利索-全球专利检索

  1. 登录
  2. 注册

搜索结果

61 条记录 (耗时 0.012 秒)

    Natural language processing system
    5.
    发明公开
    Natural language processing system 失效
    Naturspra​​chenverarbeitungssystem。

    公开(公告)号：EP0625758A1

    公开(公告)日：1994-11-23

    申请号：EP94300701.3

    申请日：1994-01-31

    申请人： International Business Machines Corporation

    发明人： Sharman, Richard

    IPC分类号： G06F15/419 ,  G06F15/38

    CPC分类号： G06F17/2735 ,  G06F17/2715 ,  G06F17/2775 ,  G06F17/2795

    摘要： A B-tree 21 is used to store natural language data, for example as part of a speech recognition or speech synthesis system. The B-tree is arranged in a hierarchy, with each node pointing to two nodes in the level below. Each node contains a test value (a word), and data relating to that test value (or a reference to a storage location where the data is maintained). The data is accessed by starting at the top of the tree and comparing the desired word with the test value for that node. Depending on the relative alphabetical ordering of the desired word and the test value, the appropriate branch from that node is followed. This process is continued down the tree until a test value corresponding to the desired word is located. The B-tree is arranged so that frequency of occurrence of the test values in natural language decreases substantially monotonically as the tree is descended.

    摘要翻译： B树21用于存储自然语言数据，例如作为语音识别或语音合成系统的一部分。 B树以层次结构排列，每个节点指向下面两个级别的节点。 每个节点包含一个测试值（一个字），以及与该测试值相关的数据（或对保存数据的存储位置的引用）。 通过从树的顶部开始并将所需的字与该节点的测试值进行比较来访问数据。 根据所需单词和测试值的相对字母排序，跟随该节点的适当分支。 该过程在树下继续下去，直到找到与所需单词对应的测试值。 布置B树，使得自然语言中测试值的出现频率随着树下降而基本上单调减少。

    Concurrent access to indexed data files
    6.
    发明公开
    Concurrent access to indexed data files 失效
    并行访问索引数据文件

    公开(公告)号：EP0559358A3

    公开(公告)日：1994-02-16

    申请号：EP93301289.0

    申请日：1993-02-22

    申请人： INTERNATIONAL BUSINESS MACHINES CORPORATION

    发明人： Scott, Jonathan Andrew

    IPC分类号： G06F15/403 ,  G06F15/419

    CPC分类号： G06F17/30327 ,  G06F17/30348 ,  G06F17/30351 ,  Y10S707/956 ,  Y10S707/99938

    摘要： A database management system including an index (10) with parallel read access for large amounts of keyed information. The index can be referenced for read-only access (217, 218, 219) by any number of simultaneous processes (200, 201, 202) in parallel with being updated by a single process (203), where the read-only access does not involve any locking or waiting.

    摘要翻译： 一种数据库管理系统，包括一个索引（10），具有对大量键控信息的并行读取访问。 该索引可以被任何数量的同时进程（200,201,202）与被单个进程（203）更新并行地进行只读访问（217,218,219）参考，其中只读访问确实 不涉及任何锁定或等待。

    Method and system for enhanced efficiency of data recovery in balanced tree memory structures
    7.
    发明公开
    Method and system for enhanced efficiency of data recovery in balanced tree memory structures 失效
    数据恢复在平衡树记忆体结构中提高效率的方法与系统

    公开(公告)号：EP0547992A3

    公开(公告)日：1993-12-01

    申请号：EP92480174.9

    申请日：1992-11-19

    申请人： INTERNATIONAL BUSINESS MACHINES CORPORATION

    发明人： Keller, Robert S. ,  Mitchell, Michael O. ,  Stringfellow, Dale B., Jr. ,  Williams, Marvin L.

    IPC分类号： G06F15/419

    CPC分类号： G06F17/30961

    摘要： A method and system for enhancing the efficiency of data recovery in a multilevel balanced tree memory structure having a root node and multiple leaf nodes stored within a data processing system. After checking to determine if the multilevel balanced tree memory structure is of sufficient size and complexity to warrant optimization, and after determining that sufficient memory space is available within the data processing system, the entire multilevel balanced tree memory structure is recopied within the data processing system utilizing a post order visitation process. This process is accomplished in a manner such that adjacent leaves are as closely located as possible within memory, thereby minimizing the amount of memory which must be traversed during a data search. In one embodiment of the present invention, a virtual memory data processing system is utilized and the entire multilevel balanced tree memory structure is recopied in a manner such that as many adjacent leaves as possible are stored within each page of memory therein.

