International Business Machines Corporation

Oki Data Corporation

Mitsubishi Heavy Industries Limited

Mitsubishi Electric Corporation

Mitsubishi Gas Chemical Company, Inc.

Mgc Filsheet Co., Ltd.

Other

Tokyo Electron Limited

Takuya Goto

Tohru Hasegawa

Makoto Hirai

Masato Ishikawa

Hisanori Ikeda

Masahide Kinami

Katsuya Ishigami

Akimoto Suzuki

Katsumi Nakamura

Hiroki Nakano

Masaharu Sakamoto

Shinji Kawai

Asako Ono

Yosuke Naito

Xiangning Liu

Takahiro Inagaki

Youhei Koyanagi

Hiroki Azuma

Kentaro Hasegawa

Naoki Goto

Yasushi Oba

Toshihiro Tombe

Kenichi Tsuruda

Tsuyoshi Kaneko

Yutaka Oishi

Hiroaki Kikuchi

Tatsuya Murakami

Katsuyuki Ito

Tetsu Koyama

Tetsuo Aoyama

Hiroshi Matsunaga

Takayuki Shimizu

Masaki Saitoh

Fumitaka Ozeki

Kojiro Abe

Chiaki Oishi

Ryuji Sotoaka

Hiroyuki Mori

Yoshiya Kimura

Takayuki Niuya

Fukusaburo Ishihara

Naoto Azusawa

Hirokazu Takenaka

Shuji Umehara

Yasuo Oose

Masanobu Mizukami

Hirotaka Tsuruya

Kazuhito Kojima

Teruo Soeda

Takuya Niimura

Hideto Gotoh

Sachio Takeyama

Hideyuki Suzuki

Hiroaki Sato

Takahiro Matsumoto

Daisuke Hayashi

Hiroshi Ohmura

Masakazu Sato

Takeshi Yamada

Junichi Ito

Takahito Tashiro

Toshiaki Yamada

Yoshiroh Kamiyama

Shunsuke Ishikawa

Takahiro Kubota

Hiroshi Omura

Kenta Watanabe

Masahiro Fujiwara

Kenichi Takasaki

Tetsuya Tazawa

Kazuto Nakamura

Takemasa Ishikuro

Shuhei Yokota

Takanori Mimura

Masukazu Hirata

Akira Kono

Hisashi Soga

Michinobu Takahagi

Susumu Yamamoto

Den Ozaki

Shota Sugai

Toshihiro Hotta

Junta Watanabe

Shigeo Horiuchi

Hiroto Mori

Jiro Wada

Kiyoshi Yasutomi

Kazuoki Katayama

Hiroshi Taneda

Ryu Iwasaki

Yuichi Sasano

Tsukasa Hirayama

Kei Yoshitomi

Kouji Takenaka

Kazuki Matsumaru

Junichi Takeshita

Syouta Sugai

Mechanisms are provided for writing files to tape storage. A set of files to be written to the tape storage are identified and, for each file, metadata associated with the file indicating identifiers of users associated with the file is retrieved. For each file, a predicted frequency metric, specifying a predicted frequency of access of the file, is determined based on the identifiers of users associated with the file specified in the associated metadata. The files are sorted to generate a ranked set of files in which each file is ranked relative to the other files in the set of files based on their corresponding predicted frequency metric. The ranked set of files is written to the tape storage in descending order of predicted frequency metric such that files having relatively larger predicted frequency metrics are written relatively closer to a starting boundary of a partition of the tape storage.

Ordering files in magnetic tape storage

A feature of a subject file is identified. The feature is compared to a historical feature of a historical file. A similarity between the subject file and the historical file is calculated based on the comparing. A historical success metric for the historical file is identified. An intensive text extraction success value for the subject file is calculated based on the similarity and the historical success metric. Based on the intensive extraction success value, whether an intensive text extraction method should be performed on the subject file is determined.

Selecting files for intensive text extraction

A method, system, and computer program product are disclosed. The method includes receiving a set of documents, selecting a topic, and determining that a first document from the set contains a topic label for the topic. The method also includes generating a topic sentiment score for the first document and adding the topic sentiment score to a set of training data. Additionally, the method includes determining that a second document from the set does not contain the topic label, generating an average sentiment score for the second document, and generating a bias factor for the average sentiment score.

Topic classifier with sentiment analysis

An indoor unit includes a rear housing in which a heat exchanger and an electrical component box are disposed, a front housing disposed in front of the heat exchanger such that the front housing is removable, a front design panel disposed in front of the front housing such that the front design panel is removable, a securing and storage part into which a communication adapter is secured, and a temporary placement part for the communication adapter. The securing and storage part is disposed in the front housing such that the securing and storage part is exposed when the front design panel is removed. The temporary placement part is used to temporarily place the communication adapter on the temporary placement part.

Indoor unit of air-conditioning apparatus

Described is a content mining system comprising a crawler configured to retrieve a plurality of files from a data storage system. The content mining system further comprises a plurality of converters configured to extract data from the plurality of files retrieved by the crawler from the data storage system, where each of the plurality of converters is configured to process a respective type of data. The content mining system further comprises a plurality of queues interposed between the crawler and the plurality of converters, where each queue is associated with a single converter.

Archived data crawling

Methods and systems are provided for training a neural network with augmented data. A dataset comprising a plurality of classes is obtained for training a neural network. Prior to initiation of training, the dataset may be augmented by performing affine transformations of the data in the dataset, wherein the amount of augmentation is determined by a data augmentation variable. The neural network is trained with the augmented dataset. A training loss and a difference of class accuracy for each class is determined. The data augmentation variable is updated based on the total loss and class accuracy for each class. The dataset is augmented by performing affine transformations of the data in the dataset according to the updated data augmentation variable, and the neural network is trained with the augmented dataset.

