Abstract: Systems and methods for pre-training and fine-tuning of neural-network-based language models to reason directly over tables without generating logical forms. In some examples, a language model can be pre-trained using masked-language modeling tasks synthetically generated from tables pulled from a knowledge corpus. In some examples, the language model may be further pre-trained using pairs of counterfactual statements generated from those tables, and/or one or more statements that compare selected data from those tables. The language model may then be fine-tuned using examples that include only a question, an answer, and a table, allowing fine-tuning examples to be harvested directly from existing benchmark datasets or synthetically generated.

Abstract: A toolbox application presents a user interface on a client device, and receives a natural language search query provided by a user of the client device that identifies a desired functionality to be performed for the user. The toolbox application determines whether any of a plurality of accessible utility components accessible is capable of performing the desired functionality specified in the search query. Responsive to determining that none of the accessible utility components is capable of performing the desired functionality, the toolbox application sends a request comprising the natural language search query to a server that maintains library of a plurality of utility components.

Abstract: Methods and system for text conversion are described. In one embodiment, free-form text associated with an item may be received. The item may be identified based on the free-form text. The item may be compatible with a parent item. The parent item may be identified based on the free-form text. An item descriptor may be identified in the free-form text. The item descriptor may be a particular term of the free-form text. Compatibility-based text may be constructed for the item based on identification of the parent item and the item descriptor. The compatibility-based text may be capable of being used to identify a plurality of matching items. Additional methods and systems are disclosed.

Abstract: In variants, a method for automatically determining responses in a messaging platform can include: receiving a set of inputs; determining a set of response options based on the set of inputs; and providing the set of response options to a user. The method can optionally further include training and/or updating a set of models, and/or updating a corpus of historical responses.

Abstract: Systems, methods, and software described herein provide enhancements of managing summaries provided to end users. In one implementation, a summary service identifies data objects that correspond to an event and classifies each of the data objects into classifications of interest. Once classified, data objects are compared between the different classifications to identify information differences, and the information differences are used to generate a summary for an end user of the summary service.

Abstract: One aspect of the present disclosure relates to a method of sentiment analysis based on ambiguity analysis, which includes analyzing information with the sentiment analysis models and the ambiguity analysis models. Another aspect of the present disclosure relates to a method of training the sentiment analysis models and ambiguity analysis models, which includes acquiring information, constructing lexicons, conducting sentiment analysis and ambiguity analysis with said lexicons, acquiring corpus, and training models, etc. Meanwhile, another aspect of the present disclosure relates to a system of sentiment analysis, which includes input, and output modules, acquisition modules, processing modules and database.

Abstract: State of art techniques that utilize spatial association based Table structure Recognition (TSR) have limitation in selecting minimal but most informative word pairs to generate digital table representation. Embodiments herein provide a method and system for TSR from an table image via deep spatial association of words using optimal number of word pairs, analyzed by a single classifier to determine word association. The optimal number of word pairs are identified by utilizing immediate left neighbors and immediate top neighbors approach followed redundant word pair elimination, thus enabling accurate capture of structural feature of even complex table images via minimal word pairs.

Abstract: Aspects described herein may allow an automated generation of an interactive multimedia content with annotations showing vehicle damage. In one method, a server may receive vehicle-specific identifying information of a vehicle. Image sensors may capture multimedia content showing aspects associated with exterior regions of the vehicle, and may send the multimedia content to the server. For each of the exterior regions of the vehicle, the server may determine, using a trained classification model, instances of damage. Furthermore, the server may generate an interactive multimedia content that shows images with annotations indicating instances of damage. The interactive multimedia content may be displayed via a user interface.

Abstract: Provided are a search method and apparatus, and an electronic device and a storage medium. The method comprises: receiving a generic query search request, the generic query search request carrying a generic query term; acquiring at least one multimedia content card corresponding to the generic query term, wherein each multimedia content card corresponds to an extension tag corresponding to the generic query search request, each multimedia content card comprises information of a plurality of multimedia content sets; and presenting information of the plurality of multimedia content sets comprised in each multimedia content card of the at least one multimedia content card. Search results related to a generic query term can be clustered from various dimensions and presented, such that a user can directly find content related to a search intention of the user, thereby improving the search efficiency and shortening a search path.

