International Business Machines Corporation

Mohamed Nooman Ahmed

Aaron K Baughman

Mauro Marzorati

John F Behnken

Andeep S Toor

Nicholas Albert McCrory

Michelle Welcks

Samuel H Woo

Charles Evan Beller

Kristen Maria Summers

Palani Sakthi

William G Dubyak

Daniel James Anderson

Anna Podgornyak

Fayin Li

Emily Fontaine

Michelle H Jung

Krista Kinnard

A system and method for classifying compartments at a security checkpoint includes classifying a compartment into a first category or a second category using a first stage neural network that analyzes a three-dimensional representation of the compartment extracted from an imaging device coupled to the computing system, and in response to classifying the compartment into the second category, screening the compartment using a second stage neural network that performs image segmentation to isolate a hazardous object present in the compartment.

Two-stage screening technique for prohibited objects at security checkpoints using image segmentation

A system and method for automatically detecting prohibited materials in a compartment at a security checkpoint includes receiving a three-dimensional representation of a compartment from an imaging device connected to the computing system, and classifying each voxel of the three-dimensional representation using a trained neural network to determine whether any voxel classifications of the three-dimensional representation correspond to a voxel classification of a prohibited material.

Classifying a material inside a compartment at security checkpoints

A system and method for automatically detecting prohibited objects in a compartment at a security checkpoint includes receiving a three-dimensional representation of a compartment from an imaging device connected to the computing system, identifying, by the processor, a region within the compartment that has a highest likelihood of containing a prohibited object, based on a voxel classification of the three-dimensional representation using a first trained neural network, and classifying, by the processor, shapes of objects located within the identified region using a second trained neural network to determine whether any of the shapes correspond to a shape of a prohibited object known by the second neural network.

Classifying compartments at security checkpoints by detecting a shape of an object

A system and method for automatically classifying compartments at a security checkpoint includes receiving a three-dimensional representation of a compartment from an imaging device connected to the computing system, and classifying the compartment into a first category or a second category using a trained neural network that analyzes the three-dimensional representation, wherein compartments classified into the first category are automatically cleared at the security checkpoint and compartments classified into the second category undergo further security checks.

Screening technique for prohibited objects at security checkpoints

A method, computer program product, and a system where a processor(s) obtains an original image. The processor(s) applies a number of filters to the original image to generate a group of filtered images. The processor(s) stacks the original image with the filtered images in a three dimensional array; each layer of the stack comprises a separate filtered image or the original image and the three dimensional array comprises an augmented version of the original image. The processor(s) facilitates classification of the original image by a deep convolution neural network, where the facilitating comprises providing the augmented version of the original image to the deep convolution neural network, and where the deep convolution neural network classifies the original image based on applying a classification model to the augmented version of the original image The processor(s) receives the classification of the original image from the deep convolution neural network.

Convolutional neural network with augmentation features

An embodiment of the invention may include a method, computer program product and system for generating follow-up questions based on machine learning utilizing a computing device. The embodiment may include receiving an input question from a user. The embodiment may include parsing the received input question to extract input question components. Parsing utilizes natural language processing techniques. The embodiment may include executing trained question component models to predict follow-up question components. The extracted input question components are utilized as inputs to the trained question component models. The embodiment may include combining the predicted follow-up question components to generate one or more follow-up questions. The embodiment may include returning the one or more follow-up questions to the user.

Suggesting follow up questions from user behavior

Natural language processing is provided. A computer processor, selects a pipeline based on an artifact that includes unstructured data, the pipeline identifying a first algorithm of a first set of algorithms of a first human language technology (HLT) component and a second algorithm of a second set of algorithms of a second HLT component; applies the first algorithm based on the artifact to generate a first cluster space associated with the artifact; amends an evidence chain associated with the artifact in response to applying the first algorithm, wherein the evidence chain includes one or more probabilistic findings of truth corresponding to the artifact; standardizes a first ontology of the first cluster space; applies the second algorithm based on the artifact to generate a second cluster space that is associated with the artifact; and identifies a set of information of one or more corpora that is relevant to the artifact.

