Abstract: In one aspect, method useful for goal-oriented dialog automation comprising incudes the step of receiving an input message. The method includes the step of implementing an entity tagging operation on the input message. The method includes the step of tagging the message context of the input message to generate a tagged message context. The method includes the step of implementing semantic frame extraction from the tagged message context. The method includes the step of implementing an entity interpretation on the extracted frame. The method includes the step of accessing a database to determine a business schedule and a client profile. The business schedule and the client profile are related to the input message. The method includes the step of implementing a retrieval engine. The retrieval engine obtains one or more response templates. The method includes the step of generating a ranked list of candidate templates from the output of the retrieval engine.

Abstract: A method for decoding a natural language user query involves obtaining the user query submitted by a user, segmenting the user query into words, generating a character embedding for each of the words, and generating a word embedding for each of the words. The method further involves obtaining a clickstream from tracked clicks of the user, generating a clickstream embedding from the clickstream, and for each of the words, generating a unified feature representation based on the character embedding and the word embedding for each of the words, and the clickstream embedding. The method also involves decoding the unified feature representations to obtain a decoded user query, and processing the user query using the decoded user query.

Abstract: Method and apparatus for training and using a classifier for words. Embodiments include receiving a first plurality of sentences comprising a first word that is associated with a class and a second plurality of sentences comprising a second word that is not associated with the class. Embodiments include training a classifier using positive training data for the class that is based on the first plurality of sentences and negative training data for the class that is based on the second plurality of sentences. Embodiments include determining a measure of correlation between a third word and the class by using a sentence comprising the third word as an input to the classifier. Embodiments include using the measure of correlation to perform an action selected from the following list: selecting content to provide to a user; determining an automatic chat response; or filtering a set of content.

Abstract: One embodiment provides a method comprising generating a conversational interface for display on an electronic device. The conversational interface facilitates a communication session between a user and an automated conversational agent. The method further comprises performing a real-time analysis of a portion of a user input in response to the user constructing the user input during the communication session, and updating the conversational interface to include real-time feedback indicative of whether the automated conversational agent understands the portion of the user input based on the analysis. The real-time feedback allows the user to adjust the user input before completing the user input.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for announcing and detecting automated conversation are disclosed. One of the methods includes initiating, over a natural language communication channel, a conversation with a communication participant using a natural language communication method that includes a dialogue of natural language communications. The communication participant is determined to be automated using a pre-defined adaptive interactive protocol that specifies natural language linguistic transformations defined in a sequence. The conversation can be transitioned to a communication method that is different form the natural language communication method in response to determining that the communication participant is automated.

Abstract: A computer-implemented learning method includes combining a first score for a word included in a dictionary of a model and a second score for the word calculated based on a length of the word and a number of remaining characters up to an upper limit number of characters of a summary, the first score being output in response to input of input text to the model, calculating, based on a combined score generated by the combining, a distribution of generation probabilities of the word, and updating, based on the calculated distribution of the generation probabilities, parameters in the model.

Abstract: Systems and methods for automatically modeling the discourse structure of psychiatric reports and segmenting these reports into various sections are provided. The systems and methods can be based around a model that learns the section types, positions, and sequence and can automatically segment unlabeled text in a psychiatric report into the corresponding sections. Knowledge of the ordering of the sections can improve the performance of a section classifier and a text segmenter. A Hierarchical Hidden Markov Model (HHMM) can be trained and can categorize sections in psychiatric reports into a predefined section label.

Abstract: Embodiments relate to a system, program product, and method for use with an intelligent computer platform and cognitive processing. The embodiments support a mechanism for employing machine learning into dynamically processing a file. File data points are converted to vector form and augmented with similar or synonymous terms. The vector representations of the data points are utilized to identify words and their frequencies, and the identified frequencies are extrapolated and applied to the synonymous terms. As communications are received, the ML model converts the communication to vector form, and compares a feature vector of the communication with data point or file feature vectors to identify and map the communication to a similarly classified file.

