Abstract: Systems and methods are described for providing recommendations for an improved user experience in online meetings. A recommendation engine can aggregate data from user devices to make recommendations before, during and after online meetings. Before a meeting, the recommendation engine can recommend which of a user's devices to use for the meeting. During the meeting, the recommendation engine can identify current or anticipated issues and recommend changes the user can make to correct or prevent the issue. After meetings, the recommendation engine can aggregate data and identify an ongoing issue for one or multiple users. The recommendation engine can identify the cause of the issue and make recommendations to the user or an administrator accordingly.

Abstract: Systems and methods are described for providing recommendations for a user experience in online meetings. A recommendation engine can aggregate data from user devices to make recommendations before, during and after online meetings. Before a meeting, the recommendation engine can recommend which of a user's devices to use for the meeting. During the meeting, the recommendation engine can identify current or anticipated issues and recommend changes the user can make to correct or prevent the issue. After meetings, the recommendation engine can aggregate data and identify an ongoing issue for one or multiple users. The recommendation engine can identify the cause of the issue and make recommendations to the user or an administrator accordingly.

Abstract: Disclosed are various approaches for performing automated actions in a conferencing service. Distractions can be detected and users can be muted. Breakout rooms can be suggested to attendees based upon the user's identity. Additionally, event summaries and recaps can be generated for users who are late-arriving or who depart and return to the event.

Abstract: Disclosed are various approaches for surfacing contextual training programs for users. In some examples, user context data is identified for a user account. The user context data is inputted into a training recommendation model. A training recommendation is generated. The training recommendation recommends a training program that is mapped to the user context data by the training recommendation model. The training recommendation or the training program is surfaced to a client device that is identified by the contextual training service.

Abstract: Disclosed are various examples for to generating personalized user experiences in the form of short-form videos that summarize informational topics that are selected to present to a user according at least one of a user context or a user interest. As the informational topics are identified for a given user, the informational topics can be applied as inputs to a recommendation model that is trained to recommend one or more of the topics to summarize and present to a user in visual form based on at least one of the user context or the user interest. Once a topic is recommended by the recommendation model, a video or other type of visual content can be automatically generated and presented to a user to allow the user to consume the information in the summarized form.

Abstract: Disclosed herein are examples of systems and methods for recommending events using machine learning. A first recommendation can be generated based at least in part on at least one user parameter associated with a user. The first recommendation can comprise a first event. The first recommendation can be provided to a client device associated with the user, and a user response to the recommendation can be received from the client device. A second recommendation can be generated based at least in part on the user response and the at least one user parameter, wherein the second recommendation comprises a second event. The second recommendation can be provided to the client device.

Abstract: Systems and methods are described for providing recommendations for an improved user experience in online meetings. A recommendation engine can aggregate data from user devices to make recommendations before, during and after online meetings. Before a meeting, the recommendation engine can recommend which of a user's devices to use for the meeting. During the meeting, the recommendation engine can identify current or anticipated issues and recommend changes the user can make to correct or prevent the issue. After meetings, the recommendation engine can aggregate data and identify an ongoing issue for one or multiple users. The recommendation engine can identify the cause of the issue and make recommendations to the user or an administrator accordingly.

Abstract: Systems and methods are described for providing recommendations for an improved user experience in online meetings. A recommendation engine can aggregate data from user devices to make recommendations before, during and after online meetings. Before a meeting, the recommendation engine can recommend which of a user's devices to use for the meeting. During the meeting, the recommendation engine can identify current or anticipated issues and recommend changes the user can make to correct or prevent the issue. After meetings, the recommendation engine can aggregate data and identify an ongoing issue for one or multiple users. The recommendation engine can identify the cause of the issue and make recommendations to the user or an administrator accordingly.

Abstract: Disclosed are various approaches for generating a text transcript of a soundtrack. The soundtrack can correspond to an event in a conferencing service. Language models can be trained on data that is specific to organizations, users within the organization, and metadata associated with an agenda for the event. The metadata can include texts, attachments, and other data associated with the event. The language models can be arranged into a convolutional neural network and output a text transcript. The text transcript can be used to retrain the language models for subsequent use.

