Abstract: A data processing platform generates visualizations for data streams to visually represent a portion of data in the data stream. The platform performs an analysis of a change in values of data contained in the data stream and generates, using a result of the analysis, metadata identifying an insight into the data in the data stream. The insight indicates a characteristic of the change in values. A natural language representation of the insight is generated using the metadata and output for display in association with the visualization.

Abstract: Systems and methods are provided for automatically recommending content. One example method includes receiving, from a computing device, a request for content and generating an initial result set in response to the request. A user profile associated with the request is identified. Based on the initial result set and the user profile, a positivity index score for the initial result set is determined. Based on the positivity index score, a type of additional content to add to the initial result set is determined. A modified result set is generated by adding the additional content to the initial result set. At least a portion of the modified result set is transmitted to the computing device.

Abstract: In a data preparation process, raw data in a database is classified into first-type data and second-type data. In a data retrieval process, a first retrieval range is determined in the first-type data, and to-be-retrieved data is retrieved in the first retrieval range, to obtain a first retrieval result; the to-be-retrieved data is retrieved from all of the second-type data, to obtain a second retrieval result; and finally, a final retrieval result is determined from the first retrieval result and the second retrieval result.

Abstract: An electronic apparatus includes a memory storing a plurality of pieces of data by a distributed data storing method; a communication interface configured to be communicatively connected to an external apparatus; and a processor configured to: based on identifying that first data corresponding to the external apparatus among the plurality of pieces of data stored in the memory has been changed to second data, change a state of the second data to a locked state; and control the communication interface to transmit the second data in the locked state to the external apparatus by a streaming method, wherein in the locked state the second data cannot be changed in response to a data change command.

Abstract: Methods and systems for connecting data with non-standardized schemas in connected graph data exchanges. For example, the system generates a custom data structure corresponding to a user identifier for a user profile that includes pointers between user profile attributes (e.g., individual fields/categories within the user profile) and existing assets in a connected graph (e.g., an existing application, software profile for an application, data set, connections, etc.). The system then connects the custom data structure corresponding to the user identifier to the existing assets in the connected graph.

Abstract: Embodiments autonomously test a logical model for inconsistencies. For example, metadata descriptive of a logical model can be received, where the logical model includes an abstraction for a database schema, the database schema is implemented at a database, and the database schema includes a fact table and a dimension table. Logical queries can be automatically generated including at least first and second logical queries based on the retrieved metadata, where the first and second logical queries target a logical object of the logical model. At least the first and second logical queries can be issued to a server that hosts the logical model, where, at the server, the first and second logical queries are translated to first and second database queries, and the first and second database queries target at least a fact table and a dimension table from the database schema. Query results received from execution of the first and second database queries can be compared.

Abstract: Denoising of technology logs and presentation of resulting user interfaces that display actions/log entries in ranked order of importance. Specifically, technology logs are received for a plurality of entities (e.g., users, devices, systems or the like). In response to receiving the technology logs, a rarity indicator is determined for the various recorded actions/entries in all of the received logs based on the number of occurrences in all of the received logs. Subsequently, an importance indicator is determined for each recorded action in a corresponding technology log based on the rarity indicator and the number of occurrences of the recorded action in the corresponding technology log. In response to determining the importance indicator, user interface(s) are generated and displayed that provide for the actions to be displayed in ranked order of their importance indicators.

Abstract: Aspects of the present disclosure relate to watermarks and watermarking techniques for data streaming pipelines. Time stamp and offset timeline data is shared by computing instances along the pipeline to enable improved watermarking of the data stream through the pipeline. The improved watermarks enable better determination of completeness for the data stream and improve materialization of the results. The watermarking techniques can include periodically publishing watermark data by processing units of a vertex, fetching a merged watermark for a vertex by a vertex, and/or watching a data storage for the watermark data for events. Consensus algorithms can be used to maintain consensus among vertices for the watermark data.

Abstract: Techniques are disclosed relating to maintaining a cache usable to locate data stored in a data structure. A computer system, in various embodiments, maintains a data structure having a plurality of levels that store files for a database. The files may include one or more records that each have a key and corresponding data. The computer system may also maintain a cache for the database whose entries store, for a key, an indication of a location of a corresponding record in a file of the data structure. In some embodiments, the computer system receives a request to access a particular record stored in the data structure where the request specifies a key usable to locate the particular record. The computer system may retrieve, from the cache via the key, a particular indication of a location of the particular record and may use the particular indication to access the particular record.

Abstract: Embodiments include technologies to apply at least one machine learning model to features of a search query, features of a searcher user, features of a searchee content item, and features of a searchee user, produce a first outcome prediction that represents a probability of a first objective relating to engagement of the searcher user with a content item in an online system and a second outcome prediction that represents a probability of a second objective relating to engagement of the searchee user with the online system responsive to the engagement of the searcher user with the content item, apply a multi-objective optimization solver to the first objective, the second objective and an outcome prediction that is a combination of the first outcome prediction and the second outcome prediction, and generate a serving function for a search engine based on the first objective, the second objective, and the outcome prediction.

