Adjusting a verbosity of a conversation turn

Patent metadata

Field	Value
Publication number	US-11238234-B2
Application number	US-201916568067-A
Country	US
Kind code	B2
Filing date	Sep 11, 2019
Priority date	Sep 11, 2019
Publication date	Feb 1, 2022
Grant date	Feb 1, 2022

Abstract

In one general aspect, a computer-implemented method includes identifying current choices with different verbosity levels for a current turn in a conversation; normalizing multi-dimensional verbosity vectors for each of the current choices to obtain a normalized value for each of the current choices; determining a state definition for the current turn in the conversation, utilizing the normalized values for each of the current choices; providing the state definition for the current turn in the conversation and the normalized values for each of the current choices to a trained reinforcement learning module; receiving, from the trained reinforcement learning module, a score associated with each of the current choices for the current turn in the conversation; and selecting one of the current choices to be entered for the current turn in the conversation, based on the score associated with each of the current choices for the current turn in the conversation.

First claim

What is claimed is:

1. A computer-implemented method, comprising:
identifying current choices with different verbosity levels for a current turn in a conversation;
normalizing multi-dimensional verbosity vectors for each of the current choices to obtain a normalized value for each of the current choices;
determining a state definition for the current turn in the conversation, utilizing the normalized values for each of the current choices;
providing the state definition for the current turn in the conversation and the normalized values for each of the current choices to a trained reinforcement learning module;
receiving, from the trained reinforcement learning module, a score associated with each of the current choices for the current turn in the conversation; and
selecting one of the current choices to be entered for the current turn in the conversation, based on the score associated with each of the current choices for the current turn in the conversation.

2. The computer-implemented method of claim 1 , wherein the conversation includes a question and answer (Q and A) conversation.

3. The computer-implemented method of claim 1 , wherein each of the current choices includes a textual word or phrase having wording and vocabulary different from all other current choices.

4. The computer-implemented method of claim 1 , wherein each of the current choices includes a multi-dimensional verbosity vector describing a verbosity of the choice.

5. The computer-implemented method of claim 1 , wherein the multi-dimensional verbosity vectors for each current choice each include a value for a number of words used within the current choice.

6. The computer-implemented method of claim 1 , wherein the multi-dimensional verbosity vectors for each current choice each include a value for a type of words used within the current choice.

7. The computer-implemented method of claim 1 , wherein normalizing the multi-dimensional verbosity vectors for each of the current choices includes normalizing a word count for each of the current choices.

8. The computer-implemented method of claim 1 , wherein normalizing the multi-dimensional verbosity vectors for each of the current choices includes normalizing one or more domain terminologies for each of the current choices.

9. The computer-implemented method of claim 1 , wherein the state definition is determined by summing an absolute distance of each of a plurality of verbosity vectors of past turns in the conversation.

10. The computer-implemented method of claim 1 , wherein the state definition is determined by calculating a difference between a normalized verbosity value for a last turn in the conversation and the normalized value for each of the current choices.

11. The computer-implemented method of claim 1 , wherein the trained reinforcement learning module includes a Q-learning method.

12. The computer-implemented method of claim 1 , wherein for each of the current choices, the score includes a value indicating a suitability of the current choice for use within the current turn in the conversation.

13. The computer-implemented method of claim 1 , wherein a current choice having a highest score is determined to be least likely to confuse a user in the conversation, while minimizing a verbosity level, and is selected to be entered for the current turn in the conversation.

14. The computer-implemented method of claim 1 , wherein the state definition is determined by summing an absolute distance of each of a plurality of verbosity vectors.

15. A computer program product for adjusting a verbosity of a conversation turn, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, wherein the computer readable storage medium is not a transitory signal per se, the program instructions executable by a processor to cause the processor to perform a method comprising:
identifying, by the processor, current choices with different verbosity levels for a current turn in a conversation;
normalizing, by the processor, multi-dimensional verbosity vectors for each of the current choices to obtain a normalized value for each of the current choices;
determining, by the processor, a state definition for the current turn in the conversation, utilizing the normalized values for each of the current choices;
providing, by the processor, the state definition for the current turn in the conversation and the normalized values for each of the current choices to a trained reinforcement learning module;
receiving, from the trained reinforcement learning module by the processor, a score associated with each of the current choices for the current turn in the conversation; and
selecting, by the processor, one of the current choices to be entered for the current turn in the conversation, based on the score associated with each of the current choices for the current turn in the conversation.

16. The computer program product of claim 15 , wherein the conversation includes a question and answer (Q and A) conversation.

17. The computer program product of claim 15 , wherein each of the current choices includes a textual word or phrase having wording and vocabulary different from the other current choices.

18. The computer program product of claim 15 , wherein each of the current choices includes a multi-dimensional verbosity vector describing a verbosity of the choice.

19. The computer program product of claim 15 , wherein the multi-dimensional verbosity vectors for each current choice each include a value for a number of words used within the current choice.

20. A system, comprising:
a processor; and
logic integrated with the processor, executable by the processor, or integrated with and executable by the processor, the logic being configured to:
identify current choices with different verbosity levels for a current turn in a conversation;
normalize multi-dimensional verbosity vectors for each of the current choices to obtain a normalized value for each of the current choices;
determine a state definition for the current turn in the conversation, utilizing the normalized values for each of the current choices;
provide the state definition for the current turn in the conversation and the normalized values for each of the current choices to a trained reinforcement learning module;
receive, from the trained reinforcement learning module, a score associated with each of the current choices for the current turn in the conversation; and
select one of the current choices to be entered for the current turn in the conversation, based on the score associated with each of the current choices for the current turn in the conversation.

Assignees

• IBM

Inventors

• Jiang Shun
• Moore Robert John
• Szymanski Margaret Helen
• Huang Lei
• Ren Guangjie
• Yin Peifeng

Classifications

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06N3/084

  Backpropagation, e.g. using gradient descent · CPC title

• G06F40/30 Primary

  Semantic analysis · CPC title

• G06N3/006

  based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO] · CPC title

• G06N20/00

  Machine learning · CPC title