Abstract: Disclosed are various approaches for authenticating a user through a voice assistant device and creating an association between the device and a user account. The request is associated with a network or federated service. A user account can be implicitly authenticated based on proximity of a client device to the voice assistant device. An association between the user account and the voice assistant device can then be created.

Abstract: Disclosed are various approaches for automating the detection and identification of anomalous devices in a management service. Device check-ins are received by a management service and housed in a data store. The quantity of device check-ins over various time periods can be analyzed using various approaches to identify anomalous devices.

Abstract: Systems and methods are described for securing confidential content in a virtual meeting. In an example, a host can create a meeting agenda that identifies attendees and access restrictions for the meeting. A server can generate encryption keys corresponding to different confidentiality levels of content identified in the agenda. The encryption keys can be provided to user devices of the attendees based on the access permissions of the attendees. As the virtual meeting takes place, the server can encrypt confidential portions of the content stream for the meeting using the encryption keys. In an example, the server can also apply a machine learning algorithm to identify any confidential content in the content stream not previously identified. The server can encrypt additional confidential content where appropriate using encryption keys corresponding to the confidentiality level of the content.

Abstract: Disclosed are various embodiments for generating a list of predicted contacts that can be provided to a client device. The predicted contacts can be generated based upon an analysis of user interaction data. The predicted contacts can be made available to a phone application or messaging application on a client device so that contact information can be displayed in response to an incoming call or a message.

Abstract: Systems and methods for 3D printer management can allow or reject printing of an object based on a model that is trained with machine learning. In one example, the model classifies the object according to object type. The object type can be compared against a list, such as a whitelist or blacklist, to determine whether to block the object from printing. The lists can be specific to users, such as based on an organizational group to which the user belongs. A print server can apply the model prior to forwarding the object to a 3D printer for printing. Both the models and the lists can evolve based on machine learning, such as based on which print decisions receive override from administrators.

Abstract: Disclosed herein are examples of systems and methods for formatting electronic messages using machine learning. An electronic message can be obtained, and a processed message can be generated based at least in part on the electronic message. At least one attribute for the processed message can be determined. A formatting specification can be generated based at least in part on the at least one attribute. A reformatted message can be generated based at least in part on the formatting specification.

Abstract: Disclosed are various approaches for automating the detection and identification of anomalous devices in a management service. Device check-ins are received by a management service and housed in a data store. The quantity of device check-ins over various time periods can be analyzed using various approaches to identify anomalous devices.

Abstract: Disclosed are various approaches for performing automated actions in a conferencing service. Distractions can be detected and users can be muted. Breakout rooms can be suggested to attendees based upon the user's identity. Additionally, event summaries and recaps can be generated for users who are late-arriving or who depart and return to the event.

Abstract: Disclosed are various approaches for generating a text transcript of a soundtrack. The soundtrack can correspond to an event in a conferencing service. Language models can be trained on data that is specific to organizations, users within the organization, and metadata associated with an agenda for the event. The metadata can include texts, attachments, and other data associated with the event. The language models can be arranged into a convolutional neural network and output a text transcript. The text transcript can be used to retrain the language models for subsequent use.

Abstract: Examples of generating contextual compliance policies are described. A meeting invitation can be created by a meeting organizer. The meeting can have compliance rules associated therewith. A compliance policy that facilitates enforcement of the compliance rules during the meeting can be transmitted to attendee client devices and enforced by a management component running on the client devices.

Abstract: Systems and methods are described for securing confidential content in a virtual meeting. In an example, a host can create a meeting agenda that identifies attendees and access restrictions for the meeting. A server can generate encryption keys corresponding to different confidentiality levels of content identified in the agenda. The encryption keys can be provided to user devices of the attendees based on the access permissions of the attendees. As the virtual meeting takes place, the server can encrypt confidential portions of the content stream for the meeting using the encryption keys. In an example, the server can also apply a machine learning algorithm to identify any confidential content in the content stream not previously identified. The server can encrypt additional confidential content where appropriate using encryption keys corresponding to the confidentiality level of the content.