Abstract: A method for communicating a playback order for a plurality of media content items to a user device operating in an online mode, the method performed at a server system and comprising receiving an indication that the user device will enter an offline mode, generating a playback order for the plurality of media content items, and transmitting the generated playback order to the user device before the user device enters the offline mode.

Abstract: A media recommendation system may score media items according to user recommendations, popularity, and/or recency. The scores may be weighted to produce an overall score for each media item. Media items may be added to a pool for a specific user, from which media items are selected for playback. The contents of the pool may be modified based upon user feedback and other data. The pool may be modified dynamically and/or in real time as media items are consumed or rated by the user.

Abstract: A machine may form all or part of a network-based system configured to provide media service to one or more user devices. The machine may be configured to define a station library within a larger collection of media files. In particular, the machine may access metadata that describes a seed that forms the basis on which the station library is to be defined. The machine may determine a genre composition for the station library based on the metadata. The machine may generate a list of media files from the metadata based on a relevance of each media file to the station library. The machine may determine the relevance of each media file based on a similarity of the media file to the genre composition of the station library as well as a comparison of metadata describing the media file to the accessed metadata that describes the seed.

Abstract: Disclosed are methods, systems, and non-transitory computer-readable medium for determining color labeling for a vehicle search service. For instance, the method may include: obtaining vehicle images for a plurality of vehicles; filtering the vehicle images to obtain exterior and/or interior images of the plurality of vehicles, determining super-pixel values for the exterior images and/or the interior images; generating a standard color mapping by associating the super-pixel values to standard color names; and storing searchable vehicle information at least based on the standard color mapping.

Abstract: Methods, apparatuses, and embodiments related to analyzing the content of digital images. A computer extracts multiple sets of visual features, which can be keypoints, based on an image of a selected object. Each of the multiple sets of visual features is extracted by a different visual feature extractor. The computer further extracts a visual word count vector based on the image of the selected object. An image query is executed based on the extracted visual features and the extracted visual word count vector to identify one or more candidate template objects of which the selected object may be an instance. When multiple candidate template objects are identified, a matching algorithm compares the selected object with the candidate template objects to determine a particular candidate template of which the selected object is an instance.

Abstract: Systems and methods are provided to process a digital photo and other media. An apparatus to process digital photos can include a tangibly embodied computer processor (CP) and a tangibly embodied database. The CP can perform processing including: (a) inputting a photo from a user device, and the photo including geographic data that represents a photo location at which the photo was generated; (b) comparing at least one area with the photo location and associating an area identifier to the photo as part of photo data; and (c) performing processing based on the area identifier and the photo data.

Abstract: An embodiment may involve, based on a profile associated with a client device, selecting an audio file containing music. Based on an attribute of the audio file containing the music, an audio file containing a story may be selected. A playlist for the client device may be generated, where the playlist includes (i) a reference to the audio file containing the music, and (ii) a reference to the audio file containing the story. A server device may transmit the playlist to the client device over a wide area network. Reception of the playlist at the client device may cause an audio player application to retrieve and play out each of the audio file containing the music and the audio file containing the story.

Abstract: A system for supporting a user's repetitive motion activity operates to manage cadence-based playlists identifying one or more media content items having a tempo corresponding to a user's cadence. The cadence-based playlists can be categorized by different tempi or tempo ranges that cover all likely cadences during the user's activities. A media-playback device is provided to acquire a user's cadence and retrieve a cadence-based playlist associated with a tempo or a tempo range corresponding to the cadence.

Abstract: Dynamic determination of video correlation is disclosed, including: obtaining a plurality of recorded video interactions; obtaining a plurality of recorded non-video interactions; storing, at a database, event data determined based at least in part on the plurality of recorded video interactions and the plurality of recorded non-video interactions; and determining a subset of the event data that comprises correlations between a video and non-video interactions.