Abstract: The disclosed computer-implemented method for performing natural language translation in AR may include accessing an audio input stream that includes words spoken by a speaking user in a first language. The method may next include performing active noise cancellation on the words in the audio input stream so that the spoken words are suppressed before reaching a listening user. Still further, the method may include processing the audio input stream to identify the words spoken by the speaking user, and translating the identified words spoken by the speaking user into a second, different language. The method may also include generating spoken words in the second, different language using the translated words, and replaying the generated spoken words in the second language to the listening user. Various other methods, systems, and computer-readable media are also disclosed.

Abstract: A multilingual exchange system and message transmission method that enables exchange of messages in a foreign language in a simple manner without any special translation skill, and, simultaneously, allowing for dramatic improvement in grammatical accuracy when converting messages into a second language. In an outer box 106, which is a display area for a modifier in a third sentence pattern sentence inner boxes 106A, 106B, and 106C are allocated to each part of speech in English, which is the second language, and inner boxes 106A, 106B, and 106C can be recognized as separate inner boxes. Even if display content in a first language is divided into three inner boxes, there is no need to divide display content in second language into three inner boxes, and even if substitutions of parts of speech or changes in content occur in one language, it is possible to maintain grammatical accuracy in another language.

Abstract: The invention relates to header and trailer record validation for batch files. According to an embodiment of the present invention, a computer implemented system implements a Header/Trailer Validation Tool. The Header/Trailer Validation Tool may read a control file containing pertinent information of each file to be validated. Key information may be determined at run time and the Header/Trailer Validation Tool may process the files dynamically. An embodiment of the present invention may also process any number of files in a single execution—this is particularly useful because in some cases a set of files may be received from another application, but not all the files may be used at the same time.

Abstract: Techniques are provided for compression and decompression of log data using pattern recognition. An exemplary method comprises: obtaining at least one log message, wherein the at least one log message comprises a pattern-based message template and one or more message variables, wherein the pattern-based message template was learned from historical log data using a log pattern mining technique; obtaining a compression index that maps a plurality of pattern-based message templates to a corresponding message signature; and writing the one or more message variables and a message signature corresponding to the pattern-based message template of the at least one log message to a log file. The log pattern mining technique learns a given pattern-based message template from the historical log data to represent a plurality of historical log messages. The at least one log message may comprise a historical log message or may be from a real-time stream of log messages.

Abstract: One example method includes implementing a function as a service (FaaS) at a datacenter by performing operations including receiving an application program interface (API) gateway call from a client application, wherein the API gateway call is associated with an object PUT request, and automatically triggering, with the API gateway call, performance of an object insertion function. The object insertion function includes retrieving, from backend object storage, a previous version of the object, differentially compressing the object relative to the previous version of the object so as to generate a differential, and storing the differential in the backend object storage.

Abstract: An embodiment of the invention may include a method, computer program product and computer system for profile picture selection. The method, computer program product and computer system may include a computing device which may classify user photographs in user account data into one or more photograph classifications. The computing device may classify an emotion conveyed by a media interaction by the user and select a classified user photograph matching the emotion conveyed by the media interaction by the user. The computing device may display selected user photograph with the media interaction by the user.

Abstract: Computer-implemented techniques for constructing structured database query language statements from natural language questions are disclosed. In some embodiments, a knowledge graph-based approach is employed to construct a structured database query language statement such as, for example, a Structure Query Language (SQL) statement, from a natural language question. The approach encompasses a domain knowledge graph model and a database schema wiring model. The domain knowledge graph model represents domain-specific knowledge about the underlying database data in the form of a graph having nodes and directed edges between the nodes. The database schema wiring model maps routes in the domain knowledge graph to database schema elements. The techniques allow a natural language interface system to accurately translate domain-specific natural language questions posed by users to structured database query language statements that can be executed against a structured database to answer the natural language question.

Abstract: A computer-implemented method for constructing and searching structured data of question sentences in a question-answer database using machine learning and natural language processing includes receiving one or more question-answer articles as input from one or more question-answer databases over at least one network, extracting one or more profile keywords from a profile associated with a user, assigning a personalization score corresponding to the user to each of the one or more question-answer articles based on the one or more profile keywords, including applying a morphological and dependency analysis to a body section of the question-answer article, receiving a query for obtaining question-answer data relevant to the user, and outputting question-answer data based at least in part on the personalization score of each question-answer article in satisfaction of the query.

