Method for fine-grained affective states understanding and prediction

1. A computer-implemented method for fine-grained affective states prediction, the computer-implemented method comprising:

creating training data comprising content conveying emotions;

creating a trained model by performing an emotion vector space model training process using the training data to train a model using a feed forward neural network, wherein the emotion vector space model training process converts discrete emotions into emotion vector representations, wherein an output layer of the feed forward neural network are word vectors that indicate probabilities of words and emojis that are within a window size of an emotion vector representation that is input for an input layer of the feed forward neural network, and wherein a hidden layer of the feed forward neural network is represented by a weight matrix with a row for each of the discrete emotions and a column for every hidden neuron in the hidden layer; and

using the trained model to predict fine-grained affective states for text conveying an emotion.

2. The method of claim 1 , further comprising applying dimension reduction and rotation techniques on the trained model to identify a principle dimension of a set of fine-grained affective states.

3. The method of claim 1 , further comprising applying clustering techniques on the trained model to identify basic emotions of a set of fine-grained affective states.

4. The method of claim 1 , further comprising generating domain-specific emotional taxonomy by aggregating predicted results.

5. The method of claim 1 , wherein the emotion vector space model training process further comprises projecting the emotion vector representations of the fine-grained affective states into a distributed vector space, wherein similar fine-grained affective states are close to each other in the distributed vector space.

6. The method of claim 5 , further comprising gathering all sentences from the content conveying emotions that express a same emotional state together into a corresponding emotional document to form an input layer of the feed forward neural network.

7. The method of claim 6 , wherein the emotion vector space model training process further comprises converting words in the corresponding emotional document into word vectors and projecting the word vectors into the distributed vector space.

8. The method of claim 1 , wherein creating training data comprising content conveying emotions comprises gathering content that includes fine-grained emotions from online content based on context dependencies.

9. The method of claim 8 , wherein the training data comprises a random sampling of a same number of the content conveying emotions for each of the fine-grained emotions.

10. The method of claim 1 , wherein creating the training data comprising crawling for online posts containing a keyword, parsing the online posts containing the keyword to construct dependency relations between words appearing in the online posts, and extracting adjective terms that have dependencies with the keyword and treating the adjective terms as discrete emotions in the training data.

11. The method of claim 10 , further comprising deleting all discrete emotions with less than predetermined number of appearances from the training data.

12. The method of claim 10 , further comprising removing duplicate online posts containing the keyword found after crawling for the online posts containing the keyword.

13. The method of claim 10 , further comprising randomly sampling up to a predetermined number of the online posts for each of the discrete emotions.

14. A system configured to predict fine-grained affective states, the system comprising a processor configured to execute instructions to:

create training data comprising content conveying emotions;

create a trained model by performing an emotion vector space model training process using the training data to train a model using a feed forward neural network, wherein the emotion vector space model training process converts the discrete emotions into emotion vector representations, wherein an output layer of the feed forward neural network are word vectors that indicate probabilities of words and emojis that are within a window size of an emotion vector representation that is input for an input layer of the feed forward neural network, and wherein a hidden layer of the feed forward neural network is represented by a weight matrix with a row for each of the discrete emotions and a column for every hidden neuron in the hidden layer; and

use the trained model to predict fine-grained affective states for text conveying an emotion.

15. The system of claim 14 , further comprising applying dimension reduction and rotation techniques on the trained model to identify a principle dimension of a set of fine-grained affective states.

16. The system of claim 14 , further comprising applying clustering techniques on the trained model to identify basic emotions of a set of fine-grained affective states.

17. The system of claim 14 , further comprising generating domain-specific emotional taxonomy by aggregating predicted results.

18. The system of claim 14 , wherein the emotion vector space model training process further comprises projecting the emotion vector representations of the fine-grained affective states into a distributed vector space, wherein similar fine-grained affective states are close to each other in the distributed vector space.

19. The system of claim 18 , further comprising gathering all sentences from the content conveying emotions that express a same emotional state together into a corresponding emotional document to form an input layer of the feed forward neural network.

20. The system of claim 19 , wherein the emotion vector space model training process further comprises converting words in the corresponding emotional document into word vectors and projecting the word vectors into the distributed vector space.