Abstract: A method, system, and computer program product for automated artificial intelligence (AI) factsheet generation for modeling and model customization in speech to text (STT) services. The method receives audio data for a user. The audio data contains human speech. Text data is generated, using a first speech to text model, to represent the human speech of the audio data. A set of transcription errors of the first speech to text model are identified. A set of AI factsheets are generated to describe model metadata for the first speech to text model. Based on the set of transcription errors and the set of AI factsheets, the method generates a second speech to text model customized to the user.

Abstract: An automatic speech recognition (ASR) computing system and methodology are provided to predict a textual representation of received input speech data. A context acoustic biasing (CAB) engine of the ASR computing system receives historical textual content and an ontology data structure. The CAB engine matches key terms identified in the historical textual content with concepts present in the ontology data structure to generate a contextual term list data structure comprising the concept terms related to concepts matching the key terms. The CAB engine generates acoustic representations of the concept terms in the contextual term list data structure and inputs them to an ASR computer model of the ASR computing system which processes an input speech signal to generate a predicted textual representation of the input speech signal. The predicted textual representation is biased towards the acoustic representations of the concept terms in the contextual term list data structure.

Abstract: Methods, systems, and computer program products for multi-modal construction of deep learning networks are provided herein. A computer-implemented method includes extracting, from user-provided multi-modal inputs, one or more items related to generating a deep learning network; generating a deep learning network model, wherein the generating includes inferring multiple details attributed to the deep learning network model based on the one or more extracted items; creating an intermediate representation based on the deep learning network model, wherein the intermediate representation includes (i) one or more items of data pertaining to the deep learning network model and (ii) one or more design details attributed to the deep learning network model; automatically converting the intermediate representation into source code; and outputting the source code to at least one user.

Abstract: Systems and methods are disclosed to implement a contextual comparison of machine registry hive files. In embodiments, the comparison process is implemented by a data collection agent that periodically uploads changes in a client machine registry to a machine assessment service. During a data collection, the agent compares a binary hive file generated from the current state of the registry with another binary hive file generated in the last period. The differences are captured in a text-encoded patch file, which is used to update a snapshot of the registry maintained by the machine assessment service. The comparison is performed directly on the two binary hive files without converting them into text files, so that the process can be performed more quickly and using less compute bandwidth. Moreover, the comparison process can be extended to implement a variety of custom behaviors based on the contents of the hive files.

Abstract: Methods, systems, and computer program products for automated domain-specific constrained decoding from speech inputs to structured resources are provided herein.

Abstract: A method, computer system, and a computer program product for debugging a deep neural network is provided. The present invention may include identifying, automatically, one or more debug layers associated with a deep learning (DL) model design/code, wherein the identified one or more debug layers include one or more errors, wherein a reverse operation is introduced for the identified one or more debug layers. The present invention may then include presenting, to a user, a debug output based on at least one break condition, wherein in response to determining the at least one break condition is satisfied, triggering the debug output to be presented to the user, wherein the presented debug output includes a fix for the identified one or more debug layers in the DL model design/code and at least one actionable insight.

Abstract: Systems and methods are disclosed to implement a delta data collection technique for collecting machine characteristics data from client machines. In embodiments, the collected data is used by a machine assessment service to maintain a virtual representation of the client machine for assessments. To initialize the collection process, the client uploads an initial copy of the data in full. Subsequently, the client determines periodic deltas between a current baseline of the data and a last reported baseline, and the deltas are uploaded as patches. The machine assessment service then applies these patches to update the virtual representation of the client machine. In embodiments, to facilitate the generation or uploading of the patches, the client may generate the baselines in a different encoding format as used by the data. For example, baselines in the new encoding format may be more easily compared and manipulated during the patch generation process.

Abstract: One embodiment provides a method, including: mining a plurality of deep-learning models from a plurality of input sources; extracting information from each of the deep-learning models, by parsing at least one of (i) code corresponding to the deep-learning model and (ii) text corresponding to the deep-learning model; identifying, for each of the deep-learning models, operators that perform operations within the deep-learning model; producing, for each of the deep-learning models and from (i) the extracted information and (ii) the identified operators, an ontology comprising terms and features of the deep-learning model, wherein the producing comprises populating a pre-defined ontology format with features of each deep-learning model; and generating a deep-learning model catalog comprising the plurality of deep-learning models, wherein the catalog comprises, for each of the deep-learning models, the ontology corresponding to the deep-learning model.

Abstract: Systems and methods are disclosed to implement a delta data collection technique for collecting machine characteristics data from client machines. In embodiments, the collected data is used by a machine assessment service to maintain a virtual representation of the client machine for assessments. To initialize the collection process, the client uploads an initial copy of the data in full. Subsequently, the client determines periodic deltas between a current baseline of the data and a last reported baseline, and the deltas are uploaded as patches. The machine assessment service then applies these patches to update the virtual representation of the client machine. In embodiments, to facilitate the generation or uploading of the patches, the client may generate the baselines in a different encoding format as used by the data. For example, baselines in the new encoding format may be more easily compared and manipulated during the patch generation process.

Abstract: Techniques for the suggestion and completion of deep learning models are disclosed including receiving a set of data and determining at least one property of the data. A plurality of characteristics of a computing device and a plurality of deep learning models are received and a score for each of the plurality of deep learning models is determined based on the received computing device characteristics and the determined at least one property of the data. The plurality of deep learning models are ranked for presentation to a user based on the determined scores. One or more of the deep learning models are presented on a display based on the ranking. A selection of one of the deep learning models is received and the selected deep learning model is trained using the set of data.

