Verifying compliance of video conference presentation with specified parameters

1 . A method, comprising:

receiving, from a host device of a video conference, an identifier of a presenter device of the video conference and a natural language presentation parameter description;

interpreting, using one or more language models, the natural language presentation parameter description to identify parameters for the video conference, wherein the parameters identified include a slide content requirement and an oral presentation requirement;

receiving, from the presenter device, presentation slides for the video conference and indicia of a trial performance of a presentation for the video conference;

determining, using the one or more language models, whether the presentation slides and the indicia of the trial performance meet at least a subset of the parameters, the subset of the parameters including the slide content requirement and the oral presentation requirement; and

transmitting, to at least one of the host device or the presenter device, an output based on determining whether the presentation slides and the indicia of the trial performance meet the at least the subset of the parameters,

wherein the one or more language models include a pre-trained model that is finetuned using labeled examples of presentation slides or trial performances that are labeled as meeting or failing to meet one or more presentation parameters.

2 . The method of claim 1 , comprising:

determining that the presentation slides and the indicia of the trial performance meet the at least the subset of the parameters; and

generating the output indicating approval of the presentation slides for the video conference.

3 . The method of claim 1 , comprising:

determining that the presentation slides and the indicia of the trial performance do not meet the at least the subset of the parameters; and

generating, using the one or more language models, the output comprising a natural language output instructing a user of the presenter device how to modify at least one of: the presentation slides or the trial performance to meet the at least the subset of the parameters.

4 . The method of claim 1 , wherein the slide content requirement comprises at least one of a number of slides, an amount of text in at least one slide, an amount of white space in the at least one slide, a parameter based on imagery in the at least one slide, a minimum amount of time per slide, or a maximum amount of time per slide, and the oral presentation requirement comprises at least one of a minimum time limit, a maximum time limit, a minimum speaking speed, or a maximum speaking speed.

5 . The method of claim 1 , wherein determining whether the presentation slides and the indicia of the trial performance meet the at least the subset of the parameters comprises:

analyzing, using the one or more language models, whether text content from the presentation slides meets the subset of the parameters;

analyzing, using the one or more language models, whether natural language audio content from the indicia of the trial performance meets the subset of the parameters; and

determining, using the one or more language models, whether the natural language audio content from the indicia of the trial performance accurately corresponds to the text content from the presentation slides.

6 . The method of claim 1 , wherein the slide content requirement comprises a parameter requiring a specific object to be present in imagery of the presentation slides, and wherein determining whether the presentation slides and the indicia of the trial performance meet the at least the subset of the parameters comprises:

analyzing, using a computer vision engine that is trained to recognize objects in images, the imagery in the presentation slides to determine whether a recognized object meets the parameter requiring the specific object; and

transmitting, by the computer vision engine, data indicating whether the imagery in the presentation slides meets the parameter requiring the specific object to the one or more language models.

7 . The method of claim 1 , wherein determining whether the presentation slides and the indicia of the trial performance meet the at least the subset of the parameters comprises:

determining an emotion corresponding to a voice tonality in the trial performance using an audio processing engine; and

transmitting, by the audio processing engine, the emotion to the one or more language models.

8 . The method of claim 1 , wherein the natural language presentation parameter description comprises a recording of a user of the host device speaking, and identifying parameters for the video conference based on the natural language presentation parameter description further comprises interpreting speech in the recording to identify the parameters from the speech.

9 . The method of claim 1 , wherein the presentation slides include text elements, the oral presentation requirement comprises a requirement that each of the text elements is to be discussed in the trial performance, and determining whether the presentation slides and the indicia of the trial performance meet the at least a subset of the parameters further comprises determining, using the one or more language models, whether each of the text elements is discussed in the indicia of the trial performance.

10 . The method of claim 1 , wherein the slide content requirement comprises a requirement that factual content in the presentation slides and the indicia of the trial performance is to be accurate based on a specified source of truth, and determining whether the presentation slides and the indicia of the trial performance meet at least a subset of the parameters further comprises:

obtaining information from the specified source of truth over a network; and

determining, using the one or more language models, whether the factual content in the presentation slides and the indicia of the trial performance is verified by the information obtained from the specified source of truth.

11 . The method of claim 1 , wherein the presentation slides include imagery and the indicia of the trial performance include natural language audio content, and wherein determining whether the presentation slides and the indicia of the trial performance meet the at least the subset of the parameters further comprises determining, using the one or more language models, whether the natural language audio content accurately corresponds to the imagery.

