Abstract: An unsupervised multi-hop search across a corpus of documents in a database or other data resource permits the identification of relationships between two entities mentioned in the corpus in cases where the two entities are not co-mentioned within any documents in the corpus (or not mentioned within document(s) with sufficient frequency or proximity to infer the relationship). The search can employ a beam search algorithm anchored by word embeddings and an A* graph traversal to calculate semantic distance between the entities as different paths through the corpus for different entity co-mentions are evaluated.

Abstract: A message batching configuration may be determined for transmitting a message to recipients. The message batching configuration may include two or more message batches, a respective recipient count for each message batch, a respective time delay between each message batch, and a performance metric for evaluating the message. The message is transmitted in accordance with the message batching configuration. The transmission of subsequent message batches is halted when it is determined that the designated performance metric fails to meet a designated performance metric threshold.

Abstract: An entity-based platform stores a database of entity mentions and related, entity-based information obtained by processing text in a corpus of documents such as content from a public network or content from a document management system. The platform may analyze changes to a previously-processed document in order to align mentions in the modified document to mentions in a prior document and/or to identify new mentions not present in the previously-processed document. As the underlying documents in the corpus change over time, this advantageously permits sustained use of knowledge graphs and other entity-based representations or analysis linked to specific entity mentions.

Abstract: Sentiment analysis is a task in natural language processing. The embodiments are directed to using a generative language model to extract an aspect term, aspect category and their corresponding polarities. The generative language model may be trained as a single, joint, and multi-task model. The single-task generative language model determines a term polarity from the aspect term in the sentence or a category polarity from an aspect category in the sentence. The joint-task generative language model determines both the aspect term and the term polarity or the aspect category and the category polarity. The multi-task generative language model determines the aspect term, term polarity, aspect category and category polarity of the sentence.

Abstract: Artificial intelligence (AI) technology can be used in combination with composable communication goal statements to facilitate a user's ability to quickly structure story outlines in a manner usable by an NLG narrative generation system without any need for the user to directly author computer code. Narrative analytics that are linked to communication goal statements can employ a conditional outcome framework that allows the content and structure of resulting narratives to intelligently adapt as a function of the nature of the data under consideration. This AI technology permits NLG systems to determine the appropriate content for inclusion in a narrative story about a data set in a manner that will satisfy a desired communication goal.

Abstract: Systems, methods, and related technologies are provided for generating interactive emails and tracking user interactions. In one implementation, an email can be received within an email client. The email can be rendered within the email client, such as by: presenting a first selectable element within the email client, presenting a second selectable element within the email client, and presenting one or more content items within the email client. A selection of the first selectable element can be received within the email client. In response to the selection of the first selectable element, a presentation of the one or more content items can be adjusted within the email client.

Abstract: Disclosed are some implementations of systems, apparatus, methods and computer program products for implementing a shim driver configurable to provide a database driver. A shim driver class is loaded from a class path and instantiated such that a shim driver is generated. A directory path and database driver class name are obtained from a configuration file. An application transmits a request for a database driver via an application programming interface (API) of the shim driver, where the request includes the directory path and database driver class name. The shim driver loads a database driver class identified by the database driver class name using the directory path, instantiates the database driver class such that a first database driver is generated, and provides the first database driver responsive to the request.

Abstract: Embodiments described herein provide an approach (referred to as “Co-training” mechanism throughout this disclosure) that jointly learns two representations of the training data, their class probabilities and low-dimensional embeddings. Specifically, two representations of each image sample are generated: a class probability produced by the classification head and a low-dimensional embedding produced by the projection head. The classification head is trained using memory-smoothed pseudo-labels, where pseudo-labels are smoothed by aggregating information from nearby samples in the embedding space. The projection head is trained using contrastive learning on a pseudo-label graph, where samples with similar pseudo-labels are encouraged to have similar embeddings.

Abstract: Techniques are described herein for a method of obtaining a token based on a conversation in real time. The method further includes predicting, using a large language model (LLM) and the token, a next token. The method further includes predicting, using a classifier and the next token, a completion of a user turn. The method further includes triggering a next turn of the conversation in real time using the completion of the user turn.

Abstract: In some embodiments, a method stores a total number of generative credits for a generative artificial intelligence (AI) solution that is integrated with a software application in a database system. Usage data is tracked for a request to the generative artificial intelligence (AI) solution in the database system. The method determines a context from the usage data and retrieves a contextual pricing model for the generative AI solution using the context. The contextual pricing model translates a model specific charging policy to generative credits. The method applies the usage data to the contextual pricing model to translate the usage data to a number of generative credits. The number of generative credits for the generative AI solution is applied to an available number of generative credits of the total number of generative credits to generate a new available number of generative credits.

