JP4691895B2 - Data processing system design apparatus and computer program therefor - Google Patents

Description

  The present invention relates to a modular system