Natural language processing with dynamic pipelines

Detection and classification of personally identifiable information includes identifying a document with a known author. A first set of features of the document is extracted using natural language processing, and a second set of features of the document is extracted based upon one or more past documents for the known author using a recurrent neural network. The first set of features and the second set of features are classified using a classifier to produce classified extracted features. Personally identifiable information is labeled in the document based upon the classified extracted features.

Machine-learning based detection and classification of personally identifiable information

A set and a second set of collections of forecasted feature vectors are selected from a repository for a future time window, a cyber-attack being in progress in a data processing environment at the present time, a collection in the set and a collection in the second set indicating an event related to the cyber-attack in a first region and a second event in a second region, respectively, of the environment at a discrete time. The events corresponding to the collections are classified, using an LTSM network, into a class of cyber-attack. From a mapping between a set of phases of the cyber-attack and a set of classes, a phase that corresponds to the class is predicted as likely to occur during the future time window in the region.

Detecting and predicting cyber-attack phases in adjacent data processing environment regions

A first collection including an analytical feature vector and a Q&A feature vector is constructed. A second collection is constructed from the first collection by inserting noise in at least one of the vectors. A third collection is constructed by crossing over at least one of vectors of the second collection with a corresponding vector of a fourth collection, migrating at least one of the vectors of the second collection with a corresponding vector of a fifth collection. Using a forecasting configuration, an analytical feature vector of the third collection is aged to generate a changed analytical feature vector containing analytical feature values expected at a future time. The changed analytical feature vector is input into a trained neural network to predict a probability of the cyber-attack occurring at the future time.

Forecasting and classifying cyber-attacks using analytical data based neural embeddings

In an approach to face recognition in an image, one or more computer processors receive an image that includes at least one face and one or more face parts. The one or more computer processors detect the one or more face parts in the image with a face component model. The one or more computer processors cluster the detected one or more face parts with one or more stored images. The one or more computer processors extract, from the clustered images, one or more face descriptors. The one or more computer processors determine a recognition score of the at least one face, based, at least in part, on the extracted one or more face descriptors.

Face detection, representation, and recognition

A method includes one or more program obtaining a natural language query, where the natural language query is comprised of a first group of terms, converting the natural language query to a machine language query, and executing machine language query on at least one computer resource. The program obtains search results responsive to the machine language query, where the search results include related terms derived from terms in the first group of terms utilizing concept expansion. The program parses the search results by applying a statistical information extraction to the terms in the first group and to the related terms to identify entities and generates at least one additional natural language query by incorporating a portion of the identified entities into the query. The identified entities in the new query are a second group of terms. At least one term in the first group is not in the second group.

Dynamic question formulation to query data sources

A first collection including a first feature vector and a Q&A feature vector is constructed. A second collection is constructed from the first collection by inserting noise in at least one of the vectors. A third collection is constructed by migrating, at least one of a vectors of the second collection with a corresponding vector of a fourth collection. The second and the fourth collections have a property distinct from one another. Using a forecasting configuration, a vector of the third collection is aged to generate a changed feature vector, the changed feature vector containing feature values expected at a future time. The changed feature vector is input into a trained neural network to predict a probability of the cyber-attack occurring at the future time.

Forecasting and classifying cyber attacks using neural embeddings migration

To compress a document, a number of edges present in a selected portion of the document are counted to determine whether the number of edges exceeds a threshold. When the number of edges exceeds the threshold, a pixel is selected from the portion and a set of neighboring pixels is identified for the pixel. For each neighboring pixel in a subset of the neighboring pixels, a corresponding label of the neighboring pixel is identified. A mask layer contains labels of pixels in the portion where a label of the selected pixel is biased using labels of neighboring pixels in the subset of the neighboring pixels. The selected pixel is designated to a foreground or a background layer of the document according to the label of the selected pixel. A compressed document is constructed corresponding to the document using the mask layer, the foreground layer, and the background layer.

Growing community of inventors

Leesburg, VA, United States of America

Mohamed Nooman Ahmed