Neural networks using intra-loop data augmentation during network training

A pipe joint structure includes: a union; a flare nut that includes a female thread formed at one of axially opposite ends of the flare nut, and a first annular surface formed at the other of the axially opposite ends of the flare nut; a cylindrical sleeve that houses the flare nut such that the flare nut can be rotate and moved forward and backward; a rotary ring that has a second annular surface located to face the first annular surface, and is attached to the sleeve such that the rotary ring can be rotated for the union; and a pressing member that presses the flare nut toward the rotary ring. At the second annular surface, a claw is formed in such a manner as to protrude towards the flare nut. The first annular surface includes an engagement surface that is formed in such a manner to face one side in a circumferential direction and a helically inclined surface that is adjacent to the engagement surface in the circumferential direction and formed helically in such a way as to incline and face the opposite side of the one side in the circumferential direction.

Pipe joint structure and air conditioner

To provide a robot control system and a robot control method capable of placing a component grasped by a robot hand at an accurate location on another member. A robot control system is provided with: a robot hand configured to grasp a clip; a camera configured to capture an image of the clip grasped by the robot hand, a calculation unit configured to calculate a position of the clip or an inclination of a component based on an imaging result of the clip captured by the camera, and a robot control unit configured to control the robot hand to adjust, based on the position of the clip or the inclination of the component calculated by the calculation unit, a position or an inclination of the robot hand and move the clip to a stringer.

Robot control system and robot control method

A method for a multi-channel search includes receiving a specific post selection submitted in a first channel and query text associated with the specific post in the first channel, where the query text includes one or more words for performing a query evaluation. The method receives a query location which includes a plurality of channels for performing the query evaluation. Responsive to extracting one or more ranking factors from the specific post selection, the method performs the query evaluation at the query location to collect a plurality of potential matching posts based on the one or more ranking factors. The method reorders the plurality of potential matches according to a plurality of determined scores for the plurality of potential matches. The method displays two-dimensional search results based on a timeline and channel relevancy, where the plurality of channels is ordered according to channel relevancy compared to the first channel.

Multi-channel search

Image classification using a generated mask image is performed by generating a mask image that extracts a target area from an input image, extracting an image feature map of the input image by inputting the input image in a first neural network including at least one image feature extracting layer, masking the image feature map by using the mask image, and classifying the input image by inputting the masked image feature map to a second neural network including at least one classification layer.

Image classification using a mask image and neural networks

An approach is provided in which a method, system, and program create a plurality of page clusters in feature space from a plurality of feature vectors corresponding to a plurality of unstructured pages. The method, system, and program product assign one of a plurality of machine learning models to each one of the plurality of page clusters based on a relationship in the feature space between the plurality of page clusters and a plurality of training clusters corresponding to the plurality of machine learning models. The method, system, and program product identify one of the plurality of page clusters that corresponds to a selected one of the plurality of unstructured pages, and transform the selected unstructured page into a structured page using a selected one of the plurality of machine learning models assigned to the identified page cluster.

Adaptive document understanding

Facilitating annotation of document elements includes obtaining annotations indicating labels applied to document elements of a digital document to be presented in an annotation graphical user interface (GUI), the digital document including the document elements, the document elements being for visual presentation in the annotation GUI to a user, and presenting the annotation GUI on a display device for the user, the annotation GUI configured to facilitate user submission of a respective label for each document element of the document elements, where the presenting presents, as part of the annotation GUI, at least a portion of the digital document including the document elements, and, for each document element, a respective label indicator correlating to the document element and visually indicating a current label selected for that document element and a label submission status for that document element that reflects whether the user has confirmed a label for that document element.

Facilitating annotation of document elements

A front panel for on-board liquid crystal displays has a high hardness resin composition (B) on at least one side of a layer containing a resin (A) having a polycarbonate resin (a1). The front panel satisfies the following conditions (i) to (iv): (i) the thickness of the layer containing a high hardness resin composition (B) is 10 to 250 &#x3bc;m, and the total thickness of the layer containing a resin (A) comprising a polycarbonate resin (a1) and the layer containing a high hardness resin composition (B) is 100 to 3,000 &#x3bc;m; (ii) the high hardness resin composition (B) consists of any one of specific resin compositions (B1) to (B3); (iii) the retardation of the front panel is 3,000 nm or more; and (iv) the standard deviation of the second derivative of the irregular shape of the hard coat layer having irregularities is 0.1 or more.

Front plate for onboard LCD device

Storage pools having categorized archival storage tapes according to content of files stored on each tape. Archiving new files into the storage pools according to content of the new files for efficient retrieval of files from storage tapes organized by category of content. Recall of multiple files from archival storage tapes to primary storage of an hierarchical storage management system is efficiently performed such that tape mounting operations are minimized.

Content driven storage and retrieval of files

Provided is a structural body, formed in a plate-like shape and positioned with respect to an assembly jig including a positioning pin and supporting a first surface of the structural body, including a positioning hole formed through an end portion, a positioning protrusion being inserted into the positioning hole; and a pair of first position detection areas formed on a second surface of the end portion through which the positioning hole is formed. The positioning hole is formed apart from an axis line connecting the pair of first position detection areas.

Structural body

Growing community of inventors

Tokyo, Japan

Takuya Goto