Abstract: Methods and systems for beam search decoding. One of the methods includes initializing beam data specifying a set of k candidate output sequences and a respective total score for each of the candidate output sequences; updating the beam data at each of a plurality of decoding steps, comprising, at each decoding step: generating a score distribution that comprises a respective score for each token in the vocabulary; identifying a plurality of expanded sequences; generating, for each expanded sequence, a respective backwards-looking score; generating, for each expanded sequence, a respective forward-looking score; computing, for each expanded sequence, a respective total score from the respective forward-looking score for the expanded sequence and the respective backwards-looking score for the expanded sequence; and updating the set of k candidate output sequences using the respective total scores for the expanded sequences.

Abstract: A system and method are provided for supporting dense retrieval of a template (e.g., a document template) for responding to a query or other textual input. The templates and past queries that were responded to using the templates are stored. A machine-learning model for matching a new query to the most appropriate template is trained using a selected subset of the stored queries as training queries. For each of one or more training batches or phases, multiple stored templates are selected (e.g., randomly) then, from among all training queries that the selected templates were used for, the same number of queries are selected (e.g., randomly), such that they represent the distribution of the training queries among the selected templates. A unique loss function is computed that leverages similarities and differences not only between each selected training query and each selected template, but also between different queries and between different templates.

Abstract: A device is disclosed for source knowledge generation. The device identifies external systems storing API documentation and crawls the external systems to extract documents representing information describing APIs for accessing the one or more data source systems. The device generates a plurality of vector representations by, for each document, providing the document to a large language model to generate a vector representation of the document and generates an index that maps each vector representation to its respective document. The device extracts information describing a data connector by querying the index, the query corresponding to a particular type of data source, and generates a connector representation based on the information describing the data connector extracted from the index, wherein the connector representation is used for generating and deploying a data connector for accessing data from a data source system of the particular type of data source.

Abstract: Techniques for handling natural language query processing are described. In some examples, a sequence-to-sequence model is used to handle a natural language query. Post-processing of a result of the sequence-to-sequence model utilizes fine-grained information from an entity linker. In some examples, the sequence-to-sequence model and aspects of a natural language query pipeline are used to handle a natural language query.

Abstract: A metafolder system manages a graph that identifies activities that are performed with respect to digital cards and that identifies relationships among the digital cards. The digital cards from a plurality of different locations in the de-centralized data storage system architecture can be aggregated for projection as coming from a single virtual location.

Abstract: Disclosed is a system and a method for semantic search in an enterprise. The method comprises defining entities in the enterprise, indexing content from one or more data sources of the enterprise to determine relationships between individual pieces of the content and one or more of the entities, providing an interface to allow a user to enter a search query, parsing the search query, determining relevant pieces of the content based on the parsed search query, identifying the one or more of the entities for each of the relevant pieces of content based on the relationships therebetween and configuring the interface to display each of the relevant pieces of the content under respective one or more of the entities in response to the search query.

Abstract: Architectures and techniques are described that can translate permissions from a first storage system to a second storage system. Such can be applied during a data migration operation that migrate storage data from the first storage system to the second storage system. A directory tree structure of the storage data can be explored to identify one or more target directories that can be used to leverage an inheritance property. Permissions can be translated based on the identification of the target directories.

Abstract: Some embodiments provide a program that receives a set of data for a first record in a first table. The set of data comprises a set of values for a set of attributes. In a data loading process configured to load a subset of the set of data into a subset of a first set of columns in the first table, the program determines that a first column in a first set of columns does not belong in the subset of the first set of columns. The program generates the first record in the first table. The program generates a value for the first column in the first set of columns that refers to a second record in the second table configured to represent a defined type of record. The program stores the value in the first column in the first set of columns of the first record.

Abstract: Committing of mappings and data lock transitioning associated with files can be performed and managed. Regarding a write operation to write a file portion to a node of a node network, network management component (NMC) can analyze write stream value associated with file portion and write stream-related index value associated with the file, where write stream-related index values can be associated with tracks of a track table. Based on the analyzing, NMC can determine whether a file access pattern associated with file portion is sequential or random; and, if sequential, whether there is an overflow condition associated with the track table. If sequential, and an overflow condition exists, NMC can commit, to data store, a mapping of another file portion associated with another stored stream value stored in track table, and/or transition exclusive data lock to shared data lock with respect to the other file portion.

Abstract: The present disclosure relates to a system and techniques for resolving dangling references resulting from a dependency relationship between computing resource objects uncovered during a harvesting process. In embodiments, a harvester application adds computing resource objects associated with a client to a resource collection as those computing resource objects are identified. Dependencies are identified as each computing resource object is added to the resource collection, which are resolved only if the computing resource objects associated with those dependencies have already been added to the resource collection. If the computing resource objects associated with the dependencies have not already been added to the resource collection, then the dependency is added to an observer pool. Observer modules are configured to check each computing resource object as it is processed during the harvest process in order to match those computing resource objects to unresolved dependencies.

