Abstract: A machine learning engine may be trained using artificial intelligence techniques and used according to techniques discussed herein. While an initial electronic transaction for a resource may be permitted, a subsequent related transaction to the initial electronic transaction may be analyzed in view of additional electronic information that was not available at the time of the initial transaction. Analysis of the subsequent related transaction, using the machine learning engine, may indicate a new classification related to the resource and/or the acquisition of the resource. Based on this new classification, usage of the resource may be restricted and/or denied, and the initial transaction for the resource may even be canceled retroactively.

Abstract: A method according to the present disclose may include presenting, on a graphical user interface (GUI), an interactive element; receiving, via the interactive element on the GUI, a research target and a type of research; autonomously retrieving, from a search engine, search results related to the research target; identifying, using a predictive machine learning model, at least one relevant portion of the search results, the at least one relevant portion comprising information related to the research target and responsive to the type of research; generating a prompt based on the type of research and the at least one relevant portion of the search results; and receiving, from a generative machine learning model in response to receipt of the generated prompt, a report indicative of the research target.

Abstract: A method of categorizing text entries on a document can include determining, for each of a plurality of text bounding boxes in the document, respective text, respective coordinates, and respective input embeddings. The method may further include defining a graph of the plurality of bounding boxes, the graph comprising a plurality of connections among the plurality of bounding boxes, each connection comprising a first and second bounding box and zero or more respective intermediate bounding boxes. The method may further include determining a respective attention value for each connection according to a quantity of intermediate bounding boxes in the connection and, based on a the respective attention values and a transformer-based machine learning model applied to the respective input embeddings and respective coordinates, determining output embeddings for each bounding box and, based on the respective output embeddings, generating a bounding box label for each bounding box.

Abstract: Techniques are disclosed for maintaining a network graph by updating labels in the graph based on features generated using at least two different feature extraction procedures. A server system accesses a machine learning model trained using features generated using multiple feature extraction procedures. Using the model, the system determines labels for unlabeled nodes in a network graph whose nodes correspond to entities and whose edges correspond to electronic communications executed between the different entities, where the determining is performed based on output of the model for values of the features corresponding to the unlabeled nodes. Based on the determining, the system updates the graph by assigning the determined labels to the unlabeled nodes in the graph. The system performs, based on the assigned labels indicating that behavior of entities corresponding to the nodes are anomalous, preventative actions for entities corresponding to the nodes with assigned labels.

Abstract: Techniques are disclosed for generating multiple different types of new features from a network graph and training a machine learning model using the newly generated features. A server system generates, from electronic communications, a network graph that includes nodes and edges. The server captures snapshots of the network graph that include nodes and edges existing at different time intervals. The system generates, for nodes included in respective snapshots, different types of features that include a neighbor convolution feature for a given node in a given snapshot by compressing node behavior features for neighbor nodes of the given node that are one hop away from the given node within the given snapshot of the network graph. The system trains, using the plurality of different types of features, a machine learning model usable to predict whether unlabeled nodes in the network graph are anomalous.

Abstract: A machine learning engine may be trained using artificial intelligence techniques and used according to techniques discussed herein. While an initial electronic transaction for a resource may be permitted, a subsequent related transaction to the initial electronic transaction may be analyzed in view of additional electronic information that was not available at the time of the initial transaction. Analysis of the subsequent related transaction, using the machine learning engine, may indicate a new classification related to the resource and/or the acquisition of the resource. Based on this new classification, usage of the resource may be restricted and/or denied, and the initial transaction for the resource may even be canceled retroactively.

Abstract: A system can comprise a processor that can facilitate performance of operations, comprising accessing a document comprising a plurality of text bounding boxes, wherein each respective text bounding box of the plurality of text bounding boxes comprises respective text, for each respective text bounding box, determining respective text bounding box coordinates and respective text bounding box input embeddings, based on the respective text bounding box coordinates, determining respective text bounding box positional encodings for each respective text bounding box, based on a transformer-based deep learning model applied to the respective text bounding box input embeddings, respective text bounding box coordinates, respective text bounding box positional encodings, and bias information representative of a modification to an attention weight of the transformer-based deep learning model, determining respective output embeddings for each respective text bounding box, and based on the respective output embeddings, ge

Abstract: Methods and systems are presented for providing a framework for analyzing graphs that exhibit non-homophilous behavior. Under the framework, a structural analysis and a feature-based analysis will be performed on a sequence of graphs. When performing the feature-based analysis, various features are extracted from each node in the sequence of graphs, and clusters of nodes are identified from each graph based on the features. A set of evolving prototypes is generated to represent evolving characteristics of the clusters of nodes, and a set of persistent prototypes is generated to represent persistent characteristics of the clusters of nodes. Information derived from the structural analysis of the graphs, the set of evolving prototypes, and the set of persistent prototypes are embedded within the nodes of the graphs. The embedded information is then used to classify the nodes.

Abstract: Techniques are disclosed relating to text sanitization. Given textual data, a computer system identifies tokens predicted to constitute sensitive information. Multi-field data structures (e.g., triplets) are generated for the identified tokens that include questions, answers, and corresponding context. These data structures are supplied to a pre-trained multiple-choice question (MCQ) reading comprehension model. The model outputs, for each data structure, a probability that the question and answer for a given data structure, provided the context, is accurate. A post-processing module can then rank probabilities in this set of probabilities and select the multi-field data structure with the highest probability (in some cases, a programmable threshold must also be met). The selected multi-field data structure is then used to select category information to be used in sanitizing the textual data.

