Conversational bot evaluation and reinforcement using meaningful automated connection scores
Disclosed embodiments provide a framework to assist bot managers and builders in identifying particular friction points between bots and customers to allow for real-time identification of bot conversation issues and to train bots to improve conversation flows. Conversation data is processed using machine learning models to detect bot states within conversations and calculate a Meaningful Automated Connection Score (MACS) for these conversations. The MACS for a conversation is provided to bot builders to allow the bot builders to identify friction points and update bots accordingly.
1. A computer-implemented method, comprising:
detecting one or more messages exchanged between a customer and a bot agent, wherein the bot agent autonomously communicates with the customer through an active communications session;
obtaining conversation data corresponding to the active communications session, wherein the conversation data includes contextual information corresponding to the active communications session;
training a machine learning model to detect different bot states from different communications sessions and to calculate corresponding scores for the different communications sessions, wherein the machine learning model is trained using sample conversation data and corresponding bot states;
detecting one or more bot states corresponding to the bot agent during the active communications session, wherein the one or more bot states are detected by the machine learning model based on the contextual information;
determining a score corresponding to the active communications session, wherein the score is determined by the machine learning model based on the one or more bot states; and
providing in real-time the one or more bot states and the score, wherein the one or more bot states and the score are dynamically updated in real-time as new messages are exchanged during the active communications session.
2. The computer-implemented method of claim 1 , further comprising:
computing different probabilities that different bot states are present in the one or more messages; and
using the different probabilities to further train the machine learning model.
3. The computer-implemented method of claim 1 , wherein the one or more bot states, the score, and messages exchanged during the active communications session are stored in a message bus, and wherein the message bus is updated in real-time as the new messages are exchanged.
4. The computer-implemented method of claim 1 , wherein the one or more messages are detected through real-time monitoring of a cache, and wherein the one or more messages are added to the cache during the active communications session and as the one or more messages are exchanged.
5. The computer-implemented method of claim 1 , further comprising:
detecting that the score is below a threshold for the active communications session; and
transferring the active communications session from the bot agent to a live agent.
6. The computer-implemented method of claim 1 , further comprising:
detecting that the one or more bot states include a particular failure state; and
transferring the active communications session from the bot agent to a live agent, wherein the active communications session is transferred as a result of the particular failure state.
7. The computer-implemented method of claim 1 , further comprising:
receiving feedback corresponding to the one or more bot states and the score; and
dynamically updating the machine learning model based on the feedback.
8. A system, comprising:
one or more processors; and
memory storing thereon instructions that, as a result of being executed by the one or more processors, cause the system to:
detect one or more messages exchanged between a customer and a bot agent, wherein the bot agent autonomously communicates with the customer through an active communications session;
obtain conversation data corresponding to the active communications session, wherein the conversation data includes contextual information corresponding to the active communications session;
train a machine learning model to detect different bot states from different communications sessions and to calculate corresponding scores for the different communications sessions, wherein the machine learning model is trained using sample conversation data and corresponding bot states;
detect one or more bot states corresponding to bot agent during the active communications session, wherein the one or more bot states are detected by the machine learning model based on the contextual information;
determine a score corresponding to the active communications session, wherein the score is determined by the machine learning model based on the one or more bot states; and
provide in real-time the one or more bot states and the score, wherein the one or more bot states and the score are dynamically updated in real-time as new messages are exchanged during the active communications session.
9. The system of claim 8 , wherein the instructions further cause the system to:
compute different probabilities that different bot states are present in the one or more messages; and
use the different probabilities to further train the machine learning model.
10. The system of claim 8 , wherein the one or more bot states, the score, and messages exchanged during the active communications session are stored in a message bus, and wherein the message bus is updated in real-time as the new messages are exchanged.
11. The system of claim 8 , wherein the one or more messages are detected through real-time monitoring of a cache, and wherein the one or more messages are added to the cache during the active communications session and as the one or more messages are exchanged.
12. The system of claim 8 , wherein the instructions further cause the system to:
detect that the score is below a threshold for the active communications session; and
transfer the active communications session from the bot agent to a live agent.
13. The system of claim 8 , wherein the instructions further cause the system to:
detect that the one or more bot states include a particular failure state; and
transfer the active communications session from the bot agent to a live agent, wherein the active communications session is transferred as a result of the particular failure state.
14. The system of claim 8 , wherein the instructions further cause the system to:
receive feedback corresponding to the one or more bot states and the score; and
dynamically update the machine learning model based on the feedback.
15. A non-transitory, computer-readable storage medium storing thereon executable instructions that, as a result of being executed by one or more processors of a computer system, cause the computer system to:
detect one or more messages exchanged between a customer and a bot agent, wherein the bot agent autonomously communicates with the customer through an active communications session;
obtain conversation data corresponding to the active communications session, wherein the conversation data includes contextual information corresponding to the active communications session;
train a machine learning model to detect different bot states from different communications sessions and to calculate corresponding scores for the different communications sessions, wherein the machine learning model is trained using sample conversation data and corresponding bot states;
detect one or more bot states corresponding to bot agent during the active communications session, wherein the one or more bot states are detected by the machine learning model based on the contextual information;
determine a score corresponding to the active communications session, wherein the score is determined by the machine learning model based on the one or more bot states; and
provide in real-time the one or more bot states and the score, wherein the one or more bot states and the score are dynamically updated in real-time as new messages are exchanged during the active communications session.
16. The non-transitory, computer-readable storage medium of claim 15 , wherein the executable instructions further cause the computer system to:
compute different probabilities that different bot states are present in the one or more messages; and
use the different probabilities to further train the machine learning model.
17. The non-transitory, computer-readable storage medium of claim 15 , wherein the one or more bot states, the score, and messages exchanged during the active communications session are stored in a message bus, and wherein the message bus is updated in real-time as the new messages are exchanged.
18. The non-transitory, computer-readable storage medium of claim 15 , wherein the one or more messages are detected through real-time monitoring of a cache, and wherein the one or more messages are added to the cache during the active communications session and as the one or more messages are exchanged.
19. The non-transitory, computer-readable storage medium of claim 15 , wherein the executable instructions further cause the computer system to:
detect that the score is below a threshold for the active communications session; and
transfer the active communications session from the bot agent to a live agent.
20. The non-transitory, computer-readable storage medium of claim 15 , wherein the executable instructions further cause the computer system to:
detect that the one or more bot states include a particular failure state; and
transfer the active communications session from the bot agent to a live agent, wherein the active communications session is transferred as a result of the particular failure state.
21. The non-transitory, computer-readable storage medium of claim 15 , wherein the executable instructions further cause the computer system to:
receive feedback corresponding to the one or more bot states and the score; and
dynamically update the machine learning model based on the feedback.