Abstract: A method, apparatus, system, and computer program code for dynamically modeling multi-tenant data in a machine learning platform. A recommendation engine receives a first data set from a user. The recommendation engine characterizes the first data set to determine data attributes and data characteristics of the first data set. The recommendation engine aligns the data attributes of the first data set with a second data set according to an ontology. Based on the data characteristics of the first data set, the recommendation engine identifies a set of pre-trained models that was trained from training parameters selected from data attributes and data characteristics of a second data set. The recommendation engine recommends the set of pre-trained models to the user.

Abstract: An illustrative embodiment provides a computer-implemented method. The method comprises using a processor set to receive a number of documents from a plurality of data sources. The processor set annotates entities in the first subset of documents from the number of documents. The processor set splits the number of documents into a number of chunks. The processor sets indexes the number of chunks according to an index structure to generate a number of indexed chunks. The processor set generates a number of prompt examples for each document in the second subset of documents based on the number of indexed chunks and annotated entities from the first subset of documents. The processor set generates a prompt for extracting entities using a large language model based on the number of prompt examples.

Abstract: A computer-implemented method for calculating risk is provided. The method comprises receiving, at a first time, respective deterministic values for a number of factors. Over an evolution time interval starting at the first time, respective stochastic quantum values are generated for each factor from the deterministic values. Each factor is represented by a respective N-qubit quantum circuit, wherein each qubit represents a risk level. The quantum values for the factors are combined into an intermediate quantum risk value at a second time at the end of the evolution time interval. The intermediate quantum risk value according is then modified to a number of classical parameters, and a final stochastic quantum risk value at the second time is generated from the modified intermediate quantum risk value.

Abstract: Detecting geofence break events is provided. The method comprises defining a geofence boundary according to a triangle comprising two lines from the center of the earth to two vertices above the earth's surface, and a third line between the vertices above the earth's surface. The geofence boundary is defined by a vector between respective intersection points of the first two lines with the earth's surface. This vector represents a great arc when projected to the curvature of the earth's surface. Origin and destination coordinates for a vehicle are received, wherein a second vector between the origin and destination coordinates represents a great arc defining a path of the vehicle. Responsive to a determination that the second vector intersects the plane of the triangle, a geofence break event notice is sent to a number of downstream consumers.

Abstract: A computer implemented method for managing tokens on a blockchain platform. The method uses a number of processor units to receive a first set of attributes associated with a number of items and a second set of attributes associated with a number of offers to purchase the number of items. The number of processor units creates an item token on the blockchain platform for each item from the number of items based on the first set of attributes to form a number of item tokens. The number of processor units create a request token on the blockchain platform for each offer to purchase based on the second set of attributes to form a number of request tokens. The number of processor units create a number of matching tokens and a number of trades based on the matching tokens.

Abstract: A method for identifying relationships between nodes in a graph is provided. The method comprises determining a count of relationships for each relationship type associated with each node in a pair of nodes on the graph. The relationship types are represented by edges connected to the nodes in the pair of nodes. A number of common relationships are determined between the pair of nodes for each relationship type. A count for each common relationship is determined for the number of common relationships between the pair of nodes. A similarity coefficient for each common relationship is determined based on the count for each common relationship and the count of relationships for each relationship type associated with each node in the pair of nodes. A weighted score is generated to represent strength for relationships between the pair of nodes using the similarity coefficients.

Abstract: A computer-implemented method of providing information to a user is provided. The method comprises receiving input of a selected operating mode and extracting, via an extraction pipeline, data from a number of data sources according to the selected operating mode. The extracted data is fed into a large language model (LLM), and the LLM generates verbal and auditory information for the user based on the data. The LLM conveys the verbal and auditory information to the user via a number of interface products according to the selected operating mode.

Abstract: Classifying cybersecurity signals from media sources into distinct categories is provided. The method comprises receiving a first data subset comprising data points labeled by subject matter experts according to a predetermined number of specified categories. The data points include information regarding cybersecurity from a set of news articles. The first subset is enriched by applying a random forest algorithm to generate synthetic data points, thereby deriving a second data subset that is augmented from the first subset. The combined first and second data subsets comprise an enhanced training dataset. A BERT model is trained with the enhanced training dataset to classify cybersecurity-related news according to the specified categories. The BERT model utilizes a specialized vector database integrating domain-specific cyber-related terminology and contextual embeddings. The trained BERT model classifies a second set of news articles according to the specified categories.

