Abstract: Techniques are provided for reinstantiating digital humans with stored session context in response to navigation to a different destination. One method comprises obtaining information characterizing a user interacting with a digital human instantiated in a first destination (e.g., a webpage or a virtual environment); and, in response to a navigation from the first destination to a different destination: automatically storing session context information in a persistent session context storage based on the obtained information; and reinstantiating the digital human in the different destination with the stored session context information. A query, based on a user input, may be provided to an information retrieval system that generates query results; and a prompt for a language model may be determined based on the query results, wherein the language model generates a response comprising a first content portion comprising a first content type and a second content portion comprising a second content type.

Abstract: Techniques are provided for automatically generating language model prompts using predicted regions of interest. One method comprises obtaining a focus of a user interacting with an image and/or a virtual environment; determining a predicted region of interest of the user based on the focus, wherein the predicted region of interest has associated metadata indicating an item of interest, in the image and/or the virtual environment, to the user; providing a query to an information retrieval system that generates query results, wherein the query is based on the item of interest to the user obtained from the metadata; determining a prompt for a language model based on the query results; and providing a message, automatically generated by the language model based on the prompt, to the user. The predicted region of interest of the user may be obtained by tracking a movement of an eye of the user.

Abstract: Techniques are provided for selecting an isolated speaker signal by comparing text obtained from audio and video streams. One method comprises transforming audio signals from at least one speaker to first sets of predicted spoken words using a speech-to-text conversion model; transforming a video signal to a second set of predicted spoken words using a lip motion-to-text conversion model, wherein the second set of predicted spoken words is based on an analysis of an image associated with a respective speaker; iteratively adjusting a steering vector associated with the audio signals to compare the first sets of predicted spoken words with the second set of predicted spoken words; and selecting an isolated speaker audio signal associated with a particular one of the first sets of predicted spoken words and the second set of predicted spoken words, wherein the selection is based on a result of the comparison.

Abstract: Techniques for pause-based text-to-speech processing for digital humans are provided. One method comprises obtaining a response to be delivered by a digital human, wherein the response comprises at least one predicted pause label identifying a respective portion of the response where a human speaker is expected to pause; processing a token, comprising at least a portion of a word, from a buffer, until a predicted pause label and/or a final token indicator is detected in the token; and providing at least a portion of the given response, from the buffer, to a text-to-speech model for presentation to a user, based on the predicted pause label and/or the final token indicator in the token, wherein the digital human transforms the portion of the given response into the spoken format, using the text-to-speech model, based on the predicted pause label.

Abstract: Techniques are provided for personalizing vehicles using a digital human to administer user preferences. One method comprises automatically detecting an identifier of a user in a vehicle, wherein the identifier of the user is automatically detected based on obtained biometrics of the at least one user and/or a proximity of a device of the user; obtaining stored preferences of the user using the identifier of the user; and automatically configuring the vehicle, by a digital human, using the obtained stored preferences of the user. One or more updated preferences of the user may be stored in one or more persistent slots of a conversation system using the identifier of the user.

Abstract: Techniques are provided for identity-based varied digital human responses. One method comprises obtaining a query, for a digital human, based on an input from a user; obtaining information characterizing permissions of the user based on an identity authentication of the user; obtaining a response to the query from a language model, wherein the response comprises content having an associated access restriction; determining whether the user is authorized to access the content by comparing the permissions of the user to the access restriction; and providing, by the digital human, a response to the user based on a result of the determining. A physical security of an environment of the user and/or a presence of one or more additional users in the environment may be evaluated to determine whether the user is authorized to access the content.

Abstract: Techniques are provided for reinstantiating digital humans with stored session context in response to a device transfer. One method comprises obtaining information characterizing a user interacting with a digital human instantiated in a first device; performing the following steps, in response to a request to transfer the digital human to a different device: automatically storing session context information in a persistent session context storage based on the obtained information; and reinstantiating the digital human in the different device with the stored session context information. The session context information may be stored with a session identifier used to reinstantiate the digital human with the stored session context information.

