Abstract: An algorithm releases answers to very large numbers of statistical queries, e.g., k-way marginals, subject to differential privacy. The algorithm answers queries on a private dataset using simple perturbation, and then attempts to find a synthetic dataset that most closely matches the noisy answers. The algorithm uses a continuous relaxation of the synthetic dataset domain which makes the projection loss differentiable, and allows the use of efficient machine learning optimization techniques and tooling. Rather than answering all queries up front, the algorithm makes judicious use of a privacy budget by iteratively and adaptively finding queries for which relaxed synthetic data has high error, and then repeating the projection. The algorithm is effective across a range of parameters and datasets, especially when a privacy budget is small or a query class is large.

Abstract: Systems, methods, and computer readable media for staging a corpus of electronic communication documents for analysis, such as, for example, via a content analysis platform. The staging may include a staging platform accessing the corpus of electronic communication document. For each electronic communication document within the corpus, the staging platform may generate a fingerprint based upon the output of a hash function executed upon a set of characteristics corresponding to each segment within the electronic communication document. The staging platform may analyze the generated fingerprints to generated a plurality of threaded conversations that do not include electronic communication documents that fail to convey any new information. The systems and methods may also include detecting and flagging any segments within an electronic communication document that may have been mutated by its author.

Abstract: A system is disclosed. The system includes a processor configured to: receive a set of data structured in a schemaless data representation; automatically translate the set of data into a relational representation by: translating an array map value in the set of data into an ordered multi-map; and converting the ordered multi-map to the relational representation. The processor is further configured to store the translated set of data in a key-value data store for a query-based retrieval.

Abstract: A computer-implemented method of executing a policy-based operation on a shared computer infrastructure includes storing in a computer memory a dynamically extensible metadata system that is in communication with a processor that executes policy-based operations, where the dynamically extensible metadata system includes a data structure, a collection method, a policy processing method, and a policy condition. The collection method is then executed to collect data from a first computer resource in the shared computer infrastructure using a first data structure and from a second computer resource in the shared computer infrastructure using a second data structure, where the first data structure and the second data structure are different data structures. The collected data is then processed with the policy processing method to determine if the collected data meets the policy condition.

Abstract: A computer-implemented method, a computer program product, and a computer system for detecting an inappropriate data type of a column in a database and building a physical access path over a correct data type. The computer system detects in a table a candidate column with a first data type that has a mismatching type definition, using database usage statistics. The computer system determines whether it is possible to build an additional index as an access path over values with a second data type. The computer system, in response to determining that it is possible to build the additional index, converts values in the candidate column to the values with the second data type. The computer system builds the additional index over the values with the second data type in the table. The computer system generates a query plan operator for the additional index.

Abstract: Methods, computer program products, computer systems, and the like are disclosed that provide for scalable deduplication in an efficient and effective manner. For example, such methods, computer program products, and computer systems can include tracking one or more write operations executed on a target data store and sending metadata regarding the one or more write operations to a source site. The tracking comprises storing information regarding the one or more write operations in a data structure. The one or more write operations cause one or more units of data to be written to the target data store. The target data store is at a target site. The metadata comprises the information.

Abstract: A method, apparatus, and system for storing indexed metadata associated with backed up data is provided. The operations comprising: for each first data item stored at a first client device to be backed up, extracting, at the first client device, an associated first metadata item; transforming, at the first client device, the first metadata item based on a predefined metadata definition; associating, at the first client device, a first client identifier and a backup identifier with the transformed first metadata item; and transferring the transformed first metadata item from the first client device to a metadata bus of a search cluster.

Abstract: Disclosed herein are systems, methods, and processes to optimize the storage of pre-allocated units of storage during a backup operation. Null units of storage are identified in pre-allocated units of storage prior to the backup operation. Upon being identified, the null units of storage are inhibited from being written to a backup image during the backup operation.

Abstract: A processing system including at least one processor may be maintain an historical graph database storing historical graph state information for a graph including a set of elements, where the historical graph state information includes, for each element, respective element state information and respective state version tracking information for the respective element. The processing system may receive an event indicative of a management action associated with a first element, where the first element may include a new element to be added to the graph or an existing element of the set of elements. The processing system may update, based on the event, the historical graph state information of the historical graph database, where the updating may include providing respective state version tracking information for the first element, and handle, based on the historical graph database, a query to retrieve a set of information of the graph.

Abstract: One or more techniques and/or computing devices are provided for non-disruptively establishing a synchronous replication relationship between a primary volume and a secondary volume and/or for resynchronizing the primary volume and the secondary volume. For example, a baseline snapshot and one or more incremental snapshots of the primary volume are used to construct and incrementally update the secondary volume with data from the primary volume. A dirty region log is used to track modifications to the primary volume. A splitter object is used to split client write requests to the primary volume and to the secondary volume. A synchronous transfer engine session is initiated to processing incoming client write requests using the dirty region log. A cutover scanner is used to transfer dirty data from the primary volume to the secondary volume. In this way, a synchronous replication relationship is established between the primary volume and the secondary volume.