Abstract: Systems and methods are described for providing a reply to a query related to a media asset. A query may be received from a user while the media asset is being played on a first device, and in response to determining that the query is related to the media asset, a snapshot of the media asset may be captured, where the snapshot comprises a depiction of a first object and a second object, and the snapshot may be generated for display at a second device. In response to determining there is ambiguity whether the query is related to the first or second object, a disambiguating query based on the first and second objects of the snapshot may be generated for simultaneous output with the snapshot. In response to receiving a reply to the disambiguating query, a response to the query may be generated for output based on the reply.

Abstract: Disclosed are systems and methods for XBRL tag outlier detection. In some embodiments, the method includes the steps of: receiving a first set of XBRL data records; generating a second set of XBRL data records based upon a subset of the first set of XBRL data records; training a machine learning model using the first set of XBRL data records and the second set of XBRL data records; receiving an XBRL document associated with one or more assigned XBRL tags; and analyzing the XBRL document using the trained machine learning model to identify a set of outlier XBRL tags in the one or more assigned XBRL tags.

Abstract: Provided is an electronic device including a file system, a memory, and a processor. The memory stores at least one instruction that instructs, when executed, the processor to receive a request for access to a file or directory stored in a user data area of the file system. In response to the reception of the request, first information associated with the file or directory is store in the memory, and the size of the file or directory is displayed using the first information. When the first information is information associated with the file, the information indicates the size of the file. When the first information is information associated with the directory, the information indicates a sum of the sizes of all subdirectories and all files included in the directory.

Abstract: Systems and methods that provide visualization of networks. Data is input into a table structure that represents any hierarchy of entities, relationships and their attributes. The content of the table is processed to extract the entities, relationships and their attributes. These are turned into nodes, edges and a visual representation of their attributes using color gradients, categorical colors, shapes, thickness, text labels, etc.

Abstract: A method for utilizing a graph path cache to facilitate real-time data consumption by a plurality of machine learning models is disclosed. The method includes receiving an input from a source, the input relating to a request to characterize a data element; retrieving a data attribute that corresponds to the data element from a data management system; determining, in real-time using the graph path cache, a graph attribute that corresponds to the data element by performing deep link analysis on a graph database; executing, in real-time, a model by using the data attribute and the graph attribute, the model corresponding to the request in the input; and transmitting, in real-time, a result of the executed model to the source in response to the input.

Abstract: One or more processing devices cause display of a user interface that identifies a service definition representing a service, receive input identifying an entity, where the service is performed at least in part by the entity, and store the service definition representing the service in association with an entity definition representing the entity. The entity definition comprises information identifying data pertaining to the entity in a datastore of machine data that reflects activity in an information technology environment produced by a plurality of components of the information technology environment. The one or more processing devices receive input pertaining to a search definition representing a search producing a measure of the service, and store the search definition representing the search, where the search produces the measure of the service using at least a portion of the data pertaining to the entity.

Abstract: There is provided an information processor, an information processing method, and a recording medium to automatically estimate a sense of values of a community. The information processor includes a control unit that performs a control. The control unit acquires sensor data obtained by sensing a member who belongs to a specific community, and automatically estimate a sense of values of the member who belongs to the specific community, on the basis of the acquired sensor data.

Abstract: Disclosed herein is a software tool that may be used to create and navigate through visualizations of an organization's data as embodied in a digital duplicate. These “smart interactions” may allow a user to visually and intuitively traverse the semantic network to dynamically provide on-demand visualization of the business data. Based on the traversal of the network, the network can be simplified and/or subnetworks can be created. As part of the visualization, “perspectives” can be created based on focal points (i.e., nodes) in the network. The focal points may be used as a starting point for the traversal, and functions (e.g., aggregates) can be dynamically applied to produce subnetworks to provide data for visualizations.

Abstract: Techniques are provided for comparing multi-level data structures using commutative digesting for unordered data collections. One method comprises obtaining two multi-level data structures, wherein at least one multi-level data structure comprises an unordered data collection; determining a data structure digest value for each the two multi-level data structures by accumulating a data element digest value for each data element of the respective multi-level data structure, wherein a data element digest value for a given data element comprising an unordered data collection is determined using a commutative accumulator function; and evaluating a similarity of the two multi-level data structures by comparing the respective data structure digest values. A data element digest value for a given data element comprising an ordered data collection can be determined using a noncommutative accumulator function that aggregates a digest value for each data element using a noncommutative operation.

