Artificial intelligence (AI) enabled prescriptive persuasion processes based on speech emotion recognition and sentiment analysis
Methods and systems are provided for the AI-based computer-aided persuasion system (CAPS). The CAPS obtains inputs from both the target and the agent for an object, dynamically generates persuasion references based on analysis of the input. The CAPS obtains content output by analyzing the agent audio stream and the target audio stream using a recurrent network (RNN) model, obtains sentiment classifiers based on a convolutional neural network (CNN LSTM) model, updates a conversation matrix, and generates a persuasion reference based on the updated conversation matrix. The persuasion reference is based on an acceptance likelihood result generated from the conversation matrix using the RNN model. The CAPS further generates a target profile using DNN (deep neural net) with input of target Big Data, wherein the target profile includes one or more objects, and wherein the agent is selected based on the generated profile and one or more selected objects.
1. A method, comprising:
obtaining, by a computer system with one or more processors coupled with at least one memory unit, a target audio stream from a target when an agent is engaging in a conversation with the target on a target topic;
obtaining a target content output by analyzing the target audio stream using a recurrent network (RNN) model;
detecting one or more off-topic items in the target content output with corresponding sentiment classifiers;
updating a conversation matrix that contains prior and current target audio stream analysis based on prior and current target content output and sentiment classifiers for the target content output;
generating prescriptive analytics result for the one or more off-topic items based on the updated conversation matrix.
2. The method of claim 1 , wherein the prescriptive analytics result includes the one or more off-topic items with corresponding likelihood of success.
3. The method of claim 2 , wherein each likelihood of success of corresponding off-topic item generated from the conversation matrix using the RNN model.
4. The method of claim 3 , further comprising: generating a persuasion reference based on the prescriptive analytics result.
5. The method of claim 4 , wherein the persuasion reference includes an exploitation guidance focused on the target topic and an exploration guidance on the one or more off-topic items.
6. The method of claim 5 , wherein the exploration guidance is allocated a percentage of time based on the prescriptive analytics result.
7. The method of claim 1 , wherein the conversation matrix further contains prior and current agent audio stream analysis based on prior and current agent content output generated using the RNN model and sentiment classifiers for the agent content output.
8. The method of claim 1 , wherein each sentiment classifier is derived from an emotion classifier resulting from a convolutional neural network (CNN LSTM) model analysis of a corresponding audio stream.
9. The method of claim 8 , wherein the emotion classifier is one selecting from an emotion group comprising angry emotion, excited emotion, frustrated emotion, happy emotion, neutral emotion, sad emotion, and surprised emotion, and the sentiment classifier is one selecting from a sentiment group comprising extremely positive, positive, neutral, negative, extremely negative, and surprised.
10. The method of claim 9 , wherein each sentiment classifier prescribes a corresponding strategy delivered by the agent through follow up conversations that updates the conversation matrix with new audio streams from the follow up conversations.
11. A system comprising:
an audio input module that obtains a target audio stream from a target when an agent is engaging in a conversation with the target on a target topic;
a content output module that an agent and target content output by analyzing the audio stream using a speech to text module and then a recurrent network (RNN) model to learn the textual content;
an off-topic module that detects one or more off-topic items in the target content output with corresponding sentiment classifiers;
a conversation handling module that updates a conversation matrix that contains prior and current target audio stream analysis based on prior and current target content output and sentiment classifiers for the target content output;
a prescriptive analytic module that generates prescriptive analytics result for the one or more off-topic items based on the updated conversation matrix.
12. The system of claim 11 , wherein the prescriptive analytics result includes the one or more off-topic items with corresponding likelihood of success.
13. The system of claim 12 , wherein each likelihood of success of corresponding off-topic item generated from the conversation matrix using the RNN model.
14. The system of claim 13 , further comprising: generating a persuasion reference based on the prescriptive analytics result.
15. The system of claim 14 , wherein the persuasion reference includes an exploitation guidance focused on the target topic and an exploration guidance on the one or more off-topic items.
16. The system of claim 15 , wherein the exploration guidance is allocated a percentage of time based on the prescriptive analytics result.
17. The system of claim 11 , wherein the conversation matrix further contains prior and current agent audio stream analysis based on prior and current agent content output generated using the RNN model and sentiment classifiers for the agent content output.
18. The system of claim 11 , wherein each sentiment classifier is derived from an emotion classifier resulting from a convolutional neural network (CNN LSTM) model analysis of a corresponding audio stream.
19. The system of claim 18 , wherein the emotion classifier is one selecting from an emotion group comprising angry emotion, excited emotion, frustrated emotion, happy emotion, neutral emotion, sad emotion, and surprised emotion, and the sentiment classifier is one selecting from a sentiment group comprising extremely positive, positive, neutral, negative, extremely negative, and surprised.
20. The system of claim 19 , wherein each sentiment classifier prescribes a corresponding strategy delivered by the agent through follow up conversations that updates the conversation matrix with new audio streams from the follow up conversations.