Abstract: Disclosed herein are methods, systems, and processes to perform granular and selective agent-based throttling of command executions. A resource consumption threshold is allocated to an agent process that is configured to perform data collection tasks on a host computing device. A desired throttle is generated for the agent process based on the resource consumption threshold allocated to the agent process and execution of the agent process is controlled in polling intervals. For each polling interval, a current throttle level for the agent process is determined based on a run count and a skip count of the agent process, the agent process is suspended if the agent process is active and the current throttle is greater than the desired throttle level, and the agent process is resumed if the agent process is idle and the current throttle level is not greater than the desired throttle level.

Abstract: Disclosed herein are methods, systems, and processes to perform granular and selective agent-based throttling of command executions. A resource consumption threshold is allocated to an agent process that is configured to perform data collection tasks on a host computing device. A desired throttle is generated for the agent process based on the resource consumption threshold allocated to the agent process and execution of the agent process is controlled in polling intervals. For each polling interval, a current throttle level for the agent process is determined based on a run count and a skip count of the agent process, the agent process is suspended if the agent process is active and the current throttle is greater than the desired throttle level, and the agent process is resumed if the agent process is idle and the current throttle level is not greater than the desired throttle level.

Abstract: One embodiment provides a method, including: receiving a training dataset to be utilized for training a deep-learning model; identifying a plurality of aspects of the training dataset, wherein each of the plurality of aspects corresponds to one of a plurality of categories of operations that can be performed on the training dataset; measuring, for each of the plurality of aspects, an amount of variance of the aspect within the training dataset; creating additional data to be incorporated into the training dataset, wherein the additional data comprise data generated for each of the aspects having a variance less than a predetermined amount, wherein the data generated for an aspect results in the corresponding aspect having an amount of variance at least equal to the predetermined amount; and incorporating the additional data into the training dataset.

Abstract: Disclosed herein are methods, systems, and processes to perform granular and selective agent-based throttling of command executions. A polling interval of an agent process executing on a protected host is monitored. If the agent process is active and a current throttle is greater than a desired throttle, the agent process and its children processes are suspended and a run count flag is incremented. However, if the agent process is inactive and the current throttle is less than or equal to the desired throttle, the agent process and its children processes are resumed and a skip count flag is incremented.

Abstract: Embodiments of the present invention provide a method, a computer program product, and a system for generating a haptic signal representing a fabric composition. Embodiments of the present invention can be used to generate a haptic signal that is based on a user selection. For example, embodiments of the present invention can combine characteristic signals corresponding to a plurality of textiles to generate the haptic signal for output to a haptic device. Embodiments of the present invention can be used to recommend similar fabric compositions based upon similarity between a characteristic signal of a fabric composition and the haptic signal.

Abstract: One embodiment provides a method, including: mining a plurality of deep-learning models from a plurality of input sources; extracting information from each of the deep-learning models, by parsing at least one of (i) code corresponding to the deep-learning model and (ii) text corresponding to the deep-learning model; identifying, for each of the deep-learning models, operators that perform operations within the deep-learning model; producing, for each of the deep-learning models and from (i) the extracted information and (ii) the identified operators, an ontology comprising terms and features of the deep-learning model, wherein the producing comprises populating a pre-defined ontology format with features of each deep-learning model; and generating a deep-learning model catalog comprising the plurality of deep-learning models, wherein the catalog comprises, for each of the deep-learning models, the ontology corresponding to the deep-learning model.

Abstract: One embodiment provides a method, including: receiving input of a learning session that is being conducted by an educator, being provided to at least one user, and being related to a subject; determining, using a knowledge base, that at least one topic relevant to the subject of the learning session is incomplete, wherein the determining comprises building a knowledge subgraph of the learning session and comparing the built knowledge subgraph to at least a portion of the knowledge base; generating at least one question to be asked of the educator relevant to the at least one incomplete topic; identifying, using at least one natural language text classifier model, a location within the learning session to ask the generated at least one question; and providing, to the educator, an output corresponding to the at least one question at the identified location within the learning session.

Abstract: A method, computer system, and a computer program product for debugging a deep neural network is provided. The present invention may include identifying, automatically, one or more debug layers associated with a deep learning (DL) model design/code, wherein the identified one or more debug layers include one or more errors, wherein a reverse operation is introduced for the identified one or more debug layers. The present invention may then include presenting, to a user, a debug output based on at least one break condition, wherein in response to determining the at least one break condition is satisfied, triggering the debug output to be presented to the user, wherein the presented debug output includes a fix for the identified one or more debug layers in the DL model design/code and at least one actionable insight.

Abstract: One embodiment provides a method, including: providing, at a collaborative deep learning model authoring tool, a dialog window that (i) receives user inputs discussing deep learning model aspects and (ii) provides recommendations from the collaborative deep learning model authoring tool; providing, at the collaborative deep learning model authoring tool, a consensus view indicating (i) a conflicting aspect identified as an aspect where more than one user selected a different aspect and (ii) the aspect selected for implementation within the deep learning model based upon that aspect having the most user selections; providing, at the collaborative deep learning model authoring tool, a model view displaying layers of the deep learning model based upon (i) aspects selected by the users in the dialog window and (ii) the aspect selected for implementation in the consensus view; and providing, at the collaborative deep learning model authoring tool, a deployment view that displays an execution of the deep learning m

Abstract: Techniques for the suggestion and completion of deep learning models are disclosed including receiving a set of data and determining at least one property of the data. A plurality of characteristics of a computing device and a plurality of deep learning models are received and a score for each of the plurality of deep learning models is determined based on the received computing device characteristics and the determined at least one property of the data. The plurality of deep learning models are ranked for presentation to a user based on the determined scores. One or more of the deep learning models are presented on a display based on the ranking. A selection of one of the deep learning models is received and the selected deep learning model is trained using the set of data.