International Business Machines Corporation

Fujitsu Corporation

Kabushiki Kaisha Toshiba

Nissan Motor Company Limited

Sumitomo Metal Industries, Inc.

Nuance Communications, Inc.

Canon Kabushiki Kaisha

Sony Corporation

Subaru Corporation

Daifuku Co., Ltd.

Fukuda Denshi Co., Ltd.

Sumitomo Industries, Ltd.

Nippon Steel & Sumitomo Metal Corporation

Denka Company Limited

Takashi Fukuda

Osamu Ichikawa

Masayuki A Suzuki

Gakuto Kurata

Toshifumi Yamamoto

Bhuvana Ramabhadran

Masafumi Nishimura

Yasuhiro Kanishima

Kazuyuki Saito

Norikatsu Chiba

Hiroaki Kikuchi

Samuel Jonathan Thomas

Toru Nagano

Yutaka Oki

Shigeyasu Iwata

Kenta Watanabe

Shimpei Yotsukura

Shunsuke Ishikawa

Futoshi Iwama

Akira Tamura

Yukihiro Sugimoto

Masaki Komedani

Mitsunori Abe

Takashi Watanabe

Ryuki Tachibana

Vaibhava Goel

Takahito Tashiro

Nobuyasu Itoh

Kimio Miseki

Hayato Chikugo

Jun Iwasaki

Yuichi Suzuki

Noriyuki Negi

Masaki Tosaka

Yohei Taniguchi

Mitsunori Kumada

Tohru Nagano

Steven John Rennie

Shigeyoshi Nishihara

Ryoichi Shimoi

Tatsuo Nagata

Daisuke Sato

Masatoshi Iio

Kazuhiro Kageyama

Daita Tsubamoto

Susumu Fukuda

Hiroo Nagano

Kazuaki Takao

Kiyoshi Sudani

Masato Kano

Toru Iwahashi

Takashi Iimori

Yasutaka Kotani

Masahiro Fukuoka

Yoshinobu Matsukawa

Sukeyoshi Yamamoto

Koji Moriguchi

Hajime Murakami

Mitsuharu Yonemura

Kenji Aihara

Yasuhiro Yamashita

Atsushi Okada

Hitoshi Igarashi

Hikoyuki Kawata

Yasuyuki Itou

Kumiko Teramae

Tsukasa Murakami

Yoshikazu Hirano

Yuta Akamatsu

Atsushi Ooma

Mitsunobu Okano

Yukio Karasawa

Hyoji Hagita

Seiho Sugawara

Masayuki Suzuki

Hirotaka Kamimura

Toru Shimadoi

Takanaga Yokosawa

Hidenobu Shiihara

Natsumi Kaneko

Akira Matsunaga

Hirohito Takeichi

Megumi Tanaka

Mitsuhiro Kokubo

Makoto Araki

Takashi Hirashima

Hiroshi Nagai

Syozi Nishimura

Satoshi Nishimura

Yasuhiro Muramatsu

Hirofumi Fujisaki

Taiichirou Sakaguchi

Yasuyuki Nakamura

Terumi Oda

Tetsu Ohno

Akira Aida

Takuo Mizoguchi

Katsuyoshi Kajiyama

Kazutosi Ogata

Hideo Imamura

A computer-implemented method for preparing training data for a speech recognition model is provided including obtaining a plurality of audio data sets, each audio data set having a different acoustic feature and sorting sentences from the plurality of audio data sets so that similar sentences from different audio data sets are positioned closely, while imposing a weak constraint on audio length, to train the speech recognition model.

Training data sequence for RNN-T based global English model

An approach to identifying alternate soft labels for training a student model may be provided. A teaching model may generate a soft label for a labeled training data. The training data can be an acoustic file for speech or a spoken natural language. A pool of soft labels previously generated by teacher models can be searched at the label level to identify soft labels that are similar to the generated soft label. The similar soft labels can have similar length or sequence at the word phoneme, and/or state level. The identified similar soft labels can be used in conjunction with the generated soft label to train a student model.

Alternative soft label generation

A driving assistance method uses a controller to perform automated lane change to control the own vehicle in such a way that the own vehicle makes a lane change under a condition that the passenger is at least in contact with the steering wheel, the driving assistance method causing the controller to perform processing including: selecting a target lane; and when automated lane change from a current lane to the target lane includes a first automated lane change from a first lane to a second lane and a second automated lane change from the second lane to a third lane, performing lane-keeping assistance under a condition that the passenger is at least in contact with the steering wheel from an end of control of the first automated lane change until a start of control of the second automated lane change.

Driving assistance method and driving assistance device

Techniques for training a neural transducer-based automatic speech recognition model to be robust against background additive noise and thereby reducing insertion errors. In one aspect, a method of training an automatic speech recognition model includes: generating a modified training data set from an initial training dataset by concatenating one-word utterances with a preceding or a succeeding sentence in the initial training dataset based on a duration of silence between the one-word utterances and the preceding or the succeeding sentence; and training the automatic speech recognition model using the modified training data set.