Abstract: Many types of data files can only be stored in SQL databases as Blobs—Binary Large Objects, that the database can only handle as a whole. This limits searches within the Blob. Here improved methods are taught for automatically analyzing potential Blob files, and assigning at least certain classes of potential Blob files to General Data Type documents. In some cases, the potential Blob may be automatically analyzed and mapped to a JSON document type, and this JSON document type then automatically mapped into appropriate lists of JSON-document-nodes and tables that can be stored in the SQL database, and subsequently queried. This enables much data within the (to prior art) Blob to be queried at more detailed granularity. This data can then be retrieved and used to reconstruct some or all of the original General Data Type document, which would otherwise be a Blob to prior art methods.

Abstract: A method may include storing a vector space representation of a set of concepts in a semantic ontology, wherein locations of the set of concepts in the vector space are based on weights of a neural network trained using triples from the semantic ontology; presenting a semantic search user interface including: a text input portion to receive a search query from an input device; a suggested search portion; and a search results portion; receiving, from a user computing device, the search query; converting, using at least one processor, the search query into a vector; computing, using the at least one processor; a set of possible concepts related to the search query based on a distance between the vector and other concepts in the vector space representation; and presenting the set of possible concepts related to the search query in the suggested search portion of the semantic search user interface.

Abstract: The present application discloses a cache access method and an associated graph neural network system. The graph neural network processor is used for performing computation upon a graph neural network. The graph neural network is stored in the memory in compressed sparse row format. The method includes: receiving an address corresponding to a node of the graph neural network and a type of the address; in response to the type is one of a first type or a second type, performing lookup by comparing the address with a tag field of a degree lookup table to at least obtain a degree of the node; determining whether the degree is greater than a predetermined value to obtain a determination result; and determining whether to perform lookup on a region of the cache corresponding to the type according to the determination result.

Abstract: Provided are techniques for data search with automated selection of one or more AI inference models and inference label indexing. Metadata for a data object that is stored in a data source is retrieved. The metadata is matched to filtering criteria. The filtering criteria is used to identify an Artificial Intelligence (AI) inference model. The data object is sent to an inference engine for the AI inference model. An inference result is received from the inference engine. The inference result is stored in a database to associate the inference result with the data object. Then, in response to receiving a search request with one or more terms, the one or more terms are matched to the inference result in the database. and the data object and the inference result are returned.

Abstract: One or more computing devices, systems, and/or methods are provided. In an example, an internet resource identification item associated with one or more internet resources may be identified. User activity information associated with a plurality of events may be analyzed to determine a plurality of sets of text associated with the internet resource identification item, wherein each set of text of the plurality of sets of text is associated with an event, of the plurality of events, associated with an internet resource of the one or more internet resources. A plurality of term representations may be determined based upon the plurality of sets of text. A user intention-based representation associated with the internet resource identification item may be generated based upon the plurality of term representations. A content item may be selected for presentation via a client device based upon the user intention-based representation.

Abstract: A query is received from a client device and a mode is selected to process the query from a set of possible modes. The possible modes include a fast mode and a low-cost mode. If the fast mode is selected, the query is forwarded to a cloud database to retrieve responsive records. If the low-cost mode is selected, the cloud database is queried for index metadata of responsive records and the index metadata is used to retrieve the responsive records from a datastore other than the cloud database. Regardless of the mode selected, the responsive records are provided to the client device.

Abstract: In some implementations, a device may identify respective sets of unique values for multiple dimensions of a multidimensional data structure. The device may identify a plurality of subsets of permutations of a set of permutations of the unique values. The plurality of subsets of permutations are to be processed in parallel. The device may obtain, based on processing the plurality of subsets of permutations in parallel, respective data associated with each permutation of the plurality of subsets of permutations. The data for a permutation, of the plurality of subsets of permutations, is obtained based on respective unique values for the permutation that are determined independently of another permutation of the plurality of subsets of permutations.

Abstract: Methods of machine learning for system deployments without performance regressions are performed by systems and devices. A performance safeguard system is used to design pre-production experiments for determining the production readiness of learned models based on a pre-production budget by leveraging big data processing infrastructure and deploying a large set of learned or optimized models for its query optimizer. A pipeline for learning and training differentiates the impact of query plans with and without the learned or optimized models, selects plan differences that are likely to lead to most dramatic performance difference, runs a constrained set of pre-production experiments to empirically observe the runtime performance, and finally picks the models that are expected to lead to consistently improved performance for deployment. The performance safeguard system enables safe deployment not just for learned or optimized models but also for additional of other ML-for-Systems features.

Abstract: A system and method for receiving a query of a partitioned table, the query including a first index predicate associated with a first partition key column; determining that one or more of the query is missing a second index predicate and the first index predicate is unusable for index probing; responsive to determining that one or more of the query is missing the second index predicate and the first index predicate is unusable for index probing, generating one or more synthesized predicates used to process the query using an index scan.

Abstract: Monitoring an industrial process by building a training dataset of system data representative of status of industrial process parameters and training a custom query engine based on the training dataset. Models are generated using the custom query engine for matching query terms to the system data in response to user input representative of the system data that the user intends to access. Executing one of the models based on the input from the user generates an output retrieving the selected system data from the data tables for visualization.