Abstract: A device may receive a merchant query including first merchant data associated with a first merchant. The first merchant data may be provided, as input, to a merchant matching model associated with a merchant data structure, the merchant matching model having been trained to determine a measure of confidence that input merchant data corresponds to an existing merchant in the merchant data structure. The device may receive, as output from the merchant matching model, a measure of confidence that the first merchant data corresponds to a second merchant, the second merchant being associated with second merchant data stored in the merchant data structure. The device may also determine, based on the measure of confidence, that the first merchant corresponds to the second merchant. Based on the determination, the device may obtain the second merchant data from the merchant data structure and perform an action based on the second merchant data.

Abstract: A system and method for automatic plant monitoring. The method comprises: identifying at least one test input respective of a test area, wherein the test area includes at least one part of a plant; and generating a plant condition prediction based on the at least one test input and on a prediction model, wherein the prediction model is based on a training set including at least one training input and at least one training output, wherein each training output corresponds to a training input.

Abstract: A reasoning engine is disclosed. Contemplated reasoning engines acquire data relating to one or more aspects of various environments. Inference engines within the reasoning engines review the acquire data, historical or current, to generate one or more hypotheses about how the aspects of the environments might be correlated, if at all. The reasoning engine can attempt to validate the hypotheses through controlling acquisition of the environment data.

Abstract: Systems and methods for monitoring the quality of document reviews used in continuous active machine learning are described herein. Two orthogonal processes may be run simultaneously, asynchronously, and continuously. The first process performs continuous active machine learning for training machine classification models. The second process classifies documents that have been reviewed as part of the first process to generate classification scores of the reviewed documents. The original review may be compared to the classification scores using false negative and a false positive thresholds to identify documents that may have been incorrectly reviewed. A master review of identified documents is used to correct original reviews that were incorrect. Original incorrect reviews may be replaced in a training corpus by corrected reviews, and the models may be retrained using the corrected reviews.

Abstract: A number of attributes of different attribute types, to be used to assign observation records of a data set to clusters, are identified. Attribute-type-specific distance metrics for the attributes, which can be combined to obtain a normalized aggregated distance of an observation record from a cluster representative, are selected. One or more iterations of a selected clustering methodology are implemented on the data set using resources of a machine learning service until targeted termination criteria are met. A given iteration includes assigning the observations to clusters of a current version of a clustering model based on the aggregated distances from the cluster representatives of the current version, and updating the cluster representatives to generate a new version of the clustering model.

Abstract: A method for creating a score or value based on the difference between non-expert and expert usage based data. The score or value may be used in variety of situations such as assessing risk, training, operator classification, and identifying expertise level of an operator.

Abstract: This disclosure relates generally to human-machine interaction. In one embodiment, an interaction device for providing the interaction between the user and the ECA is disclosed. The interaction device comprises a processor and a memory communicatively coupled to the processor. The memory stores processor instructions, which, on execution, causes the processor to receive conversation data of a user interacting with the ECA, wherein the ECA is presented on an interface of the interaction device. The processor further determines an emotional state of the user based on one or more behavioral parameters associated with the conversation data of the user. The processor identifies a response state for the ECA corresponding to the emotional state of the user, wherein the response state is identified from a plurality of response states based on a pre-defined probability for each response state. The processor further transitions behavior of the ECA based on the response state.

Abstract: Hybrid feature selection methods include methods of creating a predictive model for valve performance in a fleet of aircraft. Methods include qualifying a qualification dataset of valve-related parameters calculated from data collected during a first series of flights at least before and after a non-performance event of a valve. Methods include receiving a qualified selection of the valve-related parameters and verifying a verification dataset of the qualified selection of the valve-related parameters calculated from data collected during a second series of flights. Methods include receiving a set of verified and qualified valve-related parameters and building a predictive model for valve non-performance with a training dataset of the verified and qualified valve-related parameters calculated from data collected during additional flights of the fleet.

Abstract: A distributed multisensory system to record spatially diverse events configures a group of sensors searching sensed events utilizing random access search. The system utilizes presence and co-location detection to identify objects or people in an area. The system then processes audio signals utilizing a multimodal network and machine learning tiered with obfuscated text indexing to provide high-speed, high accuracy searching of stored data. In addition, the obfuscated text allows the system to not have to store an audio transcript, increasing the security of the system. The system utilizes beamforming to merge resultant audio streams for the recorded event.

Abstract: A processing device executing a rule engine receives a first object. The processing device determines whether the first object is represented by a relational model or an object-oriented model. The first object is determined to be represented by the relational model responsive to the first object lacking a reference to a nested object. The first object is determined to be represented by the object-oriented model responsive to the first object including a reference to a nested object. If the first object is represented by the relational model, a join is performed between the first object and a second object based on a relationship between the objects using a first node. If the first object is represented by the object-oriented model, an expression of the first object is evaluated to navigate to a third object that is a first nested object of the first object using a second node.

Abstract: Methods and systems for handling queries include extracting keywords related to a first query from one or more information sources. The keywords are classified according to a plurality of categories defined by a query schema. A plurality of the keywords are combined into a second query according to the query schema. The second query is executed to generate one or more results.

