Abstract: The technology disclosed provides a so-called “joint many-task neural network model” to solve a variety of increasingly complex natural language processing (NLP) tasks using growing depth of layers in a single end-to-end model. The model is successively trained by considering linguistic hierarchies, directly connecting word representations to all model layers, explicitly using predictions in lower tasks, and applying a so-called “successive regularization” technique to prevent catastrophic forgetting. Three examples of lower level model layers are part-of-speech (POS) tagging layer, chunking layer, and dependency parsing layer. Two examples of higher level model layers are semantic relatedness layer and textual entailment layer. The model achieves the state-of-the-art results on chunking, dependency parsing, semantic relatedness and textual entailment.

Abstract: Approaches for the translation of structured text include an embedding module for encoding and embedding source text in a first language, an encoder for encoding output of the embedding module, a decoder for iteratively decoding output of the encoder based on generated tokens in translated text from previous iterations, a beam module for constraining output of the decoder with respect to possible embedded tags to include in the translated text for a current iteration using a beam search, and a layer for selecting a token to be included in the translated text for the current iteration. The translated text is in a second language different from the first language. In some embodiments, the approach further includes scoring and pointer modules for selecting the token based on the output of the beam module or copied from the source text or reference text from a training pair best matching the source text.

Abstract: The technology disclosed provides a so-called “joint many-task neural network model” to solve a variety of increasingly complex natural language processing (NLP) tasks using growing depth of layers in a single end-to-end model. The model is successively trained by considering linguistic hierarchies, directly connecting word representations to all model layers, explicitly using predictions in lower tasks, and applying a so-called “successive regularization” technique to prevent catastrophic forgetting. Three examples of lower level model layers are part-of-speech (POS) tagging layer, chunking layer, and dependency parsing layer. Two examples of higher level model layers are semantic relatedness layer and textual entailment layer. The model achieves the state-of-the-art results on chunking, dependency parsing, semantic relatedness and textual entailment.

Abstract: An online system that allows users to interact with it using expressions in natural language form includes an intent inference module allowing it to infer an intent represented by a user expression. The intent inference module has a set of possible intents, along with a small set of example natural language expressions known to represent that intent. When a user interacts with the system using a natural language expression for which the intent is not already known, the intent inference module applies a natural language inference model to compute scores indicating whether the user expression textually entails the various example natural language expressions. Based on the scores, the intent inference module determines an intent that is most applicable for the expression. If an intent cannot be determined with sufficient confidence, the intent inference module may further attempt to determine whether the various example natural language expressions textually entail the user expression.

Abstract: Approaches for the translation of structured text include an embedding module for encoding and embedding source text in a first language, an encoder for encoding output of the embedding module, a decoder for iteratively decoding output of the encoder based on generated tokens in translated text from previous iterations, a beam module for constraining output of the decoder with respect to possible embedded tags to include in the translated text for a current iteration using a beam search, and a layer for selecting a token to be included in the translated text for the current iteration. The translated text is in a second language different from the first language. In some embodiments, the approach further includes scoring and pointer modules for selecting the token based on the output of the beam module or copied from the source text or reference text from a training pair best matching the source text.

Abstract: The technology disclosed provides a so-called “joint many-task neural network model” to solve a variety of increasingly complex natural language processing (NLP) tasks using growing depth of layers in a single end-to-end model. The model is successively trained by considering linguistic hierarchies, directly connecting word representations to all model layers, explicitly using predictions in lower tasks, and applying a so-called “successive regularization” technique to prevent catastrophic forgetting. Three examples of lower level model layers are part-of-speech (POS) tagging layer, chunking layer, and dependency parsing layer. Two examples of higher level model layers are semantic relatedness layer and textual entailment layer. The model achieves the state-of-the-art results on chunking, dependency parsing, semantic relatedness and textual entailment.