Abstract: A computer-implemented method for implementing separated attention on like and dislike items for personalized ranking includes performing an element-wise product on a user embedding and a final like item embedding to generate a first vector. The method further includes performing an element-wise product on the user embedding and a final dislike item embedding to generate a second vector. The method further includes computing a probability that the user prefers the like item to the dislike item based on the first and second vectors, and generating one or more item recommendations including one or more electronic images for the user using the probability.

Abstract: Aspects of the present invention provide devices that identify a question in a text message of a chat session between a plurality of computing devices, wherein each of the plurality of computing devices includes an identifier of a user of a corresponding computing device in the text message, analyze digital data to determine a relationship between the users of the plurality of computing devices, construct search parameters for digital content external to the chat session according to the identified question and determined relationship, search for the digital content over a network according to the constructed search parameters to obtain search results, rank the search results according to the determined relationship, and distribute the ranked search results according to the determined relationship to at least one of the plurality of computing devices in an overlay of the chat session.

Abstract: In an example embodiment, a fast and space-efficient ranking mechanism is used for vertical searches. A precedence is assigned to a plurality of features based on the relative importance of those features to the provider of a vertical search engine. In an example embodiment, this relative importance may be explicitly indicated by the provider of the vertical search engine. A unique data structure may then be created that incorporates the precedence of each feature.

Abstract: Provided is a process, including: receiving a command corresponding to user input to an interactive dashboard application from a user; producing in response to the command, instances of data visualizations depicting data to be visualized; generating, with a trained captioning model, one or more narrative captions determined to be descriptive of the produced instances of data visualizations, wherein the one or more narrative captions include a natural language description of a phenomenon exhibited, at least in part, by the data to be visualized and visually depicted in at least one of the produced instances of data visualizations; and causing the one or more narrative captions to be presented to the user.

Abstract: Methods, apparatus, and processor-readable storage media for incorporating data into search engines using deep learning mechanisms are provided herein. An example computer-implemented method includes extracting one or more features from a search query by applying one or more machine learning algorithms to the search query; generating one or more word vectors by applying at least one deep learning technique to the one or more extracted features; mapping the one or more generated word vectors to one or more words from a corpus of data by implementing at least one deep similarity network; and outputting one or more results in response to the search query, wherein the one or more results are based at least in part on the one or more words from the corpus to which the one or more generated word vectors were mapped.

Abstract: A textual content adaption method, system, and computer program product, include receiving a corpus of content, segmenting the content into sections, determining a classification for each section by determining an audience and an expertise level for the section, determining a baseline for a user, modifying a portion of content from the corpus based on the classification of the sections included in the portion of content and based on the baseline of the user, and providing the modified portion of content to the user.

Abstract: An efficient data storage system is described. An agent software application on computing devices in a first tier processes snapshot backups and pushes them to an appliance software application on a server in a second tier. The appliance software application processes archive backups and pushes them to cloud storage in a third tier. A cloud application on a management server receives storage policy specifications from customers and promulgates the policies to the agent software application and the appliance software application. The policy specifications include time periods and retention set information for the backups. When a retention set has been exceeded, the storage system is pruned to remove file references to unneeded files and delete data files no longer referenced in storage sets in the retention set.

Abstract: Methods and systems are provided for improving a journaling file system by reducing journaling overhead utilizing a remapping interface to log-structured cloud block devices such that blocks in a journaling transaction are atomically mapped to their final disk locations with minimum overhead. With log-structured cloud block devices, all blocks may be stored in logfiles and indexed for access. By modifying the index of the destination block to point to the corresponding journal block, the additional write may be avoided, and a journal block may effectively be remapped from its original LBA to its destination LBA without any actual data movement, thereby reducing the file system journaling overhead.

Abstract: In general, embodiments of the invention relate to a method and system for managing persistent storage in a local computing device. More specifically, embodiments of the invention relate to determining the amount of space that will be freed up (or become available) in the persistent storage during a data transfer using a perfect hash function. Once the amount of data to be transferred is determined, embodiments of the invention initiate the allocation of an appropriate amount of space in the remote storage device and, subsequently, initiate the transfer of the data to the remote storage device.