Abstract: A method and apparatus for avoiding exposure of data that was entered into a web application in the context of having that data scanned by a third-party web application firewall (WAF). The apparatus receives, in response to forwarding a request to a web application, a response from the web application that includes a web form. The apparatus forwards the response to a browser and in response receives a second request that includes a version of the web form modified by the browser such that data entered into a plurality of fields of the web form was extracted, aggregated, and homomorphically encrypted. The apparatus causes the second request to be scanned by the third party WAF using one or more homomorphic encryption signatures, and in response to the scanning determines whether to forward the second request to the web application or take corrective action.

Abstract: Database systems and methods are provided for managing usage of large language models (LLMs). One method involves determining a numerical representation of a conversational input to a user interface, identifying a semantically similar subset of prior conversational inputs based at least in part on the numerical representation of the conversational input, and determining numerical representations of respective conversational responses generated by a language model responsive to the respective prior conversational input of the semantically similar subset. When the numerical representations of the respective conversational responses satisfy a semantic similarity threshold, the method automatically generates an automated response to the conversational input based at least in part on one or more prior conversational responses and automatically provides the automated response to the user interface responsive to the conversational input.

Abstract: A user device (e.g., a computing device, a smart device, a mobile device, a laptop, a tablet, a set-top box, a display device, etc.) may generate, based on a request for conversational history information, a timeline for a user interface comprising an interactive timeline indicator indicating a conversation that occurred during a time window. The user device may display, based on an interaction with the interactive timeline indicator, a graphical representation of the time window that indicates a circular order of occurrence for the conversation and other conversations that occurred during the time window. The user device may display the conversation based on an interaction with an interactive element that indicates when within the circular order of occurrence the conversation occurred.

Abstract: Provided herein are system, method, and/or computer program product embodiments and/or combinations and sub-combinations thereof for updating configuration settings during the execution of an application. In a given embodiment, a central system may read a configuration value for a configuration setting of an application from a configuration file. The central system may assign the configuration value to a variable corresponding to the configuration setting. The central system may execute an operation on the variable corresponding to the configuration setting to change the variable to a new configuration value, independent of multiple threads executed by the application, while the application is being executed, and independent of the value of the configuration setting in the configuration file.

Abstract: Embodiments described herein provide systems and methods for a partially supervised training model for online action detection. Specifically, the online action detection framework may include two modules that are trained jointly—a Temporal Proposal Generator (TPG) and an Online Action Recognizer (OAR). In the training phase, OAR performs both online per-frame action recognition and start point detection. At the same time, TPG generates class-wise temporal action proposals serving as noisy supervisions for OAR. TPG is then optimized with the video-level annotations. In this way, the online action detection framework can be trained with video-category labels only without pre-annotated segment-level boundary labels.

Abstract: Disclosed are methods, apparatus, systems, and computer readable storage media for interacting with components across different domains in a single user interface in an online social network. The user interface includes a first component and a second component, where the first component exposes content from a first database system at a first network domain and the second component exposes content from a second database system at a second network domain. A first interaction with the first component is received at a computing device, followed by a reference being provided in the second component, where the reference includes information related to the first interaction. A second interaction with the second component regarding the reference can be received at the computing device. Interactions between the components hosted on different database systems can occur through an application programming interface (API).

Abstract: A method for replicating a set of parent resources from an administrator namespace to a set of tenant namespaces is described. The method includes receiving, by the administrator namespace, a global object that includes a set of object fields that reference (1) a set of parent resources and (2) the set of tenant namespaces; monitoring, by an operator controller of the administrator namespace, the global object to determine whether a change has been made to the global object; and replicating, by the operator controller, the set of parent resources to the set of tenant namespaces as child resources based on the global object in response to detecting a change to the global object.

Abstract: Techniques are disclosed relating to automating authentication decisions for a multi-factor authentication scheme based on computer learning. In disclosed embodiments, a mobile device receives a first request corresponding to a factor in a first multi-factor authentication procedure. Based on user input approving or denying the first request, the mobile device sends a response to the first request and stores values of multiple parameters associated with the first request. The mobile device receives a second request corresponding to a factor in a second multi-factor authentication procedure where the second request is for authentication for a different account than the first request. The mobile device automatically generates an approval response to the second request based on performing a computer learning process on inputs that include values of multiple parameters for the second request and the stored values of the multiple parameters associated with the first request.