Nec Corporation

Nec Laboratories America, Inc.

Christopher Malon

Eric Cosatto

Hans Peter Graf

Matthew Lynn Miller

Renqiang Min

Iain Melvin

Hans-Peter Graf

Pierre-Francois Laquerre

Asim Kadav

Atsushi Marugame

Bing Bai

Nitish Joshi

Juho Kim

Pengyu Cheng

A method trains an inference model on two-hop NLI problems that include a first and second premise and a hypothesis, and further includes generating, by the model using hypothesis reduction, an explanation from an input premise and an input hypothesis, for an input single hop NLI problem. The learning step determines a distribution over extraction starting positions and lengths from within the first premise and hypothesis of a two-hop NLI problem. The learning step k extraction output slots with combinations of words from the first premise of the two-hop NLI problem and fills another extraction output slots with combinations of words from the hypothesis of the two-hop NLI problem. The learning step trains a sequence model by using the extraction output slots and the other extraction output slots together with the second premise as an input to a single-hop NLI classifier to output a label of the two-hop NLI problem.

Extracting explanations from supporting evidence

Methods and systems for language processing include augmenting an original training dataset to produce an augmented dataset that includes a first example that includes a first scrambled replacement for a first word and a definition of the first word, and a second example that includes a second scrambled replacement for the first word and a definition of an alternative to the first word. A neural network classifier is trained using the augmented dataset.

Enhanced word embedding

A method for neural network training is provided. The method inputs a training set of textual claims, lists of evidence including gold evidence chains, and claim labels labelling the evidence with respect to the textual claims. The claim labels include refutes, supports, and not enough information (NEI). The method computes an initial set of document retrievals for each of the textual claims. The method also includes computing an initial set of page element retrievals including sentence retrievals from the initial set of document retrievals for each of the textual claims. The method creates, from the training set of textual claims, a Leave Out Training Set which includes input texts and target texts relating to the labels. The method trains a sequence-to-sequence neural network to generate new target texts from new input texts using the Leave Out Training Set.

Multi-hop evidence pursuit

A computer-implemented method is provided for disentangled data generation. The method includes accessing, by a bidirectional Long Short-Term Memory (LSTM) with a multi-head attention mechanism, a dataset including a plurality of pairs each formed from a given one of a plurality of input text structures and given one of a plurality of style labels for the plurality of input text structures. The method further includes training the bidirectional LSTM as an encoder to disentangle a sequential text input into disentangled representations comprising a content embedding and a style embedding based on a subset of the dataset. The method also includes training a unidirectional LSTM as a decoder to generate a next text structure prediction for the sequential text input based on previously generated text structure information and a current word, from a disentangled representation with the content embedding and the style embedding.

Controlled text generation with supervised representation disentanglement and mutual information minimization

Methods and systems for disentangled data generation include accessing a dataset including pairs, each formed from a given input text structure and a given style label for the input text structures. An encoder is trained to disentangle a sequential text input into disentangled representations, including a content embedding and a style embedding, based on a subset of the dataset, using an objective function that includes a regularization term that minimizes mutual information between the content embedding and the style embedding. A generator is trained to generate a text output that includes content from the style embedding, expressed in a style other than that represented by the style embedding of the text input.

Contextual text generation for question answering and text summarization with supervised representation disentanglement and mutual information minimization

A computer-implemented method and system are provided for teaching syntax for training a neural network based natural language inference model. The method includes selectively performing, by the hardware processor, person reversal on a set of hypothesis sentences, based on person reversal prevention criteria, to obtain a first training data set. The method further includes enhancing, by the hardware processor, a robustness of the neural network based natural language inference model to syntax changes by training the neural network based natural language inference model on original training data combined with the first data set.

Teaching syntax by adversarial distraction

A system verifies textual claims using a document corpus. The system includes a memory for storing program code and a processor device for running the code to retrieve documents from the corpus based on Term Frequency Inverse Document Frequency (TFIDF) similarity to a set of textual claims. The processor extracts named entities and capitalized phrases from the textual claims. The processor retrieves documents from the corpus with titles matching any of the extracted named entities and capitalized phrases. The processor extracts premise sentences from the retrieved documents. The processor classifies the premise sentences together with sources of the premises sentences against the textual claims to obtain classifications from among possible classifications including a supported, an unverified, or a contradicted classification. The processor aggregates the classifications over the premise sentences to selectively output, for each textual claim, an overall decision of the supported classification, the unverified classification, or the contradicted classification.