Abstract: An information processing method executed by a computer includes: obtaining (i) activity data of a user at a predetermined time point, (ii) profile data of one of the user and a different user who has an attribute that is identical to or similar to an attribute of the user, the profile data indicating one of an activity and a characteristic, and (iii) mobile object data; generating activity demand data of the user after the predetermined time point, in accordance with the profile data and the activity data of the user at the predetermined time point; determining a mobile object candidate that is to be recommended to the user, in accordance with the activity demand data of the user and the mobile object data; and controlling notification about the mobile object candidate.

Abstract: A method for data presentation includes assembling a set of documents in response to a search query by a user having a social network. A respective person is identified with each of a group of documents in the set. For at least one document in the group, a chain of one or more links is found in the social network between the respective person and the user. A search results page is displayed, including an identification of the at least one document together with a sequence of breadcrumbs corresponding to the one or more links in the chain.

Abstract: A method and system for controlling processing by a data server. An example method includes a computing system receiving from a client entity a request for information, and, responsive to receiving from the client entity the request, the computing system (i) determining, based on the client entity from which the computing system received the request, a processing time limit for the data server to apply in processing a query for the information and (ii) generating and sending to the data server the query, including in the query an indication of the determined processing time limit. The data server could thus receive and process the query and impose the specified processing time limit on the processing of the received query.

Abstract: Improved crawling and curation of data and metadata from diverse data sources is described. In some embodiments, improvements are achieved by interpreting the context, vocabulary and relationships of data element, to enable relational data search capability for users. The user querying process is improved by systematic identification of the data objects, context, and relationships across data objects and elements, aggregation methods and operators on the data objects and data elements as identified in the curation process. User query suggestions and recommendations can be adjusted based on the context, relationships between the data elements, user profile, and the data sources. When the user query is executed, the query text is translated into an equivalent of one or more query statements, such as SQL or PostGre statements, and the query is performed on the identified data sources. Results are assembled to present the answer in a meaningful visualization for the user query.

Abstract: To find enriching contextual information for an abbreviated domain name, a data enrichment engine can comb through web content source code corresponding to the abbreviated domain name. From textual content in the web content source code, the data enrichment engine can identify words with initial characters that match characters of the abbreviated domain name to thereby establish a relationship there-between. This relationship can facilitate more accurate and efficient domain name classification. The data enrichment engine can query a WHOIS server to find out if candidate domains having initial characters that match the characters of the abbreviated domain name are registered to the same entity. If so, keywords can be extracted from the candidate domains and used to find more relevant domains for domain risk analysis and detection. Candidate domains determined by the data enrichment engine can be provided to a downstream computing facility such as a domain filter.

Abstract: A method and server for ranking digital documents for a query are disclosed. T server hosts a search engine associated with an inverted index. The method includes accessing inverted index for retrieving query-independent data for a first document-term pair and a second document-term pair, where the query-independent data is indicative of (i) a term-specific occurrence of the first term in content of the document and (ii) a term-specific occurrence of the second term in the content of the document. The method includes generating a query-dependent feature using the query-independent data and which is indicative of a group occurrence of the first term with the second term in the content of the document. The method includes generating a ranking feature for the given document based on at least the first term, the second term, and the query-dependent feature, and ranking the given document based on at least the ranking feature.

Abstract: Methods, systems, and devices for generating contact suggestions for a user of a social network. A first score is computed for each one of the plurality of users, the first score being computed using an edge-weighted ranking algorithm based on the user graph. A second score is computed, using a machine learning model, for each user of the plurality of users, the second score of each user being, at least partially, based on the first score of said user, with the second score of each user being representative of a probability of a first user sending a connection request to said user. A ranked contact suggestion list of one or more users of the plurality of users is generated, the one or more users being ranked based on their respective second score.