Abstract: Example distributed storage systems, file system interfaces, and methods provide duplicate request checking. A file interface manages a plurality of file data requests from client systems using a file system protocol. A metadata store acts as a cross-reference between file data references and storage data references. A request data store in the metadata store includes request entries with transaction identifiers for corresponding file data requests. A duplicate checker searches the request data store and excludes incoming file data requests when it finds a matching request entry.

Abstract: A system, method, and computer-readable medium, including adding a dedicated data storage structure to a first set container, the added dedicated data storage structure to store business objects(BOs) constructed based on data foundation fragments injected into a data foundation of a semantic layer universe; generating, automatically, a BO based on a data foundation of the semantic layer universe; storing the automatically generated BO in the added storage structure and storing the added data storage structure in the first set container.

Abstract: A technique for writing data in a file system includes aggregating data received in a set of I/O requests into a batch that includes multiple extents of data. After compressing a current extent of the batch and determining that the compressed extent does not fit in a space where a previous version of the extent is stored, the technique performs an FS-relocate operation by accessing an FS-hole list provided for the file system and selecting a hole, from the FS-relocate list, which is large enough to accommodate the compressed extent. The technique then places the compressed extent in the selected hole.

Abstract: Techniques are provided for compression of log data using field types. An exemplary method comprises: obtaining at least one log message, wherein the at least one log message comprises a message template and at least one message variable, wherein the message template comprises at least one field corresponding to the at least one message variable; obtaining a compression index that maps a plurality of message templates to a corresponding message signature; identifying a predefined field type of the at least one field; selecting a compression technique to apply to the at least one message variable based on the predefined field type to obtain a compressed message variable; and writing the compressed message variable and a message signature corresponding to the message template of the at least one log message to a log file. The at least one log message can be a historical log message or part of a real-time stream of log messages.

Abstract: Techniques for data processing may include: receiving a data chunk of the data set; determining, in accordance with criteria including a compressibility ratio for the data set and a cost ratio of compression computation cost and entropy computation cost, whether to activate or deactivate entropy computation for the data set, wherein the compressibility ratio is ratio of a number of compressible data chunks of the data set and a number of uncompressible data chunks of the data set; and responsive to determining to activate entropy computation for the data set, performing first processing comprising: determining an entropy value for the data chunk; and determining, in accordance with the entropy value for the data chunk, whether to compress the data chunk.

Abstract: Techniques for determining data reduction options may include: receiving first data reduction information regarding compression and deduplication of chunks of a data set; determining, in accordance with the first data reduction information for the data set, first settings denoting whether compression and deduplication are enabled or disabled for the data set; receiving, during a first time period when the first settings are effective, writes directed to the data set; receiving second data reduction information regarding compression and deduplication of chunks of the data set modified by writes during the first time period; and determining, in accordance with the second plurality of data reduction statistics for the data set, second settings denoting whether compression and deduplication are enabled or disabled for the data set.

Abstract: Techniques for data processing may include: receiving a data chunk; determining a metric value denoting a degree of compressibility of the data chunk; selecting, in accordance with the metric value denoting the compressibility of the data chunk, a first size of a plurality of sizes, wherein each of the plurality of sizes denotes a different size of an amount of storage used for storing a value of said each size; and performing the data deduplication processing for the data chunk, wherein the data deduplication processing includes using a first hash value for the data chunk to determine whether the data chunk is a duplicate of another data chunk of a hash table, wherein the first hash value is stored in a storage location of the first size.

Abstract: Deduplication of files or other data or objects in a manner that is aware of a format of the file and the application. A file may be chunked based on the format. The chunks are more consistent and lead to higher deduplication ratios. The file may be presented as a single file. However, the file is stored in chunks or subfiles and deduplication is performed with respect to the chunks. When the file is read, the file is rebuilt from its respective chunks. The files may also be compressed using differential compression, which leverages the content of similar files to compress a current file.