Reducing insertion errors in neural transducer-based automatic speech recognition

Methods and systems for training a neural network include clustering a full set of training data samples into specialized training clusters. Specialized teacher neural networks are trained using respective specialized training clusters of the specialized training clusters. Soft labels are generated for the full set of training data samples using the specialized teacher neural networks. A student model is trained using the full set of training data samples, the specialized training clusters, and the soft labels.

Knowledge distillation using deep clustering

Systems, computer-implemented methods, and computer program products to facilitate multi-task training a recurrent neural network transducer (RNN-T) using automatic speech recognition (ASR) information are provided. According to an embodiment, a system can comprise a memory that stores computer executable components and a processor that executes the computer executable components stored in the memory. The computer executable components can include an RNN-T that can receive ASR information. The computer executable components can include a voice activity detection (VAD) model that trains the RNN-T using the ASR information, where the RNN-T can further comprise an encoder and a joint network. One or more outputs of the encoder can be integrated with the joint network and one or more outputs of the VAD model.

Voice activity detection integration to improve automatic speech detection

A method, computer system, and a computer program product for audio data augmentation are provided. Sets of audio data from different sources may be obtained. A respective normalization factor for at least two sources of the different sources may be calculated. The normalization factors from the at least two sources may be mixed to determine a mixed normalization factor. A first set of the sets may be normalized by using the mixed normalization factor and to obtain training data for training an acoustic model.

Acoustic data augmentation with mixed normalization factors

A computer-implemented method for preparing training data for a speech recognition model is provided including obtaining a plurality of sentences from a corpus, dividing each phoneme in each sentence of the plurality of sentences into three hidden states, calculating, for each sentence of the plurality of sentences, a score based on a variation in duration of the three hidden states of each phoneme in the sentence, and sorting the plurality of sentences by using the calculated scores.

Data sorting for generating RNN-T models

In a method for improving generation and relevancy of search results, a processor receives a search query comprising a search term. A processor generates a document group based on the search query and at least one synonym related to the search term in a synonym dictionary. The synonym dictionary may include search document attributes for base words and synonyms of the base words. A processor extracts, from the document group, an extracted document having a document attribute matching a search document attribute of the at least one synonym. A processor lists the extracted document as a search result.

Narrowing synonym dictionary results using document attributes

Some embodiments of the present invention are directed to techniques for training teacher neural networks (TNNs) and student neural networks (SNNs). A training data set is received with a lossless set of data and a corresponding lossy set of data. Two branches of a TNN are established, with one branch trained using the lossless data (a lossless branch) and one trained using the lossy data (a lossy branch). Weights for the two branches are tied together. The lossy branch, now isolated from the lossless branch, generates a set of soft targets for initializing an SNN. These generated soft targets benefit from the training of lossless branch through the weights that were tied together between each branch, despite isolating the lossless branch from the lossy branch during soft-target generation.

Training teacher machine learning models using lossless and lossy branches

A student neural network may be trained by a computer-implemented method, including: inputting common input data to each teacher neural network among a plurality of teacher neural networks to obtain a soft label output among a plurality of soft label outputs from each teacher neural network among the plurality of teacher neural networks, and training a student neural network with the input data and the plurality of soft label outputs.

Training of student neural network with teacher neural networks

Embodiments are directed to a system, computer program product, and method for text mining, and dynamic facet and facet value management and application to a document collection. Two or more words from a first document collection are extracted, with the extracted words being associated with an applied annotation. At least one word is selected from the extracted words, designated as a facet, and a value is selectively added to the facet. An analysis of the added value is dynamically performed, and a dictionary with the annotation, facet, and values is constructed and the dictionary is applied to the document collection. A targeted list of documents is returned from the dictionary application to the document collection.

Dictionary editing system integrated with text mining

A student neural network may be trained by a computer-implemented method, including: selecting a teacher neural network among a plurality of teacher neural networks, inputting an input data to the selected teacher neural network to obtain a soft label output generated by the selected teacher neural network, and training a student neural network with at least the input data and the soft label output from the selected teacher neural network.

Training of student neural network with switched teacher neural networks

Fusion of neural networks is performed by obtaining a first neural network and a second neural network. The first and the second neural networks are the result of a parent neural network subjected to different training. A similarity score is calculated of a first component of the first neural network and a corresponding second component of the second neural network. An interpolation weight is determined for the first and the second components by using the similarity score. A neural network parameter of the first component is updated based on the interpolation weight and a corresponding neural network parameter of the second component to obtain a fused neural network.

Fusion of neural networks

A first user request which specifies a target document set wherein a first subset of the documents is flagged by a user. A primary flag table is created for the target document set. A first document subset is created matching the first user request. It is determined whether a number of flagged documents exceeds a first threshold. If so, a secondary flag table is created for the first document subset and flag data corresponding to the first document subset is stored in the secondary flag table. The flag data in the secondary flag table is merged into the primary flag table.

Document flagging based on multi-generational complemental secondary data

Growing community of inventors

Yokohama, Japan

Takashi Fukuda