Abstract: A system and method for generating and publishing inferential newsfeeds captures actively and passively provided user data elements which are processed into user facts. A Cache Management Module maintains updated cached copies of rules and execution plans. The cached copies of rules and execution plans are applied to the user facts, to inferentially generate newsfeed posts. An Inferential Data Mining Module runs each execution plan in a multi-threaded process for each user, and processes each rule within the context of the execution plan to generate the newsfeed posts. A Newsfeed Generation Module applies newsfeed templates to format the newsfeeds for various platforms. The Inferential Data Mining Module mines the Data Store for a determination of which user facts should be used to create the newsfeed posts, and of which users should be designated as subscribers to the newsfeed posts, said determination taking place independently of individual users' direct actions.

Abstract: The present application discloses a method, system, and computer system for running a test in a database system. The method includes obtaining, by one or more processors associated with a server, a template comprising one or more variables, storing, in a cache, a mapping of field definitions for a particular dataset, populating, by the one or more processors, the template to generate a populated template based at least in part on the mapping of field definitions for the particular dataset, performing, by the one or more processors, a query with respect to the particular dataset using values obtained from the populated template, and monitoring, by the one or more processors, one or more performance characteristics associated with the performing the query.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, that identify and issue search queries expected to be issued in the future. A set of search queries that have been issued by multiple user devices can be obtained. For each query instance, contextual data can be obtained. A first query and its contextual data can be input to a model that outputs the query's likelihood of being issued in the future. The model can be trained using contextual data for training queries and a corresponding labels for the training queries. The learning model outputs the first query's likelihood of being issued in future, and this query is stored as a repeatable query if the likelihood satisfying a repeatability threshold. Subsequently, a stored repeatable query is issued upon a selection of a user selectable interface component and the search engine provides search results for the query.

Abstract: A virtual assistant server receives a user input from a virtual assistant interface available on a client system associated with a user identified by a user identifier. Upon receiving the user input, the virtual assistant server creates a search network of repositories not associated with the user identifier by determining, for the received user input, a priority score of each repository not associated with the user identifier based on interactions of the user identifier through the virtual assistant interface. Further, the virtual assistant server searches, for the received user input, in a repository associated with the user identifier and in the repositories in the search network. Based on the searching, a group of one or more results corresponding to the received user input are identified and returned to the client system associated with the user identifier.

Abstract: Methods, systems, and devices for sharing a physical location with online users by requiring the online users to be in physical proximity of a specific physical location before the users can experience media content associated with the physical location. A user can mark a physical location when that user is at the location by associating a virtual anchor with that location. The virtual anchor creates a data structure that contains the location's geographic coordinates as well as other details regarding the anchor. The user can also associate media content with the virtual anchor. The virtual anchor can be shared with other users and these other users are notified of the presence of the virtual anchor when they are within a physical proximity from the physical location. Only when the other users are at the location can these users access the media content associated with the virtual anchor.

Abstract: Disclosed are techniques for faceted search queries through interactive graphical objects. A text search query corresponding to a faceted search is received from a user, which is then parsed for keywords. Attributes associated with the keywords are then extracted and used to generate a graphical representation corresponding to the results associated with the text search query. Further user input to refine the query is provided through manipulation of the graphical representation by the user (such as dragging a leg of a table to change the length of legs on a table) corresponding to modification of one or more facets of the faceted search, narrowing the available search results corresponding to the faceted search.

Abstract: An embodiment of a system for determining the disposition of a hyperdocument and hyperdocument requests using a trained artificial neural network, including an information source, a requesting application, and a server containing a trained artificial neural network (ANN), the ANN being capable of evaluating the information, and providing results reflecting the evaluation to a requesting application is described.

Abstract: Dynamic link preview generation techniques are described that overcome the challenges of conventional techniques by supporting link preview generation by a content provider system that proves the digital content via a respective network address. In one example, a content provider system, based on a request received from a service provider system, identifies a communication platform of the service provider system that is to be used to communicate the shared link. Upon identifying the communication platform, the content provider system locates customization data that describes how the link preview is to be generated for the communication platform. In response, the content provider system renders digital content available via the network address to generate the link preview.

