Abstract: One embodiment of the present invention sets forth a technique for processing recordings of events. The technique includes applying a machine learning model to a plurality of samples from one or more recordings of the events to generate a plurality of embeddings representing the plurality of samples. The technique also includes generating a plurality of labels that identify speakers associated with the plurality of samples based on metadata comprising timestamps of voice activity during the events and participants associated with the voice activity. The technique further includes storing mappings of the plurality of embeddings to the plurality of labels.

Abstract: One embodiment of the present invention sets forth a technique for processing recordings of events. The technique includes applying a machine learning model to a plurality of samples from one or more recordings of the events to generate a plurality of embeddings representing the plurality of samples. The technique also includes generating a plurality of labels that identify speakers associated with the plurality of samples based on metadata comprising timestamps of voice activity during the events and participants associated with the voice activity. The technique further includes storing mappings of the plurality of embeddings to the plurality of labels.

Abstract: Embodiments are directed towards employing multiple profiles to track identification of users across multiple domains. Multiple application-domain profiles and multiple web-domain profiles may each include a statistical identifier. A profile enrichment store may be generated from the application-domain profiles and web-domain profiles, where each entry in the profile enrichment store may be based on statistical identifiers. The profile enrichment store may be utilized to determine categories associated with a statistical identifier for a request from a client device. A targeted advertisement may be determined based on the determined category and may be provided to the client device. A client caching identifier may be used in a mobile web domain, in combination with the statistical identifier, to track mobile web users. If a request is associated with a profile that includes an affirmative opt-out flag, categories in the profile enrichment store for a corresponding statistical identifier may be cleaned up.

Abstract: A server receives incoming data records comprising an ID value and respective user attributes corresponding to a user activity that originates at a user device. Characteristics of the activity and/or characteristics of the user device are considered in order to assign a probabilistic confidence value, which is in turn used to generate links from an incoming data record to other previously-received data records, and in so doing, generates a probabilistic link between one set of user attributes from the incoming data record and another set of user attributes from previously-received data records. A messaging campaign specification that describes target user attributes is used to identify a matching set of target audience member records. The determination of a match or not is based on the probabilistic confidence value and a threshold can be varied to extend audience reach. The identified set of target audience member records are transmitted over a network.

Abstract: In one embodiment, the interactive virtual meeting assistant implements a meeting debrief post-processing operation. For a given meeting that the interactive virtual meeting assistant participated in, a meeting post-processing engine enables one or more participants of the meeting to associate a debrief with the meeting. The debrief may be an audio recording, a video recording, text, or any file. The meeting post-processing engine stores the debrief in data stores and provides access to the debrief via the meeting GUI associated with the meeting. The meeting post-processing engine also processes the debrief to generate tasks to be assigned to participants or other entities and/or schedule reminders to be provided to the participants or other entities. The debrief may be private, such that only the participant who provided the debrief may subsequently access the debrief.

Abstract: The present disclosure provides a detailed description of techniques used in methods, systems, and computer program products for associating anonymous information to personally identifiable information without sharing any personally identifiable information. A method receives a specification record comprising one or more specified demographic attributes to be used in user record selection operations, the results of which operations include user records that comprise a user identifier and at least some non-personally-identifiable information. A candidate group is formed by applying a set of rules over the retrieved user records to reject-out one or more user records that comprise mutually-exclusive characteristics with respect to the other user records in the candidate group. An anonymity measure is calculated over the candidate group to satisfy a threshold of anonymity. If needed to satisfy the threshold of anonymity, additional user records are added to the group before any sharing operations.

Abstract: The disclosure is directed to approaches for combining server-side decisions based on user attributes and/or demographics with client-side decisions based on expressed or implied user interests determined based on the contents of a web page being visited by the user. A server is responsive to an occurrence of an operation of a client device at the web page. Processing steps determine particular executable web page content to send to the client device (e.g., where the particular executable web page content is determined in response to a particular client-side operation). The web page is rendered with the aforementioned executable web page content. Server-side processing receives messages that are emitted conditionally, based on the occurrence of a user event. An event log message comprises a portion of the web page, such as a string, indicating or implying user interest. A targeted advertisement is presented to the user based on the user interest.

Abstract: One embodiment of the present invention sets forth a technique for organizing meeting content. The technique includes generating, from a set of available meetings, a thread comprising a collection of related meetings that share one or more attributes. The technique also includes aggregating data for the related meetings, where the data comprises metadata for the related meetings and terms included in recordings of the related meetings. The technique further includes outputting at least a portion of the aggregated data within a summary of the thread.

Abstract: The present disclosure provides a detailed description of techniques used in methods, systems, and computer program products for associating anonymous information with personally identifiable information without sharing any personally identifiable information. A method receives a specification record comprising one or more specified demographic attributes to be used in user record selection operations, the results of which operations include user records that comprise a user identifier and at least some non-personally-identifiable information. A candidate group is formed by applying a set of rules over the retrieved user records to exclude one or more user records that comprise mutually-exclusive characteristics with respect to the other user records in the candidate group. An anonymity measure is calculated over the candidate group to satisfy a threshold of anonymity. If needed to satisfy the threshold of anonymity, additional user records are added to the group before any sharing operations.

Abstract: One embodiment of the present invention sets forth a technique for processing recordings of events. The technique includes applying a machine learning model to a plurality of samples from one or more recordings of the events to generate a plurality of embeddings representing the plurality of samples. The technique also includes generating a plurality of labels that identify speakers associated with the plurality of samples based on metadata comprising timestamps of voice activity during the events and participants associated with the voice activity. The technique further includes storing mappings of the plurality of embeddings to the plurality of labels.

