Abstract: A method predicts custom fields from text. Transaction text is normalized from transaction data to generate normalized text. A field prediction and a type prediction are selected using prediction data and the normalized text. The prediction data is generated using a machine learning model trained to identify field predictions from free form text. The field prediction and the type prediction are presented to a client device. In response to user input from the client device, the transaction data is updated with the field prediction.

Abstract: A method including training predictor machine learning models (MLMs) using a first data set. The trained predictor MLMs are trained to predict classifications of data items in the first data set. The method also includes training confidence MLMs using second classifications, output by the trained predictor MLMs. The method also includes generating an aggregated ranked list of classes based on third classifications output by the trained predictor MLMs and second confidences output by the trained confidence MLMs. The method also includes training an ensemble confidence MLM using the aggregated ranked list of classes to generate a trained ensemble confidence MLM. The trained ensemble confidence MLM is trained to predict a corresponding selected classification for each corresponding data item in a training data set containing second data items similar to the first data items.

Abstract: Forecasts are provided based on dynamic model selection for different sets of time series. A model comprises a transformation and a prediction algorithm. Given a time series, a transformation is selected for the time series and a prediction algorithm is selected to make a forecast based on the transformed time series. Sets of time series are distinguished from each other based on diverse sparsities, temporal scales and other time series attributes. A model is dynamically selected based on time series attributes to increase forecasting accuracy and decrease forecasting computation time. The dynamic model selection is based on the creation of a meta-model from historical sets of historical time series.

Abstract: A method and system learn functions to be associated with data fields of forms to be incorporated into an electronic document preparation system. The functions are essentially sets of operations required to calculate the data field. The method and system receive form data related to a data field that expects data values resulting from performing specific operations. The method and system utilize machine learning and training set data to generate, test, and evaluate candidate functions to determine acceptable functions.

Abstract: A method and system to learn new forms to be incorporated into an electronic document preparation system, or to learn the behavior of existing systems, receive form data related to a new form having a plurality of data fields that expect data values based on specific functions. The method and system gather training set data including previously filled forms having completed data fields corresponding to the data fields of the new form. The method and system include multiple analysis modules that each generate candidate functions for providing data values for the data fields of the new form. The method and system evaluate the candidate functions from each analysis technique and select the candidate functions that are most accurate based on comparisons with the training set data.

Abstract: Systems and methods for training machine learning models are disclosed. An example method includes receiving a plurality of first outputs and a ground truth value for each first output, each first output including an extracted string and a raw confidence score, determining, for each first output, an accuracy metric based at least in part on the extracted string and its corresponding ground truth value, for each extracted string: determining a similarity metric between the respective extracted string and each other extracted string of the plurality of first outputs, and determining a pseudo-accuracy based at least in part on the determined similarity metrics and the determined accuracy metrics, generating training data based at least in part on the determined pseudo-accuracies and the plurality of first outputs, and training the machine learning model, based on the training data, to predict pseudo-accuracies associated with subsequent outputs from a document extraction model.

Abstract: The disclosed embodiments provide a system that verifies user attributes. During operation, the system obtains financial data containing a set of strings representing a set of deposits into one or more financial accounts of one or more users. Next, the system generates one or more clusters of the deposits according to a similarity between pairs of strings in the set of strings. For each cluster from the one or more clusters, the system associates a subset of the strings in the cluster with an income source and extracts one or more employment attributes of a user from a subset of the deposits associated with the user in the cluster. Finally, the system uses the one or more employment attributes to provide an automatic verification of an income of the user.

Abstract: Systems and methods for forecasting future values of data streams are disclosed. One example method may include receiving information characterizing each of a plurality of forecasting models, retrieving historical data for each of a plurality of data streams, determining one or more constraints, dynamically selecting one of the plurality of forecasting models for each of the data streams based on accuracy metrics for the forecasting models, estimating cost metrics associated with each forecasting model, dynamically selecting the forecasting model based at least in part on the accuracy metrics, the cost metrics, and the determined constraints, and forecasting a first subsequent value of each data stream using the corresponding selected forecasting model.

Abstract: A computing system monitors interaction of a user with a dialog in an application executing on a user device associated with the user. The user is executing a workflow associated with the dialog. The computing system determines a time the user spent executing the workflow associated with the dialog based on the monitoring. The computing system determines whether the user successfully executed the workflow based on the monitoring. The computing system generates a time flow metric for the dialog based on the time the user spent interacting with the dialog. The computing system generates a flow scale metric based on whether the user successfully executed the workflow. The computing system automatically generates a user experience rating for the dialog based on the user metric values. The user experience rating is a weighted representation of the time flow metric and the flow scale metric.