    History content representation system for information processors
    8.
    发明公开
    History content representation system for information processors 失效
    信息处理器的历史内容表示系统

    公开(公告)号：EP0528704A3

    公开(公告)日：1993-10-27

    申请号：EP92402028.2

    申请日：1992-07-10

    申请人： INSTITUTE FOR PERSONALIZED INFORMATION ENVIRONMENT

    发明人： Yoshimura, Tetsuya, c/o Institute for Personalized ,  Ueda, Hirotada, c/o Institute for Personalized ,  Hidai, Yutaka, c/o Institute for Personalized ,  Uyama, Masashi, c/o Institute for Personalized

    IPC分类号： G06F15/403 ,  G06F15/419

    CPC分类号： G06F17/30014 ,  G06F11/32 ,  G06F11/3476

    摘要： In an information processor which manages operational history data or status data to revert an application back to any desired point in the past or make use of the past operations, an interface screen 1 has thereon cards 2 as an object which is manipulated by the user to make use of the history, and the contents of the history are represented, while the order of said cards 2 being put one upon another is associated with the time series of the history. A card representation is provided on the screen presented by the application to the user at any point in the past, or the screen to provide a card representation is defined by images which display alternately screen representations before and after a change in the application screen by the operation performed by the user at any point in the past, while the before- and after-change representations are displayed for different periods of time. A tag 5 to be provided to the card is incorporated in the interface screen, and objects of stacks 3 and 4, which are a location where a bundle of cards arranged according to the time series of the card history is placed and to which instructions as to an user's operation performed on said card bundle are inputted, are incorporated on the user's interface screen. The stack 3 cooperates with the application, letting the application execute an operational history corresponding to a card sequence on said stack and the results be represented on said cards, and the card sequence is inserted into the stack cooperative with the application, thereby letting the application execute the operational history.

    Hypergraphics
    9.
    发明公开
    Hypergraphics 失效

    公开(公告)号：EP0484262A3

    公开(公告)日：1993-04-21

    申请号：EP91480155.0

    申请日：1991-10-10

    申请人： INTERNATIONAL BUSINESS MACHINES CORPORATION

    发明人： Brown, Jerry Richard ,  Lascelles, Martin Christopher ,  Wong, Michael (NMI)

    IPC分类号： G06F15/419 ,  G06F15/40

    CPC分类号： G06F17/30014

    摘要： Storing records of extended graphics information and retrieving such information according to pointers asociated with selected displayed entities. Certain classes of function related to the extended graphics information may be grouped under selective application features.

    摘要翻译： 存储扩展图形信息的记录并根据与所选显示实体关联的指针检索这些信息。 与扩展图形信息相关的某些类别的功能可以在选择性应用特征下分组。

    Linked data structure
    10.
    发明公开
    Linked data structure 失效
    VerknüpfteDatenstruktur。

    公开(公告)号：EP0514112A2

    公开(公告)日：1992-11-19

    申请号：EP92304236.0

    申请日：1992-05-12

    申请人： International Business Machines Corporation

    发明人： Seigh, Joseph William, II

    IPC分类号： G06F15/40 ,  G06F12/02 ,  G06F15/419

    CPC分类号： G06F9/52 ,  G06F17/30958 ,  Y10S707/99938 ,  Y10S707/99952

    摘要： A system and method for maintaining linked data structures stored in a computer system capable of processing the stored data as addressable object nodes, so that any data object node may be modified prior to the completion of outstanding read only accesses to that node. The system and method maintain an access vector for each node. The access vectors include an access counter which counts the number of read only accesses from a present node to the next node, and a link pointing to a next node. The number of read only accesses in effect for a node is the sum of access counts of all access vectors pointing to that node, minus the ADJ and minus the sum of access counts of all access vectors with access to that node. A node in the list can be replaced or deleted by first changing the pointers of all nodes that point to the node to point to a new one. Then, modifying the access count of the node by subtracting the maximum value of the access counts pointing to the node from the count in the node. Read only accesses out of the node continue to increment the access count, and when it reaches zero the node can be deleted. Because the pointers to the node had been changed, read only processes can continue without pausing for the change or replacement event.

    摘要翻译： 一种用于维护存储在能够将存储的数据作为可寻址对象节点处理的计算机系统中的链接数据结构的系统和方法，使得可以在完成对该节点的未完成的只读访问之前修改任何数据对象节点。 系统和方法维护每个节点的访问向量。 访问向量包括访问计数器，其对从当前节点到下一个节点的只读访问次数进行计数，以及指向下一个节点的链接。 对节点有效的只读访问次数是指向该节点的所有访问向量的减去ADJ的访问计数之和，并减去访问该节点的所有访问向量的访问计数之和。 可以通过首先将指向节点的所有节点的指针改为指向新的节点，来替换或删除列表中的节点。 然后，通过从节点中的计数减去指向节点的访问计数的最大值来修改节点的访问计数。 只读访问节点继续增加访问计数，当达到零时，可以删除节点。 由于指向节点的指针已更改，只读进程可以继续，而不会暂停更改或替换事件。