Abstract: A method can include receiving a request for field equipment data; responsive to the request, automatically processing the field equipment data using a trained machine learning model to generate a quality score for the field equipment data; and outputting the quality score.

Abstract: A method and system of classifying documents is disclosed. The method includes determining line-text data for each of a plurality of lines of the document image using a text extraction technique. A set of unique keywords in the document image is determined based on a predefined list of keywords in the line-text data for each of the plurality of lines. For each keyword, two forward nodes are determined as next two subsequent keywords in the set of unique keywords based on determination of a shortest distance between the corresponding keyword and the next two subsequent keywords and based on the pre-defined reading sequence. Weights of each of the two forward nodes are determined based on the shortest distance and an angle of each of the two forward nodes with the corresponding keyword. A cluster is determined based on the feature matrix using a machine learning clustering model.

Abstract: Provided is an information system and an information collection method that are capable of acquiring desired site data while preventing an increase in cost. A link logic definition unit 17 holds a link logic that is definition information defining a correspondence relation between an item value of the site data and a reference value that is an external value of transaction data used for acquiring the item value. A link logic data acquisition unit 18 acquires the item value of the site data by using the link logic.

Abstract: A time series data injection method is provided, including: receiving time series data, where the time series data includes at least one parameter identifying a data source that generates the time series data, an indicator indicating at least one attribute of the data source, and a timestamp indicating a time point when the indicator is generated; storing a first parameter group in a row storage format, where the first parameter group includes the at least one parameter identifying the data source that generates the time series data; and storing a second parameter group in a columnar storage format, where the second parameter group includes the indicator indicating the at least one attribute of the data source and the timestamp. In the method, the time series data is injected in a hybrid row-column storage manner.

Abstract: Systems, controllers and methods for contextual-based searching are provided. A system includes one or more devices configured to collect at least one audio stream and a contextual search system. The contextual search system includes a data analyzer and a search engine. The data analyzer is configured to receive the at least one audio stream from among the one or more devices and to determine contextual information from the received at least one audio stream. The search engine is configured to perform a search of at least one search term using the contextual information, to produce a search result.

Abstract: Metric time series (MTS) data objects stored within in-memory storage are marked as inactive in response to determining that no MTS data has been received for the MTS objects within a first predetermined time period. In response to determining that an MTS object has been inactive for longer than a second predetermined time period, the MTS data object is migrated from in-memory storage to on-disk storage. Queries directed to MTS objects are first run against MTS object data stored within in-memory storage, and then against MTS object data stored within on-disk storage. In this way, an amount of in-memory storage needed to store MTS objects may be minimized, while optimizing search performance.

Abstract: Fenestration units and method of operation thereof are provided that allow for use of the fenestration units as nodes of a distributed ledger network. The fenestration units include a communication device configured for communication with at least one remote electronic device, a non-transitory, non-volatile memory configured to store computer readable instructions, and a processor in operable communication with the communication device and the memory. The processor is configured to execute the computer readable instructions to function, through the communication device, as a node in a distributed ledger network by storing a local copy of a distributed ledger on the memory, updating the local copy of the distributed ledger, and distributing the local copy to the at least one remote electronic device in response to updating the local copy of the distributed ledger.

Abstract: A computer-implemented method accelerates cognitive intelligence queries to a database using semantic keys. A computing device receives a database. The computing device inserts one or more key-value pairs into the database, where the key is a semantic key that is generated from a binary codeword built from a locality sensitive hashing of one or more vectors in a database embedding model of the database, and where the value is a tuple in the database that identifies entries in the database that share predefined features. The computing device uses the one or more key-value pairs for accelerating cognitive intelligence queries to the database.

Abstract: Systems and methods are disclosed for aggregating informational reports. In one implementation, at least one processor may be programmed to receive an informational vehicle report identifying a detected event; store the report in a database in association with a first cell; query a second cell within a predetermined distance of the first cell; and determine whether the second cell is associated with the detected event. When the second cell is associated with the detected event the processor may aggregate information from the first and second cells to provide an aggregated cluster and generate an event report based on the aggregated cluster. When the second cell is not associated with an information cluster associated with the detected event, the processor may generate the event report based on the stored informational vehicle report. The processor may then transmit the event report to one or more vehicles.

Abstract: A method for training a machine learning model using self-contrastive decorrelation is provided. The method comprises training a machine learning model by receiving a sentence including text, performing a first encoding operation on the sentence, performing a second encoding operation on the sentence, mapping the first vector representation on which a first augmentation operation is performed to a first high dimensional vector representation and the second vector representation on which a first augmentation operation is performed to a second high dimensional vector representation, generating a correlation matrix using the first high dimensional vector representation and the second high dimensional vector representation, and performing a decorrelation operation on the correlation matrix.