Abstract: An apparatus in one embodiment comprises at least one processing device comprising a processor coupled to a memory. The at least one processing device is configured to obtain a file and to determine a type of the file. A structure of the file may be determined based at least in part on the determined type of the file and at least one location in the file may be identified based at least in part on the determined structure. The file may be divided at the identified at least one location into a plurality of chunks and the plurality of chunks may be provided to a block deduplication module of a storage system where the block deduplication module is configured to perform a deduplication process based at least in part on the plurality of chunks.

Abstract: Data replication in a distributed file network. When replicating an object from a source node to target nodes, an overlay plan is developed. The plan may consider bandwidth between the nodes such that the object is replicated more effectively. As a result, chunks of the object may pass through multiple nodes. As a result, more than one node or site can serve as a source for some of the chunks. When the replication process is completed, the source node or site and the target node or site each have a copy or replica of the object.

Abstract: A method for managing storage devices includes, in response to a first determination that a first cloud storage device is inaccessible to a local storage device, obtaining metadata objects from a second cloud storage device, identifying, based on a sync time analysis, at least one metadata object of the metadata objects to be deleted, initiating the deletion of the at least one metadata object on a second cloud storage device, wherein the second cloud storage device is accessible to the local storage device, and updating, after initiating the deletion of the at least one metadata object, a namespace on a local storage device.

Abstract: An approach is provided for storing and retrieving map data using contextual information priors. The approach involves, for example, processing contextual information to determine a restricted range of location information relevant to at least one query. The approach also involves processing sensor data received from at least one sensor, the sensor data collected at at least one query location, to determine semantic information. The approach further involves filtering the map data based, at least in part, on the restricted range of location information relevant to the at least one query, the semantic information, or a combination thereof.

Abstract: A computer-implemented method for identifying an expression for a target concept, includes: obtaining a set of texts as a target set of texts, with each text being associated with one of images relevant to a target concept. Candidate expressions for the target concept are extracted from the target set of texts. The candidate expressions are characteristic of the target set of texts. Each image relevant to one of the candidate expressions is collected by using an image search engine. A target expression for the target concept is selected from the candidate expressions based on a comparison result of the target concept and the collected images.

Abstract: Digital image search training techniques and machine-learning architectures are described. In one example, a query digital image is received by service provider system, which is then used to select at least one positive sample digital image, e.g., having a same product ID. A plurality of negative sample digital images is also selected by the service provider system based on the query digital image, e.g., having different product IDs. The at least one positive sample digital image and the plurality of negative samples are then aggregated by the service provider system into a single aggregated digital image. At least one neural network is then trained by the service provider system using a loss function based on a feature comparison between the query digital image and samples from the aggregated digital image in a single pass.

Abstract: Synchronized multi-system template provisioning is provided for networked computer systems, such as connected in a cloud environment. Each template includes information about corresponding templates and databases in other connected systems. When a template is deployed, user customization information is sent to the related systems to automatically deploy corresponding templates using the same or similar user customization information. Accordingly, when a template is deployed, the system checks for compatible templates in related systems and sends all information necessary for deploying the corresponding customizable templates in the other systems.

Abstract: A method and system for constructing an evolving ontology database. The method includes: receiving a plurality of data entries; calculating semantic similarity scores between any two of the data entries; clustering the data entries into a multiple current themes based on the semantic similarity scores; selecting, new concepts from the current themes by comparing the current themes with a plurality of previous themes prepared using previous data entries; and updating the evolving ontology database using the new concepts. The semantic score between any two of the data entries are calculated by: semantic similarity score=?i=0nsie?j=0kwj×f j, where si is weight of features sources, fj is a feature similarity between the two of the data entries, wj is a weight of fj, and j, k and n are positive integers.

Abstract: Embodiments regard transfer of data streaming services to provide continuous data flow. An embodiment of an apparatus includes one or more processors to process data for database operation and a database storage, wherein the system is to: receive one or more process flows, each process flow including a plurality of nodes and including one or more flows between each of the plurality of nodes and another node of the plurality of nodes; and upon receipt of a request, automatically rearrange the one or more process flows, including the apparatus to evaluate a first process flow including traversing of each node of the first process flow, and generate a rearranged process flow based on the first process flow, including establishing a level and coordinate position for each node of first process flow, and establishing a logical direction for each flow between the plurality of nodes of the first process flow.