Abstract: Systems and methods that may be used to provide an inference-based incident detection and reporting process. In one or more embodiments, an inference model may be used to track and analyze changes and incident data from various sources of record to quickly predict potential sources of errors and help focus in on the root cause of a software incident.

Abstract: Certain aspects of the present disclosure provide techniques for determining query intent and complexity based on text input. One example method generally includes receiving, from a user device, a text query and preprocessing the text query to generate a query vector. The method further includes providing the query vector to an intent model configured to output a user intent of the text query and providing the query vector to a complexity model configured to output a complexity of the text query. The method further includes receiving the user intent of the text query from the intent model and receiving the complexity of the text query from the complexity model. The method further includes determining, based on the user intent and the complexity of the query, a routing target for the text query and routing the text query to the routing target.

Abstract: Systems and methods for manipulating a graphical user interface (GUI) are described. The GUI is configured to show, to a user, a transition from displaying an inline card to displaying a corresponding snackbar. The system uses an overlay to seamlessly show the transition without requiring significant edits to most portions of the GUI. In various implementations, a system displays an inline card in a base layer of a GUI and receives an input associated with the inline card. In response to receiving the input, the system generates a transitional object in an overlay of the GUI (with the overlay displayed over the base layer), generates a snackbar in the base layer, adjusts the transitional object in the overlay from covering the inline card to covering where the snackbar is generated, and removes the transitional object from the GUI.

Abstract: Certain aspects of the present disclosure provide techniques for extracting data from a document. An example method generally includes identifying a bounding polygon of the region from an electronic image of the document and extracting data from within the bounding polygon of the region. The method further includes generating revised extracted data based on the extracted data, and combining the revised extracted data with other data extracted from the electronic image of the document to generate input data for a data processing application.

Abstract: Systems and methods are provided to perform context-aware text extraction.

Abstract: Systems and methods for real-time churn prediction using a flexible and robust machine learning system to predict user churn propensities in real-time using user activity data. In addition to churn propensity prediction, systems and methods may provide insights based on user activity data intelligence.

Abstract: Systems and methods may be used to generate and use a structured form representation and structured metadata. The structured form representation and structured metadata may include information relevant to a particular context and may be used to update document templates, import new documents and update document versions into software, automate data entry for document completion, update records to include new and or updated information, and provide other functionality of an information service.

Abstract: A computing system receives, from a client device, an image of a content item uploaded by a user of the client devices. The computing system divides the image into one or more overlapping patches. The computing system identifies, via a first machine learning model, one or more distortions present in the image based on the image and the one or more overlapping patches. The computing system determines that the image meets a threshold level of quality. Responsive to the determining, the computing system corrects, via a second machine learning model, the one or more distortions present in the image based on the image and the one or more overlapping patches. Each patch of the one or more overlapping patches are corrected. The computing system reconstructs the image of the content item based on the one or more corrected overlapping patches.

Abstract: A method including generating a first overlap matrix from a first attribute having first measurements of data items, and generating a second overlap matrix from a second attribute having second measurements of the data items. Samples of weights are generated, each of the samples of weights including a corresponding first weight for the first overlap matrix and a corresponding second weight for the second overlap matrix. For each of the samples of weights, the first overlap matrix is combined with the corresponding first weight and the second overlap matrix is combined with the corresponding second weight. Similarity matrices are generated by combining, for each of the samples of weights, a weighted first overlap matrix with a weighted second overlap matrix. A cluster analysis is performed on the similarity matrices to generate groupings the similarity matrices. The groupings represent groups of the data items. A selected grouping is chosen.

Abstract: In general, in one aspect, one or more embodiments relate to a method including receiving, in a business rules engine, input data from disparate data sources. The input data describes a merchant and an application by the merchant to use an electronic payments system for processing transactions between the merchant and customers. Featurization is performed on the input data to form a machine readable vector. By applying the machine readable vector as input to a machine learning model in a machine learning layer, a risk score is predicted. The machine learning model is trained using training data describing use of the electronic payments system by other merchants. The risk score is an estimated probability of the merchant being unable to satisfy an obligation of using the electronic payments system. A business rules engine, based on the risk score, limits use of the electronic payments system by the merchant.