Abstract: Techniques are provided herein for utilizing a classification engine to improve a classification model. For example, a classification engine may derive a statistical model based at least in part on a synthetic data set. A misclassification may be determined based at least in part on an output of the statistical model. An audit question may be provided to an individual, the audit question being determined based at least in part on the determined misclassification. Response data related to the audit question may be received. The statistical model may be validated based at least in part on the response data.

Abstract: An application performance management system is provided, which is adapted to analyze the performance of one or more applications running on information technology (IT) infrastructure. The application performance management system includes a data collector, an anomaly detector, an anomaly correlator, an anomaly ranking unit, and a source problem detector. The data collector collects performance metrics for one or more applications running on the IT infrastructure. The anomaly detector analyzes the performance metrics and detects anomalies, which may include performance metrics whose values deviate from historic values with a deviation that exceeds a predefined threshold. The anomaly correlator detects dependencies between plural anomalies and generates anomaly clusters. Each anomaly cluster includes anomalies that are correlated through one or more of the dependencies. The anomaly ranking unit ranks anomalies within an anomaly cluster.

Abstract: In general, embodiments of the present invention provide systems, methods and computer readable media for curating a training data set to ensure that training data being updated continuously from a data reservoir of verified possible training examples remain an accurate, high-quality representation of the distribution of data that are being input to a predictive model for processing.

Abstract: A system for generating a large-scale database of heterogeneous speech is provided. The system comprises a processor a plurality of independent computation cores configured to generate signatures of a plurality of speech segments; a large scale database configured to maintain a plurality of transcribed multimedia signals; a memory, the memory containing instructions that, when executed by the processor, configure the system to: randomly select a plurality of speech segments from the plurality of multimedia signals, wherein each speech segment of the plurality of speech segments is of a random length; provide the plurality of speech segments to the plurality of independent computation cores for generation of the signatures; collect the signatures from the plurality of independent computation cores; and populate the large-scale database with the plurality of signatures respective of the plurality of multimedia signals.

Abstract: Particle Thompson Sampling for online matrix factorization recommendation is described. In one or more implementations, a recommendation system provides a recommendation of an item to a user using Thompson Sampling. The recommendation system then receives a rating of the item from the user. Unlike conventional solutions which only update the user latent features, the recommendation system updates both user latent features and item latent features in a matrix factorization model based on the rating of the item. The updating is performed in real time which enables the recommendation system to quickly adapt to the user ratings to provide new recommendations. In one or more implementations, to update the user latent features and the item latent features in the matrix factorization model, the recommendation system utilizes a Rao-Blackwellized particle filter for online matrix factorization.

Abstract: Embodiments of the invention provide techniques which utilize crowdsourced reports of environmental conditions to predict and/or prevent disease outbreaks. In one aspect, a method comprises receiving one or more crowdsourced reports about one or more environmental conditions; inferring one or more input parameters for at least one disease outbreak model based at least in part on the one or more crowdsourced reports; applying the at least one disease outbreak model to at least the one or more inferred parameters to predict one or more characteristics of at least one potential disease outbreak associated with the reported one or more environmental conditions; and, based at least in part on the predicted one or more characteristics, implementing one or more corrective actions to mitigate the at least one potential disease outbreak.

Abstract: An analog implementation is proposed of an adaptive signal processing model of a kind requiring a plurality of randomly-set variables. In particular, following a digital to analog conversion of a digital input signal, analog processing is used to transform the data input to the model into data which is subsequently processed by an adaptively-created layer of the model. In the analog processing, multiplication operations involving the randomly-set variables are performed by analog circuitry in which the randomly-set variables are the consequence of inherent tolerances in electrical components. This eliminates the need for the randomly-set variables to be implemented in some other way, for example as random variables stored in memory.

Abstract: A method for using a plurality of decision engines to produce a single decision comprises: receiving heterogeneous data from the plurality of decision engines comprising a set of class labels from each of the plurality of decision engines, wherein the set of class labels from at least a first decision engine differs from the set of class labels from at least a second decision engine; generating a single set of unified class labels from the heterogeneous data; calculating at least one value corresponding to each of at least a subset of the unified class labels; and performing decision fusion on at least the subset of the unified class labels and corresponding values to produce the single decision.

Abstract: Systems and methods of intelligent formation and acquisition of machine learning training data for implementing an artificially intelligent dialogue system includes constructing a corpora of machine learning test corpus that comprise a plurality of historical queries and commands sampled from production logs of a deployed dialogue system; configuring training data sourcing parameters to source a corpora of raw machine learning training data from remote sources of machine learning training data; calculating efficacy metrics of the corpora of raw machine learning training data, wherein calculating the efficacy metrics includes calculating one or more of a coverage metric value and a diversity metric value of the corpora of raw machine learning training data; using the corpora of raw machine learning training data to train the at least one machine learning classifier if the calculated coverage metric value of the corpora of machine learning training data satisfies a minimum coverage metric threshold.