Abstract: Computer-implemented methods, system and computer program products for identifying and communicating errors or inconsistencies in data of an electronic tax return during preparation of the electronic tax return, e.g., presenting a message to a user regarding the error or inconsistency “on the spot” or immediately in response to identifying the error or inconsistency during a currently displayed interview screen. Error messages may be communicated to an interface controller that processes error messages or non-binding suggestions concerning same from a rule module by the rule module pushing messages to the interface controller, the interface controller pulling or querying the rule module, or by use of a shared memory or data store to which the rule module pushes messages and from which the interface controller pulls or retrieves the messages generated by the rule module.

Abstract: Computer-implemented methods, articles of manufacture and computerized systems for identifying or alerting a user of certain data in electronic tax returns. A computerized tax return preparation system including a tax return preparation software application executed by a computing device receives first and second tax data and populates respective fields of the electronic tax return. The system executes a predictive model such as logistic regression, naive bayes, K-means clustering, clustering, k-nearest neighbor, and neural networks. First tax data is an input into the predictive model, which generates an output, which is compared with second tax data. An alert is generated when the second tax data does not satisfy pre-determined criteria relative to the first output generated by the predictive model. The same or other predictive model may be used as additional tax data is received for subsequent tax data analysis.

Abstract: Certain aspects of the present disclosure provide techniques for providing an automated callback service to a user. An example technique includes receiving an indication of a product installation failure, which includes an error code, a context of the computing device, and a product identifier. Based on the product identifier, a phone number is retrieved that is associated with the user of the computing device. A set of solutions predicted to resolve the installation failure is retrieved, based on the error code and the context. A callback is established to the user, and the user is connected with a virtual agent that will provide solutions from the set of solutions in the ranking order retrieved until a solution is determined to resolve the product installation error. The ranking of the predicted solutions is updated for other users in the future who may face a similar product installation error.

Abstract: Systems and methods here may include utilizing a computer with a processor and a memory for receiving a pixelated image of an original size, converting the pixelated image to grayscale, calculating a magnitude of spatial gradients in the received pixelated grayscale image, downscaling the received pixelated grayscale image, computing a multiplicative gain correction for the downscaled received pixelated grayscale image, re-enlarging a gain multiplication for the original image, and applying the gain multiplication to the image to generate a processed image with higher contrast than the received pixelated image.

Abstract: Systems, methods, and articles of manufacture for associated tax data in a tax data item with a tax entity to which the tax data item pertains. A computerized tax return system is configured to access a tax data item having tax data regarding a tax entity. The system stores tax entity record(s) for one or more tax entities related to a tax return. The system has a set of matching rules for matching tax data from the tax data item with key attributes of the tax entity record(s) for a respective tax topic, and a matching algorithm for utilizing the matching rules to compare utilize the matching rules to compute a matching score. The tax return system determines a matching tax entity using the matching score. The system associates the matching tax entity with the tax data and generates a normalized data set for the tax data item.

Abstract: A method and system trains, with a machine learning process, an analysis model to detect anomalous behavior of tax professionals affiliated with a tax return preparation system. The analysis model is trained with a training set that includes contextual and behavioral data for a plurality of historical tax professionals. The trained analysis model then analyzes and generates risk scores for current tax professionals based on current behavioral and contextual data associated with the current tax professionals.

Abstract: Arbitrary image data may be transformed into data suitable for optical character recognition (OCR) processing. A processor may generate a plurality of intermediate feature layers of an image using convolutional neural network (CNN) processing. For each intermediate feature layer, the processor may generate at least one text proposal using a region proposal network (RPN). The at least one text proposal may comprise a portion of the intermediate feature layer that is predicted to contain text. The processor may merge the text proposals with one another to form a patch of the image that is predicted to contain text. The processor may determine outer coordinates of the patch. The outer coordinates may comprise at least leftmost, rightmost, topmost, and bottommost coordinates. The processor may generate a quadrilateral of the image that is a smallest quadrilateral including the leftmost, rightmost, topmost, and bottommost coordinates.

Abstract: A computer-implemented method is provided to perform text classification with a neural network system. The method includes providing a computing device to receive input datasets including user input question text and feed the datasets to the neural network system. The neural network system includes one or more neural networks configured to extract and concatenate character-based features, word-based features from the question datasets and clickstream embeddings of clickstream data to form a representation vector indicative of the question text and user behavior. A representation vector is fed into fully connected layers of a feed-forward network. The feed-forward network is configured to predict a first class and a second class associated with respective user input questions based on the representation vector.

Abstract: A method and system determines whether or not a new data acquisition process is working for individual financial accounts of users of a data management system. The method and system trains an analysis model with a machine learning process. The trained analysis model then analyzes financial data obtained by both an old data acquisition process and the new data acquisition process. The trained analysis model identifies whether the new data acquisition process is working properly based on the analysis.

Abstract: A method may include acquiring, from a document, document elements and attributes describing the document elements. One or more of the attributes may be geometric attributes describing a placement of the corresponding document element within the document. The method may further include deriving features for the document elements using the attributes, detecting form components using the features, clustering the form components into line objects of a structured representation by applying an unsupervised machine learning model to the geometric attributes of the document elements, and populating a compliance form using the structured representation.

Abstract: Known fraudulent invoice data, including defined and known fraudulent invoice feature data, is used to train a machine learning-based fraudulent invoice detection model to generate a fraudulent invoice score for invoices indicating a determined probability that a given invoice is fraudulent. The machine learning-based fraudulent invoice detection model is then used to generate a fraudulent invoice score for subsequent invoices before those invoices are paid by, and in some cases before the invoices are provided to, the parties being asked to pay the invoices. The fraudulent invoice score for the subsequent invoice is then used to determine if the subsequent invoice should be passed on to the parties being asked to pay the invoices for payment, or if one or more protective actions should be taken.

Abstract: A processor may generate a plurality of intermediate feature layers of an image using convolutional neural network (CNN) processing. For each intermediate feature layer, the processor may generate a plurality of text proposals using a region proposal network (RPN). Each text proposal may comprise a portion of the intermediate feature layer that is predicted to contain text. The processor may perform OCR processing on image data within a plurality of regions of the image to generate a text result for each region. Each region may comprise at least one of the text proposals. The processor may assemble the text results into a text string comprising the text results ordered according to a spatial order in which the plurality of regions appear within the image.

Abstract: Customer transaction data is processed to determine transaction locations for transactions, including transactions whose locations are not initially known. The transaction location data is then utilized to identify merchants that are mobile merchants, and the mobile merchant locations are periodically recalculated and tracked. Customer transaction data is further utilized to identify relationships between mobile merchants and customers of those mobile merchants. Merchant and customer data is also analyzed to identify potential customers of mobile merchants, and data related to the mobile merchants is provided to current and potential customers of those mobile merchants.

Abstract: A method, system, and computer readable medium for automatically selecting target consumers and delivering business-initiated messages to the selected target consumers, as well as automatically selecting target businesses and delivering consumer-initiated messages to the selected target businesses. In particular, financial management applications used by the businesses and consumers are employed in targeting and delivering the business-initiated messages and consumer-initiated messages.

Abstract: Aspects of the present disclosure provide techniques for displaying reduced data sets based on pre-classification of a larger data set. Embodiments include receiving a plurality of activity records describing a plurality of activities associated with the user. Embodiments further include grouping the plurality of activities into one or more pre-classified data sets based on the plurality of activity records. Embodiments further include providing the user with a summary of a pre-classified data set of the one or more pre-classified data sets via a user interface. Embodiments further include providing the user, via the user interface, with a user interface element that allows the user to categorize all activities in the pre-classified data set together based on the summary. Embodiments further include receiving input from the user via the user interface, the input assigning a category to all activities in the pre-classified data set together based on the summary.

Abstract: A method for loading objects from hash chains. A version chain of a class for a serialized object is located in an instance block of an instance chain. A class version of the serialized object is compared to a current version of the class. When the class version of the serialized object matches the current version of the class, a runtime object is loaded by deserializing the serialized object. When the class version of the serialized object does not match the current version of the class: one or more field values are extracted from the serialized object; a converter function is applied to the one or more field values to generate one or more converted field values; and a runtime object that matches the current version is loaded with the one or more converted field values.

Abstract: A processor may identify a combination term including at least two individual terms within at least one source of truth stored in a memory in communication with the processor. The processor may identify at least one document including the at least two of the individual search terms. The processor may determine a document weight for the at least one document based on the combination search term and the at least two of the individual search terms within the combination search term. The processor may provide the document as a search result arranged according to the document weight and/or may index the document according to the document weight.

Abstract: Test deposit mechanisms used by financial institutions to link accounts are used to identify undisclosed accounts associated with users of a data management system. The potential existence of undisclosed accounts is determined based on the assumption that the presence of test deposit transactions in user account data is a strong indication that an undisclosed user account exists. Using this assumption, transaction data from user accounts disclosed to a user data management system is scanned to identify test deposit transactions listed in the transaction data. If test deposit transactions are identified, the user of the data management system is queried regarding the existence of the undisclosed user account. If the user confirms the existence of the undisclosed account, the formally undisclosed account is added to a set of disclosed user accounts with the data management system.

Abstract: A transceiver of a mobile device may receive a beacon transmitted by a mesh network member mobile device. A processor of the mobile device may extract a crew ID from the beacon. The processor may determine that the crew ID matches a crew ID of a user logged into the mobile device. In response to determining that the crew ID matches the crew ID of the user, the processor may generate a reply beacon. The transceiver may transmit the reply beacon. The transceiver may receive a response to the reply beacon. The response may include a job ID enabling the user to clock into a job. The processor may clock the user into the job. As a result of clocking the user into the job, the mobile device may become a member of the mesh network.

Abstract: Machine learning-based anomaly detection methods are used to identify a change in a user's streaming transaction data. If a threshold level of change in the user's transaction data is detected, the user is then identified as potentially having experienced a life event. Then, after a user is identified has having potentially experienced a life event, individual user transactions are processed and analyzed to determine the specific life event the user has most likely experienced. The user is then identified as having experienced the identified specific life event. This information is then used to customize the interactions between the user and the data management system such as questions asked of the user, forms or displays provided to the user, or offers made to the user.