Abstract: Dynamically down sampling vehicle tracking data is provided. The method comprises receiving an original set of position data points from a vehicle, including a first and last data point. Each data point includes a timestamp. A path for the vehicle in four-dimensional hyperspace is simplified from the original set of points according to the Ramer-Douglas-Peucker algorithm. Positions are converted to three-dimension coordinates on a sphere with the timestamp comprising the fourth dimension. The simplified path comprises a reduced set of points that describe a minimum contiguous path of line segments between the first and last data points. All other position points from the original set satisfy a proximity criterion with regard to the minimum contiguous path. The reduced set represents the original set within a specified hyperspace tolerance and has a lower transmission bandwidth requirement. The reduced set of position points is then transmitted to users.

Abstract: A method for identifying a parent company in a corporate hierarchy is provided. A neural network embeds a number of company names into respective numeric vectors and then indexes the vectors into an approximate nearest neighbors (ANN) index according to euclidean distances between them. In response to a query regarding a parent company of a specified query company, a top N number of nearest neighbor companies to the query company is extracted from the ANN index. A machine learning voting model determines which, if any, of the extracted nearest neighbor companies has a parent company that best corresponds to the query company. If an extracted nearest neighbor company has a parent company that best corresponds to the query company, the parent company is displayed to a user through a user interface.

Abstract: A method, apparatus, system, and computer program product for method for dynamically managing an ontology for classifying data is provided. The ontology is generated from the classifications of a plurality of entities. Models are generated that predict classifications according to the ontology and natural language descriptions of the entities. Unclassified entities are modelled by according to the models to identify at least one classification within the ontology. Responsive to identifying a plurality of probable classifications within the ontology, a new classification is generated based on the plurality of probable classifications, an updated ontology is generated that includes the new classification, and the unclassified entity is classified according to the new classification.

Abstract: Testing prototype indices is provided. The method comprises receiving input of a number of historical index values and performing a number of Monte Carlo simulations based on the historical index values to predict future index values. A number of attributes of the historical index values are calculated, and the Monte Carlo simulations are correlated with the attributes to identify a filtered subset of predicted future index values having an accuracy above a defined threshold. The filtered subset of predicted future index values are fed into a deep neural network (DNN) that outputs a confidence score for each of the filtered subset of predictive future index values.

Abstract: A method for point-in-time visualization system management is provided. The method includes receiving organizational structure queries from user interfaces, creating organizational hierarchy database queries from the organizational structure query, and converting the results of those queries into entity change parameters lists and organizational tree structures for display on the user interface. The displayed organizational tree structure and entity change parameters can be filtered via controls available on the user interface.

Abstract: A method, apparatus, system, and computer product for monitoring emission data transmitted over a network. A computer system collects emission data received from sensor devices over the network. The computer system compares the emission data to a set of thresholds for the emission data to form a comparison between the emission data and the set of thresholds. The computer system changes a collection of additional emission data from the sensor devices over the network based on the comparison between the emission data and the set of thresholds.

Abstract: A method, system, and product for multi-class text classification. The method includes receiving an uncoded shipping container description comprising an uncoded text; and assigning a predicted HS code segment to the uncoded container description. Assigning includes cleaning the uncoded text of the uncoded container description by removing 1-2 letter words, numbers and symbols from the text; tokenizing the cleaned uncoded text to define a plurality of unencoded tokens by parsing the cleaned uncoded text into single words and bigrams; summing the scores for each of a plurality of HS code segments associated with each of the plurality of unencoded tokens across all of the plurality of unencoded tokens; and determining the predicted HS code segment based on the highest summation. The method can include receiving an uncoded shipping container description comprising an uncoded text; and assigning a predicted HS code segment to the uncoded container description.

Abstract: A method of textual data analysis is provided. The method comprises parsing text data extracted from transcripts related to a number of companies. Intermediate metrics are created comprising numerical representations of the parsed text data and derivations. The intermediate metrics are then combined into different combinations comprising headline analytics. A machine learning model tests each headline analytic for standalone predictive efficacy. Headline analytics with standalone predictive efficacies above a first threshold are selected and then tested for additive predictive efficacy to determine if the selected headline analytic incrementally increases the predictive efficacy of a preexisting economic analytic above a second threshold.

Abstract: A method, apparatus, system, and computer program code for intelligent data discovery with dynamic ontology are provided. According to one illustrative embodiment, the method using a number of processors to perform the steps of: identifying a set of data items in unstructured content using a dynamic data schema populated from a dynamic ontology; and responsive to identifying a data item that is not recognized in the data schema: storing the data item with labels; generating a weight for the data item; and responsive to the weight exceeding a threshold, updating the schema to include the data item that was not recognized.