Abstract: Embodiments of the present disclosure are directed to methods and systems for the timely processing of records exchanged between service provider systems and responsible entity systems by a records management and processing system. More specifically, the records management and processing system can maintain a set of rules defining conditions for processing records and associated actions to affect that processing upon satisfaction of or failure to satisfy the conditions of that rule. Updates to some or all of these records can be received from a data update service. The records management and processing system can apply the rules to the records and assign tags to the records based on the received update information and the conditions defined in the applied rules. The records management and processing system can then process the updated records according to workflows for processing the updated records based on the assigned tags and applied rules.

Abstract: Techniques are provided for identifying anomalies in an Internet of Things (IoT) activity profile of a user using an analytic engine. An exemplary method comprises obtaining data from a plurality of IoT devices of a user, wherein at least one IoT device comprises an agent device that performs an action on behalf of the user; applying the obtained data to a feature engineering module to convert the obtained data into time-series features that capture behavior and/or characteristics of an IoT environment of the user, and applying the time-series features to an analytic engine comprising a multi-variate anomaly detection method that learns one or more patterns in the IoT activity profile of the user for a normal state and identifies an anomaly with respect to an action performed by the agent device based on a health score indicating a deviation from the learned patterns.

Abstract: Provided are a computer program product, system, and method for determining a storage pool in which to store a data object. Event messages are received. Each event message includes metadata on an object updated in a storage pool of a plurality of storage pools, at least two of the storage pools have different reliability ratings. The event messages are added to a message queue. Information is added to a database for the event messages in the message queue including the metadata in the event message. A determination is made of a data classification of a changed data object indicated in the database. The changed data object indicated in the database having the determined data classification is moved to a storage pool having a reliability rating satisfying a reliability rating associated with the data classification of the changed data object.

Abstract: A mechanism is provided in a data processing system comprising a processor and a memory, the memory comprising instructions that are executed by the processor to specifically configure the processor to implement a data set generator. A statistics gathering engine executing within the data set generator calculates statistics and metrics across multiple data dimensions for a plurality of clients and a plurality of each client's data sources. The statistics gathering engine stores the statistics and metrics in a client-specific data map data structure for each client. An aggregated data map engine executing within the data set generator aggregates the statistics and metrics into an aggregated data map. The aggregated data map engine stores the aggregated data map in an aggregated data map data structure. The data set generator creates a client database construct.

Abstract: Techniques described herein are directed to determining an effective tag for data assets. For instance, each tag associated with a data asset may be associated with certain attributes. Non-limiting examples for such attributes may include a confidence level associated with a source that applied the tag, an applied date at which the tag was applied to the data asset, and/or a hierarchical depth of the data asset to which the tag was applied, as well as additional and/or alternative types of attributes. When a request to determine a tag for a data asset is received, the attributes for all the tags applied to the data asset may be retrieved and/or analyzed to determine which of such tags effectively classifies the data asset. The determined effective tag may be returned to the requesting entity.

Abstract: Data provenance techniques are provided using distributed ledgers.

Abstract: An example operation may include one or more of determining a data point trigger has occurred at a particular block cycle of a blockchain, initiating a sidechain to store subsequent entries based on the data point trigger, and a genesis block of the sidechain includes one or more relevant data fields from the blockchain, initiating a sidechain smart contract to manage data entries submitted to the sidechain, storing the data entries in the sidechain for a conditional period, and when the conditional period has matured, convoluting the sidechain into the blockchain.

Abstract: Systems and methods for data analytics include retrieving a first data model that includes a first set of one or more entities. A respective entity of the first set of one or more entities relates to a data subset of a first set of one or more databases, and corresponds to a metric, a dimension, or a filter. Based on the first data model, a training set is generated for training a first agent. The first agent is configured to respond to user input queries formulated in natural language. The training set for training the first agent includes a plurality of sample requests, and a plurality of database queries for the one or more databases. At least one respective database query of the plurality of database queries corresponds to at least one respective sample request of the plurality of sample requests.