Automated generation of content-related discussion
Systems, devices and automated processes are described for automatically generating discussion topics such as questions or the like that are related to particular media programs being watched by one or more viewers. The questions are generated by a large language model (LLM) or similar artificial intelligence (AI) construct executed by a content provider, by the viewer's device, or by another service. The automatically-generated questions can greatly improve the viewer's experience by provoking thought or discussion, and by fostering a community atmosphere related to the particular program being viewed.
1 . An automated process performed by a computer system comprising a processor, a non-transitory digital storage and an interface to a network, the automated process comprising:
receiving metadata about a media program to be watched by a viewer operating a remotely-located viewer device;
providing at least some of the metadata about the media program to thereby train a large language model (LLM) about the media program;
subsequently receiving, from the remotely-located viewer device, an indication that the viewer is watching the media program;
in response to the indication that the viewer is watching the media program, formulating a natural language query based upon the metadata and placing the natural language query to the LLM and thereby instruct the LLM to create one or more automatically-generated discussion topics related to the media program, wherein the LLM responsively creates the one or more automatically-created discussion topics for the viewer in real time based upon the previously-received metadata about the media program; and
providing the automatically-generated one or more discussion topics for presentation to a viewer while the viewer is watching the media program on the remotely-located viewer device.
2 . The automated process of claim 1 wherein the LLM is trained about content of the media program by converting at least some of the metadata to mathematical vectors, the content of the media program comprising characters and plot elements occurring in the media program.
3 . The automated process of claim 1 wherein the LLM executes on a processor associated with a question generation engine that communicates with the viewer device via a network.
4 . The automated process of claim 1 wherein the indication comprises timing information describing a viewing point in the media program that has been viewed by the viewer, and wherein the natural language query instructs the LLM to consider the timing information in responsively creating the one or more discussion questions so that the one or more discussion questions relate only to characters and plot events occurring in the content of the media program that has been viewed by the viewer up to the viewing point in the media program.
5 . The automated process of claim 2 wherein the metadata comprises timed text related to the content of the media program, and wherein the automated processing comprises analyzing the timed text related to the media program to thereby identify characters and plot elements of the media program for training the LLM.
6 . The automated process of claim 2 wherein the metadata comprises audio content related to the media program, and wherein the automated processing comprises performing a speech-to-text conversion on the audio content and analyzing the resulting text related to the media program to thereby identify characters and plot elements of the media program for training the LLM.
7 . A data processing system comprising:
an input interface configured to receive metadata about a media program that is to be presented to a viewer operating a remotely-located media device;
a question generation engine implemented with a processor of the data processing system, the question generation engine configured to:
receive, from the remotely-located viewer device via a network, a natural language query comprising an indication that the viewer is watching the media program; and
place the received natural language query to a large language model (LLM) and thereby instruct the LLM to create one or more automatically-generated discussion topics related to the media program, wherein the LLM responsively creates the one or more automatically-created discussion topics for the viewer in real time based upon the previously-received metadata about the media program; and
a network interface to communicate with the remotely-located media device and thereby provide the automatically-generated one or more discussion topics to the remotely-located media device for presentation to a viewer while the viewer is watching the media program.
8 . The data processing system of claim 7 wherein the large language model (LLM) of the question generation engine is trained on the metadata about the media program and configured to generate the one or more discussion topics based upon characters and plot events of the media program described by the metadata.
9 . The data processing system of claim 7 wherein the natural language query received from the remotely-located viewer device comprises timing information describing a viewing point in the media program that has been viewed by the viewer, and wherein the natural language query instructs the LLM to consider the timing information in responsively creating the one or more discussion questions so that the one or more discussion questions relate only to characters and plot events that have occurred in the content of the media program that has been viewed by the viewer prior to the viewing point in the media program.
10 . The data processing system of claim 8 wherein the network interface is further configured to receive a viewer response to the one or more discussion topics, wherein the question generation engine is further configured to responsively create one or more follow up discussion topics based upon the characters and the plot events of the media program and the viewer response, and wherein the network interface is further configured to provide the one or more follow up discussion topics to the remotely-located media device for presentation to the viewer.
11 . The data processing system of claim 10 wherein the natural language query received from the remotely-located viewer device comprises timing information describing a viewing point in the media program that has been viewed by the viewer, and wherein the natural language query instructs the LLM to consider the timing information in responsively creating the one or more discussion questions so that the one or more discussion questions relate only to characters and plot events that have occurred in the content of the media program that has been viewed by the viewer prior to the viewing point in the media program.