Abstract: Techniques are provided for updating in-memory property graphs in a fast manner, while minimizing memory consumption. A graph is represented as delta compressed sparse rows (CSR), in which its data structure stores forward edge offsets that map reverse edges to forward edges, enabling fast traversals of graph edges in forward and reverse directions. To support fast graph updates, delta logs are used to store changes to the graph. In an embodiment, a base version of the graph data structure is initially loaded or created, and subsequent versions of the graph are created from the reference to the initial graph and a delta log data structure that records the changes compared to the base version of the graph.

Abstract: A method of configuring a search engine to classify a search query includes receiving a search query data set, the search query data set comprising a plurality of search queries, defining a first set of candidate labels and a second set of candidate labels according to the search queries in the search query data set, concatenating the first set of candidate labels with the second set of candidate labels to generate a concatenated candidate label set, generating a compatibility matrix comprising a similarity between the concatenated candidate label set and the search query data set, and training a classification network according to the compatibility matrix.

Abstract: A data classification system obtains a data set including first data associated with a first characteristic and second data associated with a second characteristic. In response to obtaining the data set, the system uses a classification model to generate a classification by classifying the first data into majority data and the second data into minority data. From the classification and using the classification model, the system determines a loss. The system modifies the classification model based on this loss to generate an updated classification model. The system uses the updated classification model to re-classify a subset of the first data into the minority data and a subset of the second data into the majority data.

Abstract: Systems, methods, and computer-readable media for computer-assisted output validation in machine learning/artificial intelligence platforms are disclosed. An application instance includes one or more machine learning models used to generate searchable data structures based on multimodal inputs.

Abstract: A method includes applying a document ranking layer of a document selection large language model (LLM) to a document list including multiple reference documents to obtain a ranked document list. The method further includes selecting a subset of reference documents from the ranked document list and processing a user prompt and the document subset by a field LLM to generate an answer. The method further includes ranking the answer with an answer score by a ranking LLM. The method further includes ranking the document subset by the ranking LLM to obtain a ranked document subset. The method further includes calculating a loss function of a preference optimization layer of the document selection LLM based on the answer score and updating at least one training parameter of a foundation model of the document selection LLM based on the loss function of the preference optimization layer.

Abstract: Systems and methods of persistent file protection integrate cloud-based backups with EDR file protection. Use of local persistent file cache (PFC) data with cloud-stored file segments, facilitates reconstruction of file states from multiple sources. File storage and retrieval efficiency are optimized, even when local session caches are invalidated or when extensive file modifications exceed storage quotas.

Abstract: A computer system for executing a processing graph including: (A) accessing a graph definition of the processing graph from a nodes database, each node of the processing graph representing a component of computation performed by at least one program referenced by that node; (B) constructing the processing graph in the at least one memory device; (C) identifying at least one group of nodes in the processing graph; (D) traversing the at least one group of nodes; (E) upon visiting a node of the at least one group of nodes, accessing a node definition of the node from the nodes database, the node definition identifying a program referenced by the node; (F) executing the program to generate an output; and (G) passing the output to a parent node of the node, the parent node using the output during processing of the parent node.

Abstract: The present disclosure relates to a data processing method, a medium and an electronic device for a database. The method includes: obtaining a first execution statement for performing a target operation; reading, when a data source corresponding to the first execution statement includes first data corresponding to a committed transaction and second data corresponding to an uncommitted transaction, a materialized view obtained after performing the target operation on the first data as a target materialized view from a pre-created materialized view set; reading the second data and performing the target operation on the second data to obtain operation result data; performing data merging processing on the target materialized view and the operation result data to obtain merged data; and performing the target operation on the merged data to obtain an execution result corresponding to the first execution statement.

Abstract: Some implementations relate to methods, systems, and computer-readable media to search for and retrieve three-dimensional (3D) objects based on user queries. An exemplary method comprises receiving a first query for one or more 3D objects. At least one constraint is identified from the first query, wherein the at least one constraint includes a particular descriptor that specifies one or more aspects of matching 3D objects. A second query is generated comprising a set of descriptors comprising the particular descriptor. A search is then performed, using the second query, to search a data repository containing 3D content, to obtain one or more 3D object search results having features responsive to the first query. A response that includes one or more 3D object search results may then be outputted as a response to the first query.

Abstract: A system and method to identify a first data stream from a data graph associated with a merchant system. A configuration file associated with a publisher system to receive data associated with the merchant system is identified. Based on a configuration file associated with the publisher system, a second data stream is generated which includes the data associated with the merchant system. The second data stream is transmitted to the publisher system.

