Abstract: Aspects of the present disclosure relate to annotating or tagging customer data. In some embodiments, the annotating can include summarizing touchpoints into k-hot encoding feature vectors, mapping the feature vectors onto an embedding layer, predicting a hierarchical data sequence using the embedding layer and the feature vectors, extracting the feature vectors that are most influential in predicting the embedding layer, and outputting the touchpoints associated with the most influential feature vectors.

Abstract: Aspects of the present disclosure relate to identifying friction points in customer data. In some embodiments, identifying friction points can include receiving a set of input sequence data and predicted class labels for the set of input sequence data; selecting input sequences, from the set of input sequence data, that have class labels matching a ground truth class label; reducing the selected sequences to anchor points; and grouping the reduced selected sequences into critical data set signatures using discriminatory subsequence mining.

Abstract: An action recommendation system uses reinforcement learning that provides a next action recommendation to a traffic controller to give to a vehicle pilot such as an aircraft pilot. The action recommendation system uses data of past human actions to create a reinforcement learning model and then uses the reinforcement learning model with current ABS-B data to provide the next action recommendation to the traffic controller. The action recommendation system may use an anisotropic reward function and may also include an expanding state space module that uses a non-uniform granularity of the state space.

Abstract: Embodiments for implementing customized traversal routes by at least a portion of a processor in a computing environment. One or more traversal routes may be generated from a route network for a user to traverse according to one or more parameters and user preferences. One or more edge metrics may be assigned to the one or more traversal routes while periodically updating the one or more parameters, the user preferences, or a combination thereof associated with the one or more traversal routes.

Abstract: Systems, methods, and computer-readable media are described for automatically identifying potential errors in the text output of a domain-agnostic speech-to-text engine and generating text snippets that contain words representative of the potential errors and other words in the neighborhoods of such words for context. In this manner, a substantially reduced amount of text (i.e., the text snippets) can be reviewed for errors in the speech-to-text conversion rather than the entire text output, thereby significantly reducing the burden associated with error identification in the text output.

Abstract: A parse tree corresponding to a portion of narrative text is constructed. The parse tree includes a data structure representing a syntactic structure of the portion of narrative text as a set of tokens according to a grammar. Using a token in the parse tree as a focus word, a context window comprising a set of words within a specified distance from the focus word is generated, the distance determined according to a number of links of the parse tree separating the focus word and a context word in the set of words. A weight is generated for the focus word and the context word. Using the weight, a first vector representation of a first word is generated, the first word being within a second portion of narrative text.

Abstract: Embodiments for implementing intelligent customer journey prediction and customer segmentation of a processor in a computing environment. A response outcome of a customer journey for a user may be predicted according to an assigned score based on one or more discriminatory sequence patterns identified between one or more groups of customers using one or more machine learning operations.

Abstract: Embodiments for implementing customized traversal routes by at least a portion of a processor in a computing environment. One or more traversal routes may be generated from a route network for a user to traverse according to one or more parameters and user preferences. One or more edge metrics may be assigned to the one or more traversal routes while periodically updating the one or more parameters, the user preferences, or a combination thereof associated with the one or more traversal routes.

Abstract: Systems, methods, and computer-readable media are described for automatically identifying potential errors in the text output of a domain-agnostic speech-to-text engine and generating text snippets that contain words representative of the potential errors and other words in the neighborhoods of such words for context. In this manner, a substantially reduced amount of text (i.e., the text snippets) can be reviewed for errors in the speech-to-text conversion rather than the entire text output, thereby significantly reducing the burden associated with error identification in the text output.

Abstract: A computer-implemented method (and structure) includes receiving information. The received information is converted into a strictly hierarchical data format. A precision for a releasing the strictly hierarchical data is calculated based on privacy protection levels and a reward for different precision levels. The strictly hierarchical data is sequentially released at the calculated precision.

Abstract: Embodiments of the invention include method, systems and computer program products for document vectorization. Aspects include receiving, by a processor, a plurality of documents each having a plurality of word. The processor utilizing a vector embeddings engine generates a vector to represent each of the plurality of words in the plurality of documents. An image representation for each document in the plurality of documents is created and a word probability for each of the plurality of words in the plurality of documents is generated. A position for each word probability is determined in the image based on the vector associated with each word and a compression operation on the images is performed to produce a compact representation for the plurality of documents.