Abstract: Techniques for providing phoneme-based pronunciations for digital humans are provided. One method comprises obtaining a response, generated by a language model, to be delivered by a digital human in a spoken format, wherein the response comprises multiple words; obtaining a phoneme-based pronunciation for one or more of the words, wherein the phoneme-based pronunciation is based on a user-provided pronunciation obtained from a user prior to the obtaining the response; and providing the phoneme-based pronunciation to the digital human in a processor-readable format, wherein the digital human transforms the processor-readable format into a spoken format using a text-to-speech model. The user-provided pronunciation may be provided by the user in a feedback manner to update a pronunciation employed by the digital human. The phoneme-based pronunciation may be obtained from a hierarchical phoneme repository that employs inheritance across hierarchical levels.

Abstract: Techniques for gesture-based processing of digital human responses are provided. One method comprises obtaining a response, generated by a language model, to be delivered by a digital human to a user, wherein the response comprises a predicted gesture label identifying a gesture associated with the response; presenting a virtual touch display element to the user, based on the predicted gesture label, wherein the virtual touch display element comprises selectable portions; determining coordinates of a gesture of the user, in connection with a given selectable portion; mapping the determined coordinates of the gesture to a selection of a given item associated with a corresponding one of the selectable portions; and initiating an automated action based on the selected given item. Virtual interactions between the user and the digital human may also be processed. A haptic feedback response may be provided to the user in response to a given gesture.

Abstract: Techniques are provided for reinstantiating digital humans with stored session context in response to navigation to a different destination. One method comprises obtaining information characterizing a user interacting with a digital human instantiated in a first destination (e.g., a webpage or a virtual environment); and, in response to a navigation from the first destination to a different destination: automatically storing session context information in a persistent session context storage based on the obtained information; and reinstantiating the digital human in the different destination with the stored session context information. A query, based on a user input, may be provided to an information retrieval system that generates query results; and a prompt for a language model may be determined based on the query results, wherein the language model generates a response comprising a first content portion comprising a first content type and a second content portion comprising a second content type.

Abstract: Techniques are provided for sentiment-based adaptation of digital human responses. One method comprises determining a sentiment of a user by analyzing a vocal sentiment, a text sentiment and/or a facial sentiment of the user; applying the determined sentiment of the user to a language model that determines a sentiment-tagged response to an input of the user based on the determined sentiment, wherein the sentiment-tagged response comprises a predicted sentiment label identifying a sentiment to be employed by a digital human when delivering the sentiment-tagged response to the user; and providing the sentiment-tagged response to the digital human for delivery to the user, wherein the digital human transforms at least a portion of the sentiment-tagged response into a spoken format using the predicted sentiment label and a text-to-speech model. A vocal tone, a facial expression and/or a body positioning of the digital human may be adjusted based on the determined sentiment.

Abstract: Techniques are provided for automatically generating language model prompts using predicted regions of interest. One method comprises obtaining a focus of a user interacting with an image and/or a virtual environment; determining a predicted region of interest of the user based on the focus, wherein the predicted region of interest has associated metadata indicating an item of interest, in the image and/or the virtual environment, to the user; providing a query to an information retrieval system that generates query results, wherein the query is based on the item of interest to the user obtained from the metadata; determining a prompt for a language model based on the query results; and providing a message, automatically generated by the language model based on the prompt, to the user. The predicted region of interest of the user may be obtained by tracking a movement of an eye of the user.

Abstract: Techniques are provided for orienting a digital human towards an isolated human speaker. One method comprises obtaining coordinates associated with a human speaker; storing the coordinates as a variable in a memory of a multi-dimensional graphics engine; instantiating a multi-dimensional plane within the multi-dimensional graphics engine; positioning a multi-dimensional shape in the multi-dimensional plane based on the stored coordinates; and binding a digital human to the multi-dimensional shape to orient the digital human towards a position of the multi-dimensional shape. The coordinates may be obtained by comparing text obtained from one or more video streams to text obtained from one or more audio signals.

