sentiment type of the associated context related to the piece of comment data comprises:

determining whether the context sentiment type of the associated context related to the piece of comment data is positive sentiment-entailing; and

in response to a determination that the context sentiment type is determined to be positive sentiment-entailing classifying the piece of comment data includes using a positive model-based comment classifier, comprising to:

performing an initial classification; and

in response to a determination that the initial classification is positive and a positive context model-based comment classification of the positive model-based comment classifier is not-positive, determining a final classification to be neutral; and

classifying the piece of comment data, the classification being based at least in part on the piece of comment data and the context sentiment type.

16. The system of claim 1 , wherein to determine a context sentiment type includes using a learning machine that is trained using sample context known to be positive sentiment-entailing, negative sentiment-entailing, or neutral.

17. The system of claim 16 , wherein the learning machine includes a decision tree, a neural network, a Bayesian classifier, or a support vector machine.

18. The method of claim 8 , wherein determining a context sentiment type includes using a learning machine that is trained using sample context known to be positive sentiment-entailing, negative sentiment-entailing, or neutral.

19. The method of claim 18 , wherein the learning machine includes a decision tree, a neural network, a Bayesian classifier, or a support vector machine.