Abstract: A method comprises collecting, by a computing device located at an edge of a network, data items corresponding to information transmitted by endpoints using the network, generating, by the computing device, a probabilistic hierarchy using the data items, generating, by the computing device using the probabilistic hierarchy and natural language data, a similarity metric, generating, by the computing device using the probabilistic hierarchy, the natural language data, and the similarity metric, an ontology, detecting, by the computing device using the ontology, an anomaly, and in response to detecting the anomaly, sending a notification.

Abstract: Techniques for searching documents are described. An exemplary method includes receiving a document search query; querying at least one index based upon the document search query to identify matching data; fetching the identified matched data; determining one or more of a top ranked passage and top ranked documents from the set of documents based upon one or more invocations of one or more machine learning models based at least on the fetched identified matched data and the document search query; and returning one or more of the top ranked passage and the proper subset of documents.

Abstract: A computing system may store a first file in a system repository that is to be synchronized with at least a first repository associated with a first client device and a second repository associated with a second client device. The computing system may determine that the first repository is to receive a downloaded copy of at least a portion of the first file, and may cause a copy of the at least a portion of the first file to be downloaded from the system repository to the first repository. Based at least in part on the computing system determining that the first repository is to receive the downloaded copy, an identifier of the at least a portion of first file may be stored in the second repository without storing a copy of the at least a portion of the first file in the second repository.

Abstract: Disclosed herein are systems and methods for database replication to a remote deployment with automated fulfillment. In an embodiment, a data platform detects receipt of a database-replication request associated with a data consumer, requesting availability of a local instance of a primary-deployment database in a remote-deployment account of a data provider at a remote deployment of the data platform. The data platform responsively performs remote-deployment-activation operations including provisioning the remote-deployment account with a set of one or more replication-preparation objects associated with the primary-deployment database, and also including invoking at least one refresh command with respect to the set of one or more replication-preparation objects, which makes available to the data consumer a local instance of the primary-deployment database in the remote-deployment account of the data provider at the remote deployment of the data platform.

Abstract: Various embodiments herein each include at least one of systems, methods, and software for generating, storing, and using non-unique secondary indexes of semi-structured data in database tables. One method embodiment includes creating and storing a non-unique secondary index (NUSI) for a database table based on column data stored in rows of the database table. The NUSI in some such embodiments identifying an offset to specific data included in the column data of at least some rows of the database table where the specific data is located. Other embodiments include building a NUSI index that includes values embedded therein.

Abstract: Efficient rolling transactions for a data storage system is described herein. A method as described herein can include structuring, by a device operatively coupled to a processor, a data stream according to a first structure, the first structure comprising respective stream segments that are associated with respective first key ranges; writing, by the device, a transaction to the data stream, wherein the transaction comprises events arranged in a second structure that is distinct from the first structure, and wherein the second structure comprises respective transaction segments that are associated with respective second key ranges; and merging, by the device, respective ones of the transaction segments into respective ones of the stream segments, resulting in merged transaction segments, in response to the second key ranges of the respective merged transaction segments at least partially matching the first key ranges of the respective ones of the stream segments.

Abstract: In accordance with an embodiment, described herein is a system and method for delivery of content based on matching of user profiles with content metadata. The system enables delivery of personalized content, without the overhead of managing segment targeting rules, while providing content publishers or marketers with complete control over such personalization. A recommendation service or application program interface, provided by a computer, cloud computing environment, or other type of computer system, enables receipt and processing of requests, from client devices, for personalized content. A recommendation engine delivers content assets in response to a request from a client device. The recommendation engine determines a content channel and a user identity associated with the request, and then delivers content assets based on rules governing the matching of content asset metadata with the user profile. While content classification evolves over time, so also does the personalization of delivered content.

Abstract: A non-transitory computer-readable medium stores program code including instructions that, when executed by one or more processors, cause the one or processors to perform operations including storing and managing social network data in a social network database; receiving location information from a plurality of electronic devices associated with the social network data; specifying users of electronic devices present in a specific space at a specific time among the plurality of electronic devices; filtering relation information between the specified users in the social network database; and calculating an analysis index for the specific space and the specific time based on the filtered relation information.

Abstract: Systems and methods are described for processing records associated with a query that identifies an association between two data fields. The system can obtain a chunk of data that includes multiple records based on a query received by a data intake and query system. At least one record can include multiple sub-records that share a field value for at least one field. The system can generate a record from each sub-record and assign the generated records to one or more groups of partitions. The system can combine record data of generated records assigned to one partition of a group of partitions and then combine record data across the group of partitions. The system can process the results of the combination of records across the group of partitions based on the query.

Abstract: Methods, systems, and computer-readable media for linking multiple data entities. The method collects a snapshot of data from one or more data sources and converts it into a canonical representation of records expressing relationships between data elements in the records. The method next cleans the records to generate output data of entities by grouping chunks of records using a machine learning model. The method next ingests the output data of entities to generate a versioned data store of the entities and optimizes versioned data store for real-time data lookup. The method then receives a request for data pertaining to a real-world entity and presenting relevant data from the versioned data store of entities.