Abstract: In a boundary line recognition apparatus, based on luminance levels of an image captured by a camera, candidate edge points of a boundary line sectioning a travel road are extracted, and a candidate line of the boundary line is extracted. An apparent width of the candidate line on an image is calculated, from a width of the candidate line in a horizontal direction of the image and an angle of the candidate line relative to a vertical direction of the image. A probability of a candidate line being a boundary line is calculated to be higher, as a degree of the candidate line having characteristics as a boundary line is higher. The calculated probabilities are integrated in respect of a plurality of characteristics to recognize a boundary line. The characteristics include a ratio of the calculated apparent width to an image blur degree is larger than a predetermined value.

Abstract: Embodiments described herein provide a pipelined natural language question answering system that improves a BERT-based system. Specifically, the natural language question answering system uses a pipeline of neural networks each trained to perform a particular task. The context selection network identifies premium context from context for the question. The question type network identifies the natural language question as a yes, no, or span question and a yes or no answer to the natural language question when the question is a yes or no question. The span extraction model determines an answer span to the natural language question when the question is a span question.

Abstract: A method for evaluating robustness of one or more target neural network models using natural typos. The method includes receiving one or more natural typo generation rules associated with a first task associated with a first input document type, receiving a first target neural network model, and receiving a first document and corresponding its ground truth labels. The method further includes generating one or more natural typos for the first document based on the one or more natural typo generation rules, and providing, to the first target neural network model, a test document generated based on the first document and the one or more natural typos as an input document to generate a first output. A robustness evaluation result of the first target neural network model is generated based on a comparison between the output and the ground truth labels.

Abstract: The technology disclosed provides a so-called “joint many-task neural network model” to solve a variety of increasingly complex natural language processing (NLP) tasks using growing depth of layers in a single end-to-end model. The model is successively trained by considering linguistic hierarchies, directly connecting word representations to all model layers, explicitly using predictions in lower tasks, and applying a so-called “successive regularization” technique to prevent catastrophic forgetting. Three examples of lower level model layers are part-of-speech (POS) tagging layer, chunking layer, and dependency parsing layer. Two examples of higher level model layers are semantic relatedness layer and textual entailment layer. The model achieves the state-of-the-art results on chunking, dependency parsing, semantic relatedness and textual entailment.

Abstract: A driving supporter includes: a departure-possibility-value obtainer that obtains a departure-possibility value; a relative positional relationship obtainer that detects another vehicle located diagonally at a rear of an own vehicle and obtain a relative positional relationship between them; a support executer that executes a lane-departure prevention support; and a support controller that controls the support executer to execute the lane-departure prevention support when the departure-possibility value is greater than or equal to a threshold value. The support controller includes a threshold-value determiner that determines the threshold value to a smaller value when the relative positional relationship is a set relationship than when the relative positional relationship is not the set relationship. The set relationship is a relationship in which there is a possibility of collision of the own vehicle with the other vehicle in an event of departure of the own vehicle from a lane.

Abstract: Approaches for the translation of structured text include an embedding module for encoding and embedding source text in a first language, an encoder for encoding output of the embedding module, a decoder for iteratively decoding output of the encoder based on generated tokens in translated text from previous iterations, a beam module for constraining output of the decoder with respect to possible embedded tags to include in the translated text for a current iteration using a beam search, and a layer for selecting a token to be included in the translated text for the current iteration. The translated text is in a second language different from the first language. In some embodiments, the approach further includes scoring and pointer modules for selecting the token based on the output of the beam module or copied from the source text or reference text from a training pair best matching the source text.

Abstract: When switching from lane departure suppression control (LDA control) to lane keep assist control (LKA control), a self-vehicle is suppressed from deviating out of a traveling lane. A driving assist preferentially carries out the LKA control, but carries out the LDA control instead of the LKA control when there is a possibility that the self-vehicle may deviate out of the traveling lane under the LKA control. When the LDA control is returned to the LKA control after the lane departure avoidance operation of the self-vehicle is completed, the LDA control is continued while an angle between an orientation of the self-vehicle and the traveling lane is not a predetermined angle or less, and it is switched to the LKA control after the angle becomes the predetermined angle or less.