Verifying textual claims with a document corpus

Semantic indexing methods and systems are disclosed. One such method is directed to training a semantic indexing model by employing an expanded query. The query can be expanded by merging the query with documents that are relevant to the query for purposes of compensating for a lack of training data. In accordance with another exemplary aspect, time difference features can be incorporated into a semantic indexing model to account for changes in query distributions over time.

Query generation and time difference features for supervised semantic indexing

Systems and methods are disclosed for classifying histological tissues or specimens with two phases. In a first phase, the method includes providing off-line training using a processor during which one or more classifiers are trained based on examples, including: finding a split of features into sets of increasing computational cost, assigning a computational cost to each set; training for each set of features a classifier using training examples; training for each classifier, a utility function that scores a usefulness of extracting the next feature set for a given tissue unit using the training examples. In a second phase, the method includes applying the classifiers to an unknown tissue sample with extracting the first set of features for all tissue units; deciding for which tissue unit to extract the next set of features by finding the tissue unit for which a score: S=U&#x2212;h*C is maximized, where U is a utility function, C is a cost of acquiring the feature and h is a weighting parameter; iterating until a stopping criterion is met or no more feature can be computed; and issuing a tissue-level decision based on a current state.

Computationally efficient whole tissue classifier for histology slides

Disclosed is a computer implemented method for fully automated tissue diagnosis that trains a region of interest (ROI) classifier in a supervised manner, wherein labels are given only at a tissue level, the training using a multiple-instance learning variant of backpropagation, and trains a tissue classifier that uses the output of the ROI classifier. For a given tissue, the method finds ROIs, extracts feature vectors in each ROI, applies the ROI classifier to each feature vector thereby obtaining a set of probabilities, provides the probabilities to the tissue classifier and outputs a final diagnosis for the whole tissue.

Whole tissue classifier for histology biopsy slides

Methods and systems for interactive image analysis include receiving a selection of a region of an image and a request for analysis of the selection at an interface layer, transferring the selection and the request to an interpretation layer for analysis, dividing the selected region of the image into a plurality of sub-sections optimized for parallel computation to provide an analysis result that minimizes perceptible delay between receiving the request and receipt of results, analyzing the sub-sections using one or more execution nodes using a copy of the image stored in a shared memory, and providing combined analysis results to the interface layer for display.

Interactive analytics of digital histology slides

A method and systems for cloud-based digital pathology include scanning received slides that include a pathology sample to produce a sample image in a shared memory, analyzing the sample image using one or more execution nodes, each including one or more processors, according to one or more analysis types to produce intermediate results, transmitting some or all of the sample image to a client device, further analyzing the sample image responsive to a request from the client device to produce a final analysis based on the intermediate results, and transmitting the final analysis to the client device.

Cloud-based digital pathology

Methods and systems for digital pathology with low-latency analytics include determining potential regions of interest within an image in accordance with one or more high-priority analyses, dividing the potential regions of interest into a plurality of sub-sections optimized for parallel computation, analyzing the sub-sections using one or more execution nodes, each including one or more processors, using a copy of the image stored in a shared memory according to the one or more high-priority analyses, and storing an intermediate analysis result based on analysis results from the one or more execution nodes in a shared memory.

Digital pathology system with low-latency analytics

A method for training a classifier to be operative as an epithelial texture classifier, includes obtaining a plurality of training micrograph areas of biopsy tissue and for each of the training micrograph areas, identifying probable locations of nuclei that form epithelia, generating a skeleton graph from the probable locations of the nuclei that form the epithelia, manually drawing walls on the skeleton graph outside of the epithelia to divide the epithelia from one another, and manually selecting points that lie entirely inside the epithelia to generate open and/or closed geodesic paths in the skeleton graph between pairs of the selected points. Data is obtained from points selected from the walls and the paths and applied to a classifier to train the classifier as the epithelial texture classifier. A method and detector for detecting epithelial structures includes applying a sample micrograph area of biopsy tissue to an epithelial texture classifier; identifying probable locations of nuclei that form epithelia of the sample micrograph area with the epithelial texture classifier, generating a skeleton graph from the probable locations of the nuclei that form the epithelia of the sample micrograph area, determining a set of open and/or closed geodesic paths in the skeleton graph of the sample micrograph area; and determining a set of the epithelial masks using the open and/or closed epithelial paths of the sample micrograph area.

Epithelial structure detector and related methods

A detector and method for automatically detecting signet ring cells in an image of a biopsy tissue sample, includes finding in the image, points about which cell membranes appear in radial symmetry; selecting as candidate points, at least ones of the points that have an adjacent nuclei with a predetermined shape feature; and applying a convolutional neural network to the candidate points to determine which of the candidate points are signet ring cells.

Growing community of inventors

Fort Lee, NJ, United States of America

Christopher Malon