Abstract: Disclosed herein are system, method, and computer program product embodiments for accessing a data service using a data service Uniform Resource Identifier (URI) object. An embodiment operates extracting a name-value pair of each parameter stored in the data service class object. The embodiment generates a parameter context URI using the one or more extracted name-value pairs according to a data service specification of the data service. The embodiment extracts one or more filter range items for each filter option stored in the data service class object. The embodiment generates a filter context URI using the one or more extracted filter range items according to the data service specification of the data service. The embodiment generates a context URI using the parameter context URI and the filter context URI. The embodiment then accesses the data service according to the context URI.

Abstract: In a computer; a computer readable medium for providing rapid syndrome analysis includes instructions for measuring a plurality of CUSUMs simultaneously and generating CUSUM data in response to an observed value. It also includes instructions for comparing the CUSUM data with a predetermined plurality of threshold values and instructions for generating a signal when the CUMSUM data exceeds any of the plurality of threshold values by a predetermined amount.

Abstract: A system and method is described herein for data filtering to reduce functional, and trend line outlier bias. Outliers are removed from the data set through an objective statistical method. Bias is determined based on absolute, relative error, or both. Error values are computed from the data, model coefficients, or trend line calculations. Outlier data records are removed when the error values are greater than or equal to the user-supplied criteria. For optimization methods or other iterative calculations, the removed data are re-applied each iteration to the model computing new results. Using model values for the complete dataset, new error values are computed and the outlier bias reduction procedure is re-applied. Overall error is minimized for model coefficients and outlier removed data in an iterative fashion until user defined error improvement limits are reached. The filtered data may be used for validation, outlier bias reduction and data quality operations.

Abstract: Hardware implementations of, and methods for processing, a convolution layer of a DNN that comprise a plurality of convolution engines wherein the input data and weights are provided to the convolution engines in an order that allows input data and weights read from memory to be used in at least two filter-window calculations performed either by the same convolution engine in successive cycles or by different convolution engines in the same cycle. For example, in some hardware implementations of a convolution layer the convolution engines are configured to process the same weights but different input data each cycle, but the input data for each convolution engine remains the same for at least two cycles so that the convolution engines use the same input data in at least two consecutive cycles.

Abstract: A server is provided and is programmed to automatically determine, based on the identification data that uniquely identifies a patient device, whether there is an upgrade to the code or data (e.g., software or firmware) stored in the electronic memory of the patient device that can be applied thereto. In response to determination that there is an upgrade that can be applied, a data package is delivered to the patient device via the first wireless transceiver, the data package to be automatically applied to the electronic memory of the patient device.

Abstract: A neural network includes a drop layer configured to drops feature values. A method of computation using the neural network includes extracting feature data from input data using a first portion of a neural network, generating compressed representation data of the extracted feature data by dropping a feature value from the extracted feature data at a drop layer of the neural network based on a drop probability corresponding to the feature value, and indicating an inference result from the compressed representation data using a second portion of the neural network.

Abstract: An online system accesses a list of features used as input into a predictor to predict a performance metric for content presented to users. The online system computes importance scores for one or more of the features. A ranked list of categories is created, with each category having one or more sub-categories. For each feature having a computed importance score, the online system assigns, for each attribute in the ranked list of attributes for that feature, the feature to a sub-category in one of the categories in the ranked list of categories that has the same rank as the attribute in the ranked list of attributes for the feature, where the sub-category is associated with a label that corresponds with the attribute. For each sub-category in each category, a cumulative score is computed for the sub-category based on the importance scores of the features of that sub-category.

Abstract: According to one embodiment, an image processing device includes first storage and a processor. The first storage is configured to store a first statistical model generated by learning bokeh which occurs in a first image. The processor is configured to acquire a second image, acquire a bokeh value which indicates bokeh occurring in the second image and an uncertainty level which indicates a level of uncertainty for the bokeh value, generate a first bokeh map based on the bokeh value and uncertainty level, and acquire a second bokeh map obtained by interpolating a bokeh value with respect to the first bokeh map. The bokeh value and the uncertainty level are output from the first statistical model by inputting the second image into the first statistical model.