Abstract: In a boundary line recognition apparatus, based on luminance levels of an image captured by a camera, candidate edge points of a boundary line sectioning a travel road are extracted, and a candidate line of the boundary line is extracted. An apparent width of the candidate line on an image is calculated, from a width of the candidate line in a horizontal direction of the image and an angle of the candidate line relative to a vertical direction of the image. A probability of a candidate line being a boundary line is calculated to be higher, as a degree of the candidate line having characteristics as a boundary line is higher. The calculated probabilities are integrated in respect of a plurality of characteristics to recognize a boundary line. The characteristics include a ratio of the calculated apparent width to an image blur degree is larger than a predetermined value.

Abstract: A lane keeping traveling support apparatus includes a driving support ECU. The driving support ECU is configured to to determine a lane departure prevention torque in such a manner that a magnitude of the lane departure prevention torque becomes smaller or to stop performing a lane departure prevention control, when the lane departure prevention control is performed in place of a lane keeping assist control and a specific operation of a steering wheel is performed by a driver so as to have a direction of an own vehicle head to a lane departure direction.

Abstract: A lane keeping traveling support apparatus includes a driving support ECU. The driving support ECU is configured to determine a lane departure prevention torque in such a manner that a magnitude of the lane departure prevention torque becomes smaller or to stop performing a lane departure prevention control, when the lane departure prevention control is performed in place of a lane keeping assist control and a specific operation of a steering wheel is performed by a driver so as to have a direction of an own vehicle head to a lane departure direction.

Abstract: In a boundary line recognition apparatus, based on luminance levels of an image captured by a camera, candidate edge points of a boundary line sectioning a travel road are extracted, and a candidate line of the boundary line is extracted. An apparent width of the candidate line on an image is calculated, from a width of the candidate line in a horizontal direction of the image and an angle of the candidate line relative to a vertical direction of the image. A probability of a candidate line being a boundary line is calculated to be higher, as a degree of the candidate line having characteristics as a boundary line is higher. The calculated probabilities are integrated in respect of a plurality of characteristics to recognize a boundary line. The characteristics include a ratio of the calculated apparent width to an image blur degree is larger than a predetermined value.

Abstract: A driving supporter includes: a departure-possibility-value obtainer that obtains a departure-possibility value; a relative positional relationship obtainer that detects another vehicle located diagonally at a rear of an own vehicle and obtain a relative positional relationship between them; a support executer that executes a lane-departure prevention support; and a support controller that controls the support executer to execute the lane-departure prevention support when the departure-possibility value is greater than or equal to a threshold value. The support controller includes a threshold-value determiner that determines the threshold value to a smaller value when the relative positional relationship is a set relationship than when the relative positional relationship is not the set relationship. The set relationship is a relationship in which there is a possibility of collision of the own vehicle with the other vehicle in an event of departure of the own vehicle from a lane.

Abstract: When switching from lane departure suppression control (LDA control) to lane keep assist control (LKA control), a self-vehicle is suppressed from deviating out of a traveling lane. A driving assist preferentially carries out the LKA control, but carries out the LDA control instead of the LKA control when there is a possibility that the self-vehicle may deviate out of the traveling lane under the LKA control. When the LDA control is returned to the LKA control after the lane departure avoidance operation of the self-vehicle is completed, the LDA control is continued while an angle between an orientation of the self-vehicle and the traveling lane is not a predetermined angle or less, and it is switched to the LKA control after the angle becomes the predetermined angle or less.

Abstract: The technology disclosed provides a so-called “joint many-task neural network model” to solve a variety of increasingly complex natural language processing (NLP) tasks using growing depth of layers in a single end-to-end model. The model is successively trained by considering linguistic hierarchies, directly connecting word representations to all model layers, explicitly using predictions in lower tasks, and applying a so-called “successive regularization” technique to prevent catastrophic forgetting. Three examples of lower level model layers are part-of-speech (POS) tagging layer, chunking layer, and dependency parsing layer. Two examples of higher level model layers are semantic relatedness layer and textual entailment layer. The model achieves the state-of-the-art results on chunking, dependency parsing, semantic relatedness and textual entailment.