Abstract: An input dataset for training a new machine learning model is received by a processor. For each of a plurality of trained machine learning models, a hash function and a sketch of a training dataset used to train the machine learning model is retrieved. A sketch of the input dataset is computed based on the hash function and the input dataset, along with a distance between the sketch of the training dataset and the sketch of the input dataset. The computed distances of the trained machine learning models are ranked from smallest to largest, and a seed machine learning model for the input dataset is selected from the trained machine learning models based at least in part on the ranking. A training process of the new machine learning model using the selected seed machine learning model and the input dataset is initiated.

Abstract: Techniques for inserting and extracting geolocation data using spatial indexing in a key value database are provided. In an embodiment, a system is provided for generating one or more geohashes for a geometry object, wherein the one or more geohashes comprises encoded bits that are stored as keys in a key value database. In one example, the system comprises a geometry indexing component that generates a spatial index, wherein the spatial index is based on a total number of the encoded bits generated for the one or more geohashes. In one example, the system comprises a geometry storing component that stores the geometry object and the one or more geohashes in the key value database using the spatial index to allow for faster retrieval of the geometry object. The advantage is that properly inserted and indexed spatial data can be quickly retrieved.

Abstract: A common infrastructure collects data from a plurality of mobile devices and traditional sensors at Internet scale to respond to natural language queries received at different applications. The infrastructure includes a semantic interpreter to translate the natural language query to a data request specification that is processed by the data collection system. The data collection system includes a phenomenon layer that expresses data and information needs in a declarative fashion and coordinates data collection and processing for queries. An edge layer manages devices, receives collection requirements from the backend layer, configures and instructs devices for data collection, and conducts aggregation and primitive processing of data. This layer contains network edge nodes, such as base stations in a cellular network. Each node manages a set of local data generating networked devices.

Abstract: A common infrastructure collects data from a plurality of mobile devices and traditional sensors at Internet scale to respond to natural language queries received at different applications. The infrastructure includes a semantic interpreter to translate the natural language query to a data request specification that is processed by the data collection system. The data collection system includes a phenomenon layer that expresses data and information needs in a declarative fashion and coordinates data collection and processing for queries. An edge layer manages devices, receives collection requirements from the backend layer, configures and instructs devices for data collection, and conducts aggregation and primitive processing of data. This layer contains network edge nodes, such as base stations in a cellular network. Each node manages a set of local data generating networked devices.

Abstract: A common infrastructure collects data from a plurality of mobile devices and traditional sensors at Internet scale to respond to natural language queries received at different applications. The infrastructure includes a semantic interpreter to translate the natural language query to a data request specification that is processed by the data collection system. The data collection system includes a phenomenon layer that expresses data and information needs in a declarative fashion and coordinates data collection and processing for queries. An edge layer manages devices, receives collection requirements from the backend layer, configures and instructs devices for data collection, and conducts aggregation and primitive processing of data. This layer contains network edge nodes, such as base stations in a cellular network. Each node manages a set of local data generating networked devices.

Abstract: A spatial-temporal storage method, system, and non-transitory computer readable medium, include, in a first layer, a geometric translation circuit configured to split spatial-temporal information into row keys and translate a geometry query into a range scan, and a multi-scan optimization circuit configured to compute an optimal read strategy to optimize the range scan translated by the geometric translation circuit into a series of block starting offsets and block sizes, and, in a second layer, a block grouping circuit configured to allow grouping of blocks in the second layer while preserving spatial data locality when splits of spatial-temporal information occur in the first layer.

Abstract: A spatial-temporal storage method, system, and non-transitory computer readable medium, include, in a first layer, a geometric translation circuit configured to split spatial-temporal information into row keys and translate a geometry query into a range scan, and a multi-scan optimization circuit configured to compute an optimal read strategy to optimize the range scan translated by the geometric translation circuit into a series of block starting offsets and block sizes, and, in a second layer, a block grouping circuit configured to allow grouping of blocks in the second layer while preserving spatial data locality when splits of spatial-temporal information occur in the first layer.

Abstract: A computer-implemented method (and structure) includes receiving information. The received information is converted into a strictly hierarchical data format. A precision for a releasing the strictly hierarchical data is calculated based on privacy protection levels and a reward for different precision levels. The strictly hierarchical data is sequentially released at the calculated precision.

Abstract: A spatial-temporal storage method, system, and non-transitory computer readable medium, include, in a first layer, a geometric translation circuit configured to split spatial-temporal information into row keys and translate a geometry query into a range scan, and a multi-scan optimization circuit configured to compute an optimal read strategy to optimize the range scan translated by the geometric translation circuit into a series of block starting offsets and block sizes, and, in a second layer, a block grouping circuit configured to allow grouping of blocks in the second layer while preserving spatial data locality when splits of spatial-temporal information occur in the first layer.