Abstract: A method of categorizing text entries on a document can include determining, for each of a plurality of text bounding boxes in the document, respective text, respective coordinates, and respective input embeddings. The method may further include defining a graph of the plurality of bounding boxes, the graph comprising a plurality of connections among the plurality of bounding boxes, each connection comprising a first and second bounding box and zero or more respective intermediate bounding boxes. The method may further include determining a respective attention value for each connection according to a quantity of intermediate bounding boxes in the connection and, based on a the respective attention values and a transformer-based machine learning model applied to the respective input embeddings and respective coordinates, determining output embeddings for each bounding box and, based on the respective output embeddings, generating a bounding box label for each bounding box.

Abstract: In an embodiment, a first machine learning (ML) model is trained using a first portion of a training data set and a second ML model is trained using a second portion of the training data set. A prediction on data samples in the second portion by the first ML model is used to correct labels on noisy data samples in the second portion. A prediction on data samples in the first portion by the second ML model is used to correct labels on noisy data samples in the first portion. The first and second ML models are retrained after the labels of the noisy data samples have been replaced with corrective labels. After a number of iterations in retraining, the cross-label-correction may be performed again. After a certain number of cross-label-corrections, the training data in the first portion and the second portion is swapped to further train the models.

Abstract: A system can comprise a processor that can facilitate performance of operations, comprising accessing a document comprising a plurality of text bounding boxes, wherein each respective text bounding box of the plurality of text bounding boxes comprises respective text, for each respective text bounding box, determining respective text bounding box coordinates and respective text bounding box input embeddings, based on the respective text bounding box coordinates, determining respective text bounding box positional encodings for each respective text bounding box, based on a transformer-based deep learning model applied to the respective text bounding box input embeddings, respective text bounding box coordinates, respective text bounding box positional encodings, and bias information representative of a modification to an attention weight of the transformer-based deep learning model, determining respective output embeddings for each respective text bounding box, and based on the respective output embeddings, ge

Abstract: Systems and methods for data augmentation in a neural network system includes performing a first training process, using a first training dataset on a neural network system including an autoencoder including an encoder and a decoder to generate a trained autoencoder. A trained encoder is configured to receive a first plurality of input data in an N-dimensional data space and generate a first plurality of latent variables in an M-dimensional latent space, wherein M is an integer less than N. A sampling process is performed on the first plurality of latent variables to generate a first plurality of latent variable samples. A trained decoder is used to generate a second training dataset using the first plurality of latent variable samples. The second training dataset is used to train a first classifier including a first classifier neural network model to generate a trained classifier for providing transaction classification.

Abstract: A fraud detection model is used by a computer system to evaluate whether to grant a request to access a secure electronic resource. Before granting the request, the computer system evaluates the request using a multi-partite graph model generated using a plurality of previous requests. The multi-partite graph model includes at least a first set of nodes for sender accounts, a second set of nodes for recipient accounts, and a third set of nodes.

Abstract: Methods and systems for training a computer-based classification model for classifying data are presented. The computer-based classification model is configured to classify data into one of a plurality of classifications. An initial training data set for training the classification model is obtained. In some embodiments, the training data within the initial training data set is grouped into multiple clusters, and training data within one or more clusters having corresponding ratio between a first classification and a second classification below a threshold ratio is removed from the initial training data set to generate the modified training data set. The modified training data set, instead of the initial training data set, is used to train the classification model.

Abstract: A fraud detection model is used by a computer system to evaluate whether to grant a request to access a secure electronic resource. Before granting the request, the computer system evaluates the request using a multi-partite graph model generated using a plurality of previous requests. The multi-partite graph model includes at least a first set of nodes for sender accounts, a second set of nodes for recipient accounts, and a third set of nodes.

Abstract: A machine learning engine may be trained using artificial intelligence techniques and used according to techniques discussed herein. While an initial electronic transaction for a resource may be permitted, a subsequent related transaction to the initial electronic transaction may be analyzed in view of additional electronic information that was not available at the time of the initial transaction. Analysis of the subsequent related transaction, using the machine learning engine, may indicate a new classification related to the resource and/or the acquisition of the resource. Based on this new classification, usage of the resource may be restricted and/or denied, and the initial transaction for the resource may even be canceled retroactively.

Abstract: A fraud detection model is used by a computer system to evaluate whether to grant a request to access a secure electronic resource. The fraud detection model is generated by the computer using a plurality of previously received requests. Each request is associated with a sender account of the computer system and a recipient account to which the computer system previously sent a message containing a link to the secure electronic resource.

Abstract: A machine learning engine for fraud detection following link selection may be trained using artificial intelligence techniques and used according to techniques discussed herein. A buyer account may be used to establish and generate a digital gift card having a particular value specified by the buyer. The digital gift card may then be conveyed to another account, such as an email address. The digital gift card may be provided with an online electronic process for redemption and use of the value, for example, by selecting a link and navigating to the process. When the claimer account attempts to utilize the value of the gift card by navigating to the process or otherwise engaging in the electronic process through a device, a risk and fraud analysis engine may execute to determine, based on real-time data of the claimer account, the buyer account, and/or device, whether the digital gift card was generated fraudulently or is being used fraudulently.

Abstract: A fraud detection model is used by a computer system to evaluate whether to grant a request to access a secure electronic resource. The fraud detection model is generated by the computer using a plurality of previously received requests. Each request is associated with a sender account of the computer system and a recipient account to which the computer system previously sent a message containing a link to the secure electronic resource.