Method and system of automated question generation for speech assistance
A method and system for generating one or more questions relating to a presentation session includes receiving audio data from the presentation session, retrieving a transcript for the audio data, receiving other data relating to the presentation session, providing at least one of the transcript and the other data to a machine-learning (ML) model as input for automatically generating the one or more questions relating to the presentation session, receiving from the ML model the one or more questions, and providing the one or more questions for display on a user interface associated with the presentation session.
1. A data processing system comprising:
a processor; and
a memory in communication with the processor, the memory comprising executable instructions that, when executed by the processor or another element, cause the data processing system to perform functions of:
receiving audio data from a practice presentation session;
retrieving a transcript for the audio data;
receiving other data relating to the practice presentation session, the other data being extracted from content of presentation materials used to conduct the practice presentation session;
generating, at a questions engine, a prompt to be provided as an input to a natural language generation (NLG) model, by utilizing a prompt script that includes blank spaces for inserting data relating to at least one of the transcript and the other data, wherein the prompt script allows the NLG model to perform zero-shot adaptation in real time, and the NLG model includes a model that receives the prompt and generates one or more questions that are likely to be asked during a live presentation session based on the transcript of the practice presentation session and the content of the presentation materials;
providing the prompt to the NLG model as input for automatically generating the one or more questions;
receiving from the NLG model the one or more questions;
providing the one or more questions for display on a user interface associated with the practice presentation session; and
training the NLG model using a training data set of textual data.
2. The data processing system of claim 1 , wherein the other data includes as least one of content of a presentation document, multimodal content, content of a document shared during the presentation session or one or more questions from a previous presentation session.
3. The data processing system of claim 1 , wherein the practice presentation session is at least one of a speech rehearsal session or a virtual meeting.
4. The data processing system of claim 1 , wherein the audio data is received in real time while the practice presentation session is occurring.
5. The data processing system of claim 1 , wherein the one or more questions are displayed in real time during the practice presentation session.
6. The data processing system of claim 1 , wherein the user interface includes a UI element for receiving a response to the one or more questions from a user.
7. The data processing system of claim 6 , wherein the audio data includes data relating to the response and the executable instructions, when executed by the processor, further cause the data processing system to perform functions of:
providing at least one of the transcript and the other data to a second ML model as input for evaluating the response;
receiving from the second ML model one or more evaluation results for the response; and
providing the one or more evaluation results for display on the user interface associated with the practice presentation session.
8. The data processing system of claim 7 , wherein the evaluation results include at least one of a first indication of clarity of the response, and a second indication of a length of the response.
9. A method for generating one or more questions relating to a presentation session comprising:
receiving audio data from a practice presentation session;
retrieving a transcript for the audio data;
receiving other data relating to the practice presentation session, the other data being extracted from content of presentation materials used to conduct the practice presentation session;
generating, at a questions engine, a prompt to be provided as an input to a natural language generation (NLG) model, by utilizing a prompt script that includes blank spaces for inserting data relating to at least one of the transcript and the other data wherein the prompt script allows the NLG model to perform zero-shot adaptation in real time, and the NLG model includes a model that receives the prompt and generates one or more questions that are likely to be asked during a live presentation session based on the transcript of the practice presentation session and the content of the presentation materials;
providing the prompt to the NLG model as input for automatically generating the one or more questions;
receiving from the NLG model the one or more questions;
providing the one or more questions for display on a user interface associated with the practice presentation session; and
training the NLG model using a training data set of textual data.
10. The method of claim 9 , wherein the other data includes as least one of content of a presentation document, multimodal content, content of a document shared during the presentation session or one or more questions from a previous presentation session.
11. The method of claim 9 , wherein the practice presentation session is at least one of a speech rehearsal session or a virtual meeting.
12. The method of claim 9 , wherein the audio data is received in real time while the practice presentation session is occurring.
13. The method of claim 9 , wherein the one or more questions are displayed in real time during the practice presentation session.
14. The method of claim 9 , wherein the one or more questions are displayed after the practice presentation session is completed.
15. The method of claim 9 , wherein the user interface includes a UI element for receiving a response to the one or more questions from a user.
16. The method of claim 15 , wherein the audio data includes data relating to the response and the method further comprises:
providing at least one of the transcript and the other data to a second ML model as input for evaluating the response;
receiving from the second ML model one or more evaluation results for the response; and
providing the one or more evaluation results for display on the user interface associated with the practice presentation session.
17. A non-transitory computer readable medium on which are stored instructions that, when executed, cause a programmable device to:
receive audio data from a practice presentation session;
retrieve a transcript for the audio data;
receiving other data relating to the practice presentation session, the other data being extracted from content of presentation materials used to conduct the practice presentation session;
generating, at a questions engine, a prompt to be provided as an input to a natural language generation (NLG) model, by utilizing a prompt script that includes blank spaces for inserting data relating to at least one of the transcript and the other data wherein the prompt script allows the NLG model to perform zero-shot adaptation in real time, and the NLG model includes a model that receives the prompt and generates one or more questions that are likely to be asked during a live presentation session based on the transcript of the practice presentation session and the content of the presentation materials;
providing the prompt to the NLG model as input for automatically generating the one or more questions;
receive from the NLG model the one or more questions;
provide the one or more questions for display on a user interface associated with the practice presentation session; and
training the NLG model using a training data set of textual data.
18. The non-transitory computer readable medium of claim 17 , wherein the other data includes as least one of content of a presentation document, multimodal content, content of a document shared during the presentation session or one or more questions from a previous presentation session.
19. The non-transitory computer readable medium of claim 17 , wherein the user interface includes a UI element for receiving a response to the one or more questions from a user.
20. The non-transitory computer readable medium of claim 19 , wherein the audio data includes data relating to the response and the instructions when executed, further cause a programmable device to:
provide at least one of the transcript and the other data to a second ML model as input for evaluating the response;
receive from the second ML model one or more evaluation results for the response; and
provide the one or more evaluation results for display on the user interface associated with the practice presentation session.