Abstract: An apparatus comprises a processing device configured to obtain first parameters characterizing an operating state of information technology (IT) resources of a data center and second parameters characterizing an operating state of cooling systems of the data center, to determine an overall operating state of the data center by aggregating the first and second parameters, to identify a power consumption profile based on the overall operating state, and to perform a joint training of first and second reinforcement learning agents based on the overall operating state and the power consumption profile. The processing device is also configured to generate first controls for the heterogeneous IT resources utilizing the first reinforcement learning agent and second controls for the cooling systems utilizing the second reinforcement learning agent, the first and second controls being configured to reduce power consumption while maintaining specified performance benchmarks for workloads executing in the data center.

Abstract: An apparatus comprises a processing device configured to obtain first parameters characterizing an operating state of information technology (IT) resources of a data center and second parameters characterizing an operating state of cooling systems of the data center, to determine an overall operating state of the data center by aggregating the first and second parameters, to identify a power consumption profile based on the overall operating state, and to perform a joint training of first and second reinforcement learning agents based on the overall operating state and the power consumption profile. The processing device is also configured to generate first controls for the heterogeneous IT resources utilizing the first reinforcement learning agent and second controls for the cooling systems utilizing the second reinforcement learning agent, the first and second controls being configured to reduce power consumption while maintaining specified performance benchmarks for workloads executing in the data center.

Abstract: A method and a non-transitory computer readable medium for generating polygons within a geographic region that satisfy a query are disclosed. For example, the method extracts a plurality of nodes, a plurality of service providers and information associated with the plurality of nodes and the plurality of service providers of the geographic region from a geographic data source, creates a node table based on the information that was extracted, receives the query for one or more nodes having a predefined number of service providers within a travel budget, determines the one or more nodes from the node table that satisfy the query, and generates one or more generate polygons around the one or more nodes that satisfy the query within the geographic region.

Abstract: A method and a non-transitory computer readable medium for generating polygons within a geographic region that satisfy a query are disclosed. For example, the method extracts a plurality of nodes, a plurality of service providers and information associated with the plurality of nodes and the plurality of service providers of the geographic region from a geographic data source, creates a node table based on the information that was extracted, receives the query for one or more nodes having a predefined number of service providers within a travel budget, determines the one or more nodes from the node table that satisfy the query, and generates one or more generate polygons around the one or more nodes that satisfy the query within the geographic region.

Abstract: A system generates an automated survey and causes a user interface of an electronic device to present survey to a patient, so that the system can use the survey responses to identify and provide wellness or disease management content items that are tailored to the patient's behavioral needs. The system automatically generates an electronic survey document with data entry fields that each correspond to a question for a category of behavioral influence. The document includes fillable fields as presented on a user interface. When the system receives data via the fillable fields as presented on the user interface of the electronic device, it will assign a value to each response, construct a patient model, and it will use the patient model to extract intervention content items and provide the patient with those items.

Abstract: Defective memory may cause expensive and unnecessary replacements of the memory especially for higher density dynamic random access memory that has ever shrinking topologies. Running memory stress tests in the background for a period of time at set intervals while the operating system is idle may detect and identify memory problems in real-time without requiring a re-boot of the information handling system. The memory defects may be repaired in real-time so as not to cause loss of data by future read or write requests to the identified defective memory.

Abstract: Defective memory may cause expensive and unnecessary replacements of the memory especially for higher density dynamic random access memory that has ever shrinking topologies. Running memory stress tests in the background for a period of time at set intervals while the operating system is idle may detect and identify memory problems in real-time without requiring a re-boot of the information handling system. The memory defects may be repaired in real-time so as not to cause loss of data by future read or write requests to the identified defective memory.