12 . The method of claim 11 , wherein the imagery comprises graph imagery, and wherein determining whether the natural language audio content accurately corresponds to the imagery further comprises determining, using the one or more language models, whether the natural language audio content accurately corresponds to the graph imagery.

13 . The method of claim 1 , wherein identifying the parameters for the video conference based on the natural language presentation parameter description further comprises:

using the one or more language models to interpret the natural language presentation parameter description to generate suggested parameters;

causing presentation of the suggested parameters on a graphical user interface; and

receiving a user input approving or modifying the suggested parameters to establish the parameters for the video conference.

14 . A non-transitory computer readable medium storing instructions operable to cause one or more processors to perform operations comprising:

receiving, from a host device of a video conference, an identifier of a presenter device of the video conference and a natural language presentation parameter description;

interpreting, using one or more language models, the natural language presentation parameter description to identify parameters for the video conference, wherein the parameters identified include a slide content requirement and an oral presentation requirement;

receiving, from the presenter device, presentation slides for the video conference and indicia of a trial performance of a presentation for the video conference;

determining, using the one or more language models, whether the presentation slides and the indicia of the trial performance meet at least a subset of the parameters, the subset of the parameters including the slide content requirement and the oral presentation requirement; and

transmitting, to at least one of the host device or the presenter device, an output based on determining whether the presentation slides and the indicia of the trial performance meet the at least the subset of the parameters,

wherein the one or more language models include a pre-trained model that is finetuned using labeled examples of presentation slides or trial performances that are labeled as meeting or failing to meet one or more presentation parameters.

15 . The non-transitory computer readable medium of claim 14 , the operations comprising:

determining that the presentation slides and the indicia of the trial performance meet the at least the subset of the parameters; and

generating the output comprising text indicating approval of the presentation slides for the video conference.

16 . The non-transitory computer readable medium of claim 14 , the operations comprising:

determining that the presentation slides and the indicia of the trial performance do not meet the at least the subset of the parameters; and

generating, using the one or more language models, the output comprising a natural language output instructing a user of the presenter device how to modify at least one of: the presentation slides or the trial performance to meet the at least the subset of the parameters.

17 . The non-transitory computer readable medium of claim 14 , wherein the slide content requirement comprises a parameter requiring a specific object to be present in imagery of the presentation slides, and wherein determining whether the presentation slides meet the at least the subset of the parameters comprises:

analyzing, using a computer vision engine that is trained to recognize objects in images, the imagery in the presentation slides to determine whether a recognized object meets the parameter requiring the specific object; and

transmitting, by the computer vision engine, data indicating whether the imagery in the presentation slides meets the parameter requiring the specific object to the one or more language models.

18 . The non-transitory computer readable medium of claim 14 , wherein determining whether the presentation slides and the indicia of the trial performance meet the at least the subset of the parameters comprises:

determining an emotion corresponding to a voice tonality in the trial performance using an audio processing engine; and

transmitting, by the audio processing engine, the determined emotion to the one or more language models.

19 . A system, comprising:

a memory subsystem; and

processing circuitry configured to execute instructions stored in the memory subsystem to:

receive, from a host device of a video conference, an identifier of a presenter device of the video conference and a natural language presentation parameter description;

interpret, using one or more language models, the natural language presentation parameter description to identify parameters for the video conference, wherein the parameters identified include a slide content requirement and an oral presentation requirement;

receive, from the presenter device, presentation slides for the video conference and indicia of a trial performance of a presentation for the video conference;

determine, using the one or more language models, whether the presentation slides and the indicia of the trial performance meet at least a subset of the parameters, the subset of the parameters including the slide content requirement and the oral presentation requirement; and

transmit, to at least one of the host device or the presenter device, an output based on determining whether the presentation slides and the indicia of the trial performance meet the at least the subset of the parameters,

wherein the one or more language models include a pre-trained model that is finetuned using labeled examples of presentation slides or trial performances that are labeled as meeting or failing to meet one or more presentation parameters.

20 . The system of claim 19 , the processing circuitry configured to execute the instructions stored in the memory subsystem to:

determine that the presentation slides and the indicia of the trial performance do not meet the at least the subset of the parameters; and

generate, using the one or more language models, the output comprising a natural language output instructing a user of the presenter device how to modify at least one of: the presentation slides or the trial performance to meet the at least the subset of the parameters.