Abstract: Systems and methods for detecting and using semantic relatedness to classify segments of digital text are disclosed. More particularly, embodiments determine the semantic relatedness of segments of text to abstract categories where the abstract categories are not defined by a single word or semantic concept. Detecting semantic relatedness includes analyzing text, embedding the text, and determining semantic relatedness to a set of concepts for a category where each concept may include a set of words/phrases embedded in a similar fashion. The text embedding can be projected onto each concept embedding and reduced to a score representing semantic relatedness. The text is classified based on the semantic relatedness.

Abstract: A data security system interfaces with a plurality of heterogeneous online platforms, each having its own platform specific set of data compliance or security settings. The data security system maps a set of common data compliance or security settings to the platform specific data compliance or security settings and monitors a state of the platform specific set of data compliance or security settings. If the data security system determines that the state of a platform specific data compliance or security settings do not comply with a specified common data compliance or security setting state, the system automatically changes the state of the platform specific data compliance or security settings to comply with the common data compliance or security settings setting state.

Abstract: A data security system uses API connections to heterogeneous collaboration platforms to analyze elements of conversations to identify regulatory, security, and privacy risks and take corresponding actions to limit and remediate risks in real-time. In some embodiments, the data security system can use API event triggers to determine when an action is taken, apply a machine learning-based detections to identify relevant risks, and take a corresponding remedial action such as removing a user from a conversation.

Abstract: A data security system interfaces with a plurality of heterogeneous online collaboration platforms, each having its own platform specific set of collaboration settings to control access to collaborative data. The data security system maps a set of common collaboration settings to the platform specific set of collaboration settings of each heterogeneous online collaboration platform and monitors a state of the platform specific set of collaboration settings of each heterogeneous online collaboration platform. If the data security system determines that the state of a platform specific collaboration setting does not comply with a specified common collaboration setting state, the data security system automatically changes the state of the platform specific collaboration setting to comply with the common collaboration setting state.

Abstract: Systems, methods and computer program products for sample efficient training of machine learning models are provided. A process may proceed, starting with an initial set of labeled examples and the initial set of unlabeled examples, to label unlabeled examples in an iterative manner, with the input dataset for a next iteration comprising an augmented set of labeled examples from a current iteration and selected unlabeled examples, until a final set of labeled examples is created. The final set of labeled examples is used to train a machine learning model. Each iteration includes mapping the input dataset to a reduced dimension space and using the reduced dimension space to identify high value examples to label.

Abstract: A security and compliance platform ingests content from heterogeneous collaboration platforms and processes the content to detect potential regulatory, privacy, and security risks in the conversations. One of the detections that can be applied is the identification of application windows appearing during a collaboration screen share. Because these applications may contain sensitive personally identifiable information (PII), technical trade secrets, or highly confidential corporate information, there is meaningful utility in identifying instances when these applications are shared. If applications shared on screen is detected, a corresponding workflow action may be implemented, for example flagging a session for further analysis in a review screen of the security and compliance platform.

Abstract: A data security system interfaces with a plurality of heterogeneous online collaboration platforms, each having its own platform specific set of collaboration settings to control access to collaborative data. The data security system maps a set of common collaboration settings to the platform specific set of collaboration settings of each heterogeneous online collaboration platform and monitors a state of the platform specific set of collaboration settings of each heterogeneous online collaboration platform. If the data security system determines that the state of a platform specific collaboration setting does not comply with a specified common collaboration setting state, the data security system automatically changes the state of the platform specific collaboration setting to comply with the common collaboration setting state.

Abstract: A data security system collects collaboration platform collaboration settings through interfaces and analyzes those settings against a policy of common collaboration settings for security and compliance. The data security system interfaces with a plurality of heterogeneous online collaboration platforms, each having its own platform specific set of collaboration settings to control access to collaborative data. The data security system maps the set of common collaboration settings to the platform specific set of collaboration settings of each heterogeneous online collaboration platform from the plurality of heterogeneous online collaboration platforms and monitors a state of the platform specific set of collaboration settings of each heterogeneous online collaboration platform. Based on a determination that a platform specific collaboration setting does not comply with the collaboration setting policy, the data security system can take a remedial action or other action.

Abstract: Systems and methods for enhanced rule-based querying of unstructured text using graph analysis are provided. Some embodiments apply machine learning to the analysis of unstructured text to identify regulatory, privacy, and cybersecurity risks. In some embodiments, a query string is provided to identify desired content in the unstructured text. The query string is tokenized and the tokenized parts are converted into nodes of a query graph and relationships are defining between the nodes. After identifying matches between terms in target text with nodes of the query graph, a graph is generated using matched terms as nodes. Based on the defined relationships, a determination is made whether any group of nodes on the graph satisfy the query string. Any matched nodes are mapped to the target text, and this information is provided to an analyst or reviewer for further consideration or action.

Abstract: The disclosed invention provides methods supporting systems that processes video communication(s) provided by an organization subject to internal or external regulations such as a specific financial services firm or a specific consumer product goods company. In various embodiments, video content is analyzed through a pipeline process of multiple extraction scripts, programs, and/or services and processes the content for participant likeness through facial recognition, voice recognition, and voice-to-text and stores the output in a normalized, searchable data store. A policy analysis is performed based on the stored data and external regulatory compliance, internal rules such as acceptable use policies, community-based regulations, or any combination thereof, to determine a risk score.

Abstract: The disclosed invention provides methods supporting systems that processes video communication(s) provided by an organization subject to internal or external regulations such as a specific financial services firm or a specific consumer product goods company. In various embodiments, video content is analyzed through a pipeline process of multiple extraction scripts, programs, and/or services and processes the content for participant likeness through facial recognition, voice recognition, and voice-to-text and stores the output in a normalized, searchable data store. A policy analysis is performed based on the stored data and external regulatory compliance, internal rules such as acceptable use policies, community-based regulations, or any combination thereof, to determine a risk score.

Abstract: Systems and methods for detecting and using semantic relatedness to classify segments of digital text are disclosed. More particularly, embodiments determine the semantic relatedness of segments of text to abstract categories where the abstract categories are not defined by a single word or semantic concept. Detecting semantic relatedness includes analyzing text, embedding the text, and determining semantic relatedness to a set of concepts for a category where each concept may include a set of words/phrases embedded in a similar fashion. The text embedding can be projected onto each concept embedding and reduced to a score representing semantic relatedness. The text is classified based on the semantic relatedness.

Abstract: Methods of and systems for identifying a discrete participant among a plurality of participants in an audio/video communication (e.g., video content, digital video content, audio content, and audio-visual content) that include: receiving the audio/video communication; identifying from the audio/video communication content (e.g., using metadata from the broadcast content to identify each participant) multiple identification-related features for each participant; associating a first identification-related feature of a first participant to a second identification-related feature of the first participant; organizing the associated identification-related features of the first participant into a feature class(es); logically relating a first feature class to a second feature class; and attributing a grouping of logically-related feature classes to the discrete participant.

Abstract: Methods of and systems for identifying a discrete participant among a plurality of participants in an audio/video communication (e.g., video content, digital video content, audio content, and audio-visual content) that include: receiving the audio/video communication; identifying from the audio/video communication content (e.g., using metadata from the broadcast content to identify each participant) multiple identification-related features for each participant; associating a first identification-related feature of a first participant to a second identification-related feature of the first participant; organizing the associated identification-related features of the first participant into a feature class(es); logically relating a first feature class to a second feature class; and attributing a grouping of logically-related feature classes to the discrete participant.