Abstract: In one embodiment, the interactive virtual meeting assistant implements a meeting debrief post-processing operation. For a given meeting that the interactive virtual meeting assistant participated in, a meeting post-processing engine enables one or more participants of the meeting to associate a debrief with the meeting. The debrief may be an audio recording, a video recording, text, or any file. The meeting post-processing engine stores the debrief in data stores and provides access to the debrief via the meeting GUI associated with the meeting. The meeting post-processing engine also processes the debrief to generate tasks to be assigned to participants or other entities and/or schedule reminders to be provided to the participants or other entities. The debrief may be private, such that only the participant who provided the debrief may subsequently access the debrief.

Abstract: One embodiment of the present invention sets forth a technique for organizing meeting content. The technique includes generating, from a set of available meetings, a thread comprising a collection of related meetings that share one or more attributes. The technique also includes aggregating data for the related meetings, where the data comprises metadata for the related meetings and terms included in recordings of the related meetings. The technique further includes outputting at least a portion of the aggregated data within a summary of the thread.

Abstract: One embodiment of the present invention sets forth a technique for processing recordings of events. The technique includes applying a machine learning model to a plurality of samples from one or more recordings of the events to generate a plurality of embeddings representing the plurality of samples. The technique also includes generating a plurality of labels that identify speakers associated with the plurality of samples based on metadata comprising timestamps of voice activity during the events and participants associated with the voice activity. The technique further includes storing mappings of the plurality of embeddings to the plurality of labels.

Abstract: The present disclosure provides a detailed description of techniques used in systems, methods, and in computer program products for building and operating a match cooperative without handling personally identifiable information. The various embodiments address the problem of discovering attributes pertaining to a particular user without sharing personally identifiable information pertaining to that particular user. More specifically, the claimed embodiments are directed to approaches for receiving online and offline PII and NPII associated with various users, obfuscating (e.g., hashing) the PII, and matching the obfuscated PII to the NPII based on various data (e.g., common attributes, etc.) and methods (e.g., deterministic matching, probabilistic matching, etc.). The matched NPII attributes can then be used to target the user associated with the obfuscated PII in online advertising campaigns.

Abstract: One embodiment of the present invention sets forth a technique for processing recordings of events. The technique includes applying a machine learning model to a plurality of samples from one or more recordings of the events to generate a plurality of embeddings representing the plurality of samples. The technique also includes generating a plurality of labels that identify speakers associated with the plurality of samples based on metadata comprising timestamps of voice activity during the events and participants associated with the voice activity. The technique further includes storing mappings of the plurality of embeddings to the plurality of labels.

Abstract: Embodiments are directed towards employing a plurality of tag states to control tag suspension based on an asynchronous process that proactively monitors tag performance, response times, privacy safeguards, and latency. Tags may be in one of multiple states. Tags in a NORMAL state or a FLAGGED state may be enabled for deployment, and tags in a SUSPENSION A state or a SUSPENSION B state may be blocked from deployment. A tag's state may change based on monitoring the tag's latency to determine if latent events occur. Tag latency may be asynchronously monitored independent of web page requests or requests to execute an application. For example, if latent events occur, then the tag's state may change from NORMAL to FLAGGED, and if non-latent events occur, then the tag's state may change from FLAGGED to NORMAL.

Abstract: The disclosure is directed to approaches for combining server-side decisions based on user attributes and/or demographics with client-side decisions based on expressed or implied user interests determined based on the contents of a web page being visited by the user. A server is responsive to an occurrence of an operation of a client device at the web page. Processing steps determine particular executable web page content to send to the client device (e.g., where the particular executable web page content is determined in response to a particular client-side operation). The web page is rendered with the aforementioned executable web page content. Server-side processing receives messages that are emitted conditionally, based on the occurrence of a user event. An event log message comprises a portion of the web page, such as a string, indicating or implying user interest. A targeted advertisement is presented to the user based on the user interest.

Abstract: A server receives incoming data records comprising an ID value and respective user attributes corresponding to a user activity that originates at a user device. Characteristics of the activity and/or characteristics of the user device are considered in order to assign a probabilistic confidence value, which is in turn used to generate links from an incoming data record to other previously-received data records, and in so doing, generates a probabilistic link between one set of user attributes from the incoming data record and another set of user attributes from previously-received data records. A messaging campaign specification that describes target user attributes is used to identify a matching set of target audience member records. The determination of a match or not is based on the probabilistic confidence value and a threshold can be varied to extend audience reach. The identified set of target audience member records are transmitted over a network.

Abstract: Embodiments are directed towards employing multiple profiles to track identification of users across multiple domains. Multiple application-domain profiles and multiple web-domain profiles may each include a statistical identifier. A profile enrichment store may be generated from the application-domain profiles and web-domain profiles, where each entry in the profile enrichment store may be based on statistical identifiers. The profile enrichment store may be utilized to determine categories associated with a statistical identifier for a request from a client device. A targeted advertisement may be determined based on the determined category and may be provided to the client device. A client caching identifier may be used in a mobile web domain, in combination with the statistical identifier, to track mobile web users. If a request is associated with a profile that includes an affirmative opt-out flag, categories in the profile enrichment store for a corresponding statistical identifier may be cleaned up.

Abstract: The disclosure is directed to approaches for combining server-side decisions based on user attributes and/or demographics with client-side decisions based on expressed or implied user interests determined based on the contents of a web page being visited by the user. A server is responsive to an occurrence of an operation of a client device at the web page. Processing steps determine particular executable web page content to send to the client device (e.g., where the particular executable web page content is determined in response to a particular client-side operation). The web page is rendered with the aforementioned executable web page content. Server-side processing receives messages that are emitted conditionally, based on the occurrence of a user event. An event log message comprises a portion of the web page, such as a string, indicating or implying user interest. A targeted advertisement is presented to the user based on the user interest.