Abstract: A computing system receives a selection of a parameter-setting rule associated with a specific unique identifier and automatically assigns data parameters to the data records according to the parameter-assigning rule. The computing system determines a domain for each data record based on data parameters associated with the occurred instance. A collection of data records is displayed to a user of the computing system. The collection of data records is displayed with a set of third-party defined fields. Values for the set of third-party defined fields are pre-filled based on the parameter-assigning rules. Upon receiving the request to export the collection of data records, a payload corresponding to the collection of data records for the third-party system is created. The payload includes at least the set of third-party defined fields and the domain associated with each data record automatically assigns data records to the corresponding domain in the third-party system.

Abstract: Systems and methods are provided for storing a first data object comprising a first set of immutable components, the first data object being associated with a corresponding second data object stored by a remote replication system. A difference is determined between the first set of immutable components of the first data object and a second set of immutable components of the corresponding second data object. A subset of immutable components is identified from the first set of immutable components based on the difference. The subset of immutable components from the first set of immutable components is provided to the remote replication system over a communication network.

Abstract: In an example embodiment, a solution for matching entities in a query table with one or more entities in a target table, in the presence of a constraint on the value sum of the matching targets (hereinafter called the “value constraint”), using machine learning techniques, is provided. Specifically, a non-linear objective function is converted to a linear objective function and a machine learning model is trained using the linear objective function, allowing for the use of solver functions from libraries in order to speed matching over existing methods.

Abstract: A method includes receiving a domain search query from a first user, generating a prompt based on the query to instruct a first model to provide a domain recommendation, where the first model is trained by utilizing a dataset including: first data reflecting interactions of users with domain portfolios, second data reflecting domain-related actions performed by the users, third data reflecting behaviors of a second user during a current domain search session, fourth data reflecting behaviors of the second user over a predetermined period covering multiple search sessions, and fifth data reflecting search scenarios of multiple users, determining the domain recommendation against a predetermined rule to form a sanitized domain recommendation, verifying an availability to register the sanitized domain recommendation to form an available domain recommendation, adding industry information to the available domain recommendation to form an enriched domain recommendation, and transmitting the enriched domain recommen

Abstract: Methods and systems described in this disclosure are directed to electronically generating narratives and tracking a user's actions in responses to receiving the narratives. The disclosed system can rank narratives for effectiveness of a narrative based on the responses. In some embodiments, the narratives can be classified into categories based on criteria pertaining to user population data or user-specific data.

Abstract: The present disclosure relates to maintaining extracted data in a group node profile from electronic activities. A group node profile associated with a group entity is maintained. A first data point including an electronic activity or a record object is accessed. The first data point is parsed using a value detection policy to obtain a first value. The group node profile is updated based on the first value. An association score between the group node profile and the first value is generated. A second data point is accessed. The second data point is parsed. The association score between the group node profile and the first value is updated.

Abstract: Systems and methods herein describe embodiments for handling a data drift. An example system accesses the data pipeline, which is comprised of a plurality of stages. For each stage of the plurality of stages in the data pipeline, the system identifies stage schema fields for processing data in the data pipeline and generates a set of stage schema fields comprising the identified stage schema fields in the stage. In response to detecting an origin stage, the system generates a set of pipeline schema fields, whereby the set of pipeline schema fields comprise a union of the generated sets of stage schema fields. The set of pipeline schema fields are then stored.

Abstract: Disclosed is a system to provide information associated with a user device from a server to the user device without disclosing the association to the server. Data structures, created at the server, contain information such as information about restaurants, mechanics, medical conditions, etc. The server creates bundles including two or more data structures containing disparate information, and a unique identifier for each bundle. The server creates a table of contents indicating the unique identifier of a bundle and the information contained in the bundle and sends the table of contents to the device. The server receives the unique identifier of the bundle and provides the bundle having the unique identifier to the device. While the bundle contains the information associated with the user device, because the bundle contains disparate information. the server does not know which information is associated with the user device.

Abstract: A method for labeling data items includes receiving, from a client, user input generated by a user of the client. The method also includes generating query conditions based on the user input. Further, the method includes retrieving, from a custom knowledge system integration module and using the query conditions, a list of label candidates. In addition, the method includes determining, via the client, that the user approves the list of label candidates. Moreover, the method includes generating, based on the determination, a label candidate query using at least the list of label candidates. Further, the method includes submitting the label candidate query to a large language model (LLM) engine. Moreover, the method includes obtaining, in response to the label candidate query, a list of target labels from the LLM engine. Also, the method includes labeling a plurality of unlabeled data items using the list of target labels.

Abstract: A database system comprised of a decoupled compute layer and storage layer is implemented to store, build, and maintain a canonical dataset, a temporary buffer, and projection datasets. The canonical dataset is a set of batch updated data. The data is appended in chunks to the canonical dataset such that the canonical dataset becomes a historical dataset over time. The buffer is a write ahead log that contains the most recent chunks of data and provides atomicity and durability for the database system. The projection datasets are indexes of the canonical dataset and/or the buffer that may have single or multiple column sort-orders and/or particular data formats. The writes to the canonical dataset, projection datasets, and buffer may be asynchronous and therefore the database system is advantageously less resource constrained.