Hitachi, Ltd.

Hitachi Seibu Soft Ware Co., Ltd.

N.v. Philips' Gloeilampenfabrieken

Junichi Higashino

Atsushi Hatakeyama

Hiromichi Fujisawa

Yuki Aoyama

Yasuaki Nakano

Hitoshi Matsushima

Kanji Kato

Hisao Ogata

Toshihiro Hananoi

Masahiro Abe

Hiroshi Sakou

Takafumi Miyatake

Satoshi Yoshizawa

Seiji Kashioka

Masakazu Ejiri

Yoshiki Matsuda

Eric H Persoon

Tetsuya Hashimoto

Akisuke Naruse

Kazuhiko Kawasaki

Miyuki Hara

Takako Morita

Shigeko Ohtani

A document difference extraction method and apparatus which is used for extracting the difference between structured documents properly meeting the sense of a document editor taking the logical meaning and structure of the structured documents into consideration. Structured documents are edited and stored in a memory unit by a document editing program. With reference to a comparison criterion set for the logical structure of each structured document before and after edition, the logical structure of the structural documents before and after edition read from the memory unit is analyzed by a structured document parsing program, and the difference between the structured documents is extracted by a structured document difference extraction program in such a manner as to satisfy the comparison criterion in accordance with the result of parsing. The comparison criterion assumes the form of a table containing a plurality of tags representing logical structures and types of tags for the comparison criterion. The tag types for comparison criterion include tags having contents which are compared only when the particular tags are coincident with each other, tags having contents which are ignored at the time of comparison, a set of tags having the same logical meaning, and a set of tags having contents which are not compared with each other.

Method and apparatus for structured document difference string extraction

A document information classification method and apparatus for classifying a document group and arranging a classified result hierarchically on the basis of key words given to the document group and words appearing in documents without dependence on a prescribed classification system. The document group of a document data base and a key word group given to each document of a key word data base are managed by a data management unit. A document classification unit classifies documents into folders on the basis of individual key words and stores them. The folders having similar document groups are integrated. Whether the integration is effective or not is judged upon integration. Whether the inside of the integrated folder and the inside of unintegrated folders can be classified in detail or not is judged and a hierarchical classification system is prepared. A classified result is produced in CRT by a classified result output unit to provide environment in which a user can read out the classified result.

Method and apparatus for classifying document information

A perceptron device has processing elements arranged in an input layer, one or more hidden layers, and an output layer. The training method of the perceptron device uses the back propagation method. In order to increase learning speed, the learning rate for the updates of a particular processing element is a function f(M,N,K) of the number N of inputs to that element, the number K of outputs of that element, and the number N of inputs to the next layer. Function f increases with increasing M, and decreases with increasing K and N.

Method for adjusting network parameters in a multi-layer perceptron

A document storage and retrieval system is provided with means for storing a document body in the form of image, means for storing text information in the form of a character code string for retrieval, means for executing a retrieval with reference to the text information, and means for displaying a document image relating thereto on a retrieval terminal according to the retrieval result. Such a form of the system is available for retrieving the full contents of a document and also for displaying the document body printed in a format easy to read straight in the form of image. Accordingly, users are capable of retrieving documents with arbitrary words and also capable of reading even such a document as is complicated to include mathematical expressions and charts through a terminal in the form of image, the same as on paper. Further, the invention provides a system wherein the text information for retrieval is extracted automatically from the document image through character recognition. Since a precision of the character recognition has not been satisfactory hitherto, a visual retrieval and correction have been carried out without fail by operators. However, there is no necessity for the operators to attend therefor according to the invention. Thus, the text information for retrieval can be generated at the cost of practical time and money even in case of volumes of documents.

Document storage and retrieval system for storing and retrieving

A neural network learning system using back propagation and a method of learning pattern showing are disclosed. In the case where a supervised signal contains an error or a pattern difficult to learn, the error is detected and the particular pattern is automatically removed during calculations of learning iterations to conduct rightly and accelerate the learning. The learning history for each pattern is stored to detect inconsistent and difficult-to-learn patterns, which are prevented from being shown to the network by a pattern showing control during the next learning iteration. As a result, an inconsistent or difficult-to-learn pattern which may be contained in a learning pattern set of input and supervised patterns is removed during learning iterations thereby to permit early completion of the learning process.

Learning system and a learning pattern showing method for a neural

A neural network (100) has an input layer, a hidden layer, and an output layer. The neural network stores weight values which operate on data input at the input layer to generate output data at the output layer. An error computing unit (87) receives the output data and compares it with desired output data from a learning data storage unit (105) to calculate error values representing the difference. An error gradient computing unit (81) calculates an error gradient, i.e. rate and direction of error change. A ratio computing unit (82) computes a ratio or percentage of a prior conjugate vector and combines the ratio with the error gradient. A conjugate vector computing unit (83) generates a present line search conjugate vector from the error gradient value and a previously calculated line search gradient vector. A line search computing unit (95) includes a weight computing unit (88) which calculates a weight correction value. The weight correction value is compared (18) with a preselected maximum or upper limit correction value (.kappa.). The line search computing unit (95) limits adjustment of the weight values stored in the neural network in accordance with the maximum weight correction value.

Learning method and apparatus for neural networks and simulator with

In a document information system and a method for realizing such a system, document data are transferred and displayed at a high speed. In response to an operation made from a terminal apparatus, the document data which have been stored via communication apparatus into a data storage apparatus are read out, and the read document data are temporarily stored before being transmitted by employing the communication apparatus. In the terminal apparatus, the received document data are temporarily stored as display data, and additionally the display data are swapped to a secondary storage device in such a case that a total amount of the temporarily stored data exceeds a predetermined capacity. Then, the display data may be transferred from either the terminal apparatus, or the secondary storage device to a display buffer and then may be displayed in a unit of representation.

Document information system and document data transfer and display method

A learning control method reduces overall learning time by displaying data related to an appropriate determination of learning protraction and a proper restoring method. Prior to initiating the learning, the user is inquired about the current problem and a problem data set representing items associated with the problem is obtained. Evaluation data indicating a state of learning obtained during the learning on the current problem is sequentially stored and displayed. When there is a high possibility of learning protraction during the learning, a message informing the user is displayed. When the learning is stopped by the user in this case, the problem data set and evaluation data set are stored. Then, a list of restoring methods is displayed and a particular restoring method is selected by the user once the learning is stopped. The learning is restarted on the current problem in accordance with the selected restoring method.

Method of and system for controlling learning in neural network

A multi-layer neural network comprising an input layer, a hidden layer and an output layer and a learning method for such a network are disclosed. A processor belonging to the hidden layer stores both the factors of multiplication or weights of link for a successive layer nearer to the input layer and the factors of multiplication or weights of link for a preceding layer nearer to the output layer. Namely, the weight for a certain connection is doubly stored in processors which are at opposite ends of that connection. Upon forward calculation, the access to the weights for the successive layer among the weights stored in the processors of the hidden layer can be made by the processors independently from each other. Similarly, upon backward calculation, the access to weights for the preceding layer can be made by the processors independently from each other.

Multi-layer network and learning method therefor

A document storage and retrieval system stores a document body in the form of an image, storing text information in the form of a character code string for retrieval, and executing a retrieval with reference to the text information, followed by displaying a document image relating thereto on a retrieval terminal according to the retrieval result. Such a form of the system is available for retrieving the full contents of a document and also for displaying the document body printed in a format easy to read straight in the form of an image.

Document storage and retrieval system

An image understanding system of this invention uses a grammer describing a document image, and represents the structure of an unknown input image by parsing a statement (the structure of the grammar) written in accordance with this grammer. In other words, the grammer describes an image as substructures and the relative relation between them, and when the substructures and their relative relation are identified in parsing, search is then made whether or not the substructures and their relative relation exist in an unknown input image. The structure of the unknown input image is represented on the basis of the result of this search.

Image understanding system

A graphic data management system including a segment table in which several kinds of graphic data placed in a multi-dimensional space are stored correspondingly to graphic element numbers, and an index table in which, for each of the cell numbers provided when the multi-dimensional coordinate system is partitioned into predetermined cells, and all the graphic element numbers at least a part of which included in a cell are stored. The graphic data management system carries out the search, addition and deletion referring to a segment table and the index table. The index table includes a cell pointer table in which pointers indicative of the addresses where the graphic element numbers are stored are listed for each cell number, a cell table in which the graphic element number and concatenation pointers which indicate the storing addresses of the subsequent graphic element numbers are listed as pairs, and a space management pointer indicative of the address of a free space area in the cell table.

Management system of graphic data

A method for controlling a multi-joint robot having redundant operating joints, wherein joint velocities are calculated for non-redundant joints with redundant joints being fixed for each of at least a portion of the combinations of the non-redundant joints excluding redundant joints capable of attaining the movement of the robot end at a command velocity specified to the robot, by inverse-coordinate-transformation from the spatial coordinate to a joint coordinate inherent to the robot, on the basis of joint angles detected by an encoder mounted on the joints of the robot and the joint velocities are averaged or nearly averaged to obtain the average value or nearly average value as a desired joint velocity for the joints. The desired joint velocity is supplied to a velocity control unit of the robot as a controlled variable.

Growing community of inventors

Kodaira, Japan

Junichi Higashino