Abstract: Systems and methods that may be used to generate and use a social graph generated by user financial transaction data (i.e., a financial transaction-based social graph). Connections and other data within the financial transaction-based social graph can be used for targeted product offerings, other offerings, and or advertisements via e.g., collaborative filtering and user segmentation and profiling.

Abstract: Certain aspects of the present disclosure provide techniques for improved retrieval of data from documents. Embodiments include receiving, from a user, a definition of a document region, wherein the definition comprises coordinates relative to a location on a document page. Embodiments include receiving, from the user, an identifier associated with the document region. Embodiments include receiving a document comprising one or more elements. The document may not support queries for the one or more elements. Embodiments include building a tree based on the document, the tree including one or more elements with element coordinates. Embodiments include retrieving an item of data associated with the identifier by determining that the element coordinates of an element in the tree are within the document region associated with the identifier and retrieving the element as the item of data. Embodiments include using the item of data to perform an action.

Abstract: Techniques are disclosed for managing workflows using block chains. An entire workflow may be stored as a finite state machine on a block chain. Each data block of the block chain may correspond to a state and related operation of a workflow. Each data block of the block chain may be used as an input to a subsequent state of the workflow. Thus, any input may be used to compute the workflow even if the input is not designed to be used by a workflow. Nested block chains, each corresponding to a different workflow or different part of a workflow may be used to increase performance of computing the workflows.

Abstract: A method for modeling risk of a derogatory financial event. The method includes generating a risk profile of a small business (SMB) using a risk model based on accounting data and other third party business management application (BMA) data of the SMB. In particular, the risk model is generated using a training data set that includes accounting data and other third party BMA data of multiple SMBs over particular time periods related to historic derogatory events (e.g., to be delinquent and/or to default on a loan) of these SMBs. Generated using the risk model, the risk profile of the SMB represents the likelihood of a derogatory financial event occurring to the SMB.

Abstract: A feature generator generates a feature data structure (FDS) by pre-processing a tax return of a user by extracting inherent features and generating derived features based on the tax return; and populating entries of the FDS based on the inherent and derived features. A prediction engine generates a prediction data structure (PDS) based on the FDS. The PDS includes a prediction that the user has a credit score above a first credit score threshold by applying the features to a machine learning model (MLM). The MLM is trained using first tax returns for first tax payers, each having a credit score above the first threshold, and second tax returns for second tax payers, each having a credit score below the first threshold. A task manager executes, based on the prediction, a task that requires the credit score of the user exceed a second credit score threshold less than the first threshold.

Abstract: The present disclosure relates to a trustworthy data exchange. Embodiments include receiving, from a device, a query, wherein the query comprises a question. Embodiments include identifying particular information related to the query. Embodiments include receiving credentials from a user for retrieving the particular information related to the query. Embodiments include retrieving, using the credentials, the particular information related to the query from one or more data repositories that are part of a distributed database comprising an immutable data store that maintains a verifiable history of changes to information stored in the distributed database. Embodiments include determining, based on the particular information related to the query, an answer to the query. Embodiments include providing the answer to the device.

Abstract: Certain aspects of the present disclosure provide techniques for generating geofences for an address using an iterative process. An example method generally includes receiving an address to generate a geofence around. A position associated with the received address is obtained from a map service. A set of perimeter points is generated for the received address. Each perimeter point is associated with a latitude and longitude position determined based on a traversal direction and an iteratively determined distance from the obtained position associated with the received address. A geofence is generated for the received address based on the set of perimeter points.

Abstract: Certain aspects of the present disclosure provide techniques for detecting errors in account numbers. One example method generally includes receiving, from a user device, an entered number associated with a user and determining, based on a first portion of the entered number, an entity associated with the entered number. The method further includes obtaining, from an account number database, a plurality of account numbers associated with the entity and generating, from the plurality of account numbers, an account number matrix. The method further includes attempting to solve a multiplication equation of the account number matrix, wherein a solution of the multiplication equation is a vector of constants, upon determining a solution to the multiplication equation, determining whether the entered vector is a valid number for the entity and upon determining the entered vector is a valid number for the entity, storing the entered number in the account number database.

Abstract: Aspects of the present disclosure provide techniques for intelligently presenting targeted offers to a user of a software application. Embodiments include receiving received request from a user of the software application to access a portion of the software application. Using a predictive model, a predictive score is generated for the user. The predictive model generally may be a model trained using user account data reduced into n-tuples of predictive attributes representative of a plurality of users, correlated with clickstream data associated with the plurality of users indicating whether a user interacted with a targeted offer. The predictive score generally represents a likelihood that the user will interact with the targeted offer based on data stored by the software application for the user. The targeted offer is presented to the user based, at least in part, on a determination that the predictive score for the user exceeds a threshold score.

Abstract: A method includes importing personal and financial information from an image of a third-party tax form associated with a taxpayer. The method further includes assigning, based at least in part on the personal and financial information, the taxpayer to one of a plurality of user classes. The user class of the taxpayer determines at least one subsequent interaction between the taxpayer and an online tax application. And the method includes using at least some of the personal and financial information to pre-populate one or more fields of an online tax return for the taxpayer. The online tax return is presented in a view in a graphical user interface displayed by the online tax application.

Abstract: A method and system adaptively improves potential customer conversion rates, revenue metrics, and/or other target metrics by providing effective user experience options to some users while concurrently testing user responses to other user experience options, among a variety of user experience options, according to one embodiment. The method and system selects the user experience options by applying user characteristics data to an analytics model, according to one embodiment. The method and system analyzes user responses to the user experience options to update the analytics model, and to dynamically adapt the personalization of the user experience options, at least partially based on feedback from users, according to one embodiment. The method and system determines bias weights from characteristics of the analytics model and uses the bias weights to compensate for data biases when updating or generating analytics models.

Abstract: Techniques are disclosed for performing optical character recognition (OCR) by assessing and improving quality of electronic documents to perform the OCR. For example a method for identifying information in an electronic document includes obtaining a reference image of the electronic document, distorting the reference image by adjusting different sets of one or more parameters associated with a quality of the reference image to generate a plurality of distorted images, analyzing each distorted image to detect the adjusted set of parameters and corresponding adjusted values, determining an accuracy of detection of the set of parameters and the adjusted values, and training a model based at least on the plurality of distorted images and the accuracy of the detection, wherein the trained model determines at least a first technique for adjusting a set of parameters in a second image to prepare the second image for optical character recognition.

Abstract: A method for mitigating cold starts in recommendations includes receiving a request that identifies a requested page and identifying a content vector of the requested page. The content vector is generated based on providing text of the requested page to a neural network text encoder. The method further includes selecting, based on a rank engine and the content vector, a link to a cold start page that does not satisfy a threshold level of interaction data. The rank engine ranks the selected link above a second link to a warm page that does satisfy the threshold level of the interaction data. The method further includes presenting the requested page with the selected link.

Abstract: A method may include receiving an event from an event source. The event may correspond to event data. The event source may be a container executing an image. The image may correspond to image metadata including attributes describing the image. The method may further include combining the event data with the image metadata to obtain enriched data, detecting, using the enriched data, a deviation from a policy, and in response to detecting the deviation from the policy, performing an action to enforce the policy.

Abstract: A method for generating a synthetic dataset involves generating discretized synthetic data based on driving a model of a cumulative distribution function (CDF) with random numbers. The CDF is based on a source dataset. The method further includes generating the synthetic dataset from the discretized synthetic data by selecting, for inclusion into the synthetic dataset, values from a multitude of entries of the source dataset, based on the discretized synthetic data, and providing the synthetic dataset to a downstream application that is configured to operate on the source dataset.

Abstract: The invention relates to a method. The method includes receiving a flawed input comprising a domain specific misspelling. The method further includes encoding, by an encoder machine learning model executing on a computer processor, the flawed input on a per character basis to create a context vector. The method further includes decoding, by a decoder machine learning model executing on the computer processor, the context vector on the per character basis to create a rephrased input lacking the domain specific misspelling. The method further includes presenting the rephrased input.

Abstract: At least one processor may obtain a document comprising text tokens. The at least one processor may determine, based on a pre-trained language model, word embeddings corresponding to the text tokens. The at least one processor may determine, based on the word embeddings, named entities corresponding to the text tokens; and one or more accuracy predictions corresponding to the named entities. The at least one processor may compare the one or more accuracy predictions with at least one threshold. The at least one processor may associate, based on the comparing, the named entities with one or more confidence levels. The at last one processor may deliver the named entities and the one or more confidence levels.

Abstract: Transactions include text fields, such as description fields. Transactions are extracted from financial institutions using web-scraping extraction. In the process of extracting transactions, errors can be introduced into text fields, such as the inclusion of a space within a word or the removal of a space between words. A statistical approach is applied to the text fields. When two alternative text fields are presented, the alternative that statistically includes more common tokens, such as unigrams and bigrams, is chosen as the correct alternative. The incorrect alternative is replaced by the correct alternative in the text field.

Abstract: A system for targeted communication includes a computer processor and a platform configured to perform transactions between first users and second users. A management server is separate from the platform. The management server includes a first management application used by the first users and configured to generate a first user profile for each of the first users and a second management application used by the second users. The second management application, logically separate from the first management application, is configured to generate a second user profile for each of the second users. The management server also includes a third application configured to permit the first application to communicate with the second application through the third application and to enable targeted communication, between the first users and the second users, based on the first user profile and the second user profile.

Abstract: A machine learning method. A source domain data structure and a target domain data structure are combined into a unified data structure. First data in the source domain data structure are latent with respect to second data in the target domain data structure. The unified data structure includes user vectors that combine the first data and the second data. The user vectors are transformed into a transformed data structure by applying a mapping function to the user vectors. The mapping function relates, using at least one parameter, first relationships in the source domain data structure to second relationships in the target domain data structure. The at least one parameter is based on a combination of affinity scores relating items with which the user interacted and did not interact. The transformed data structure is input into a machine learning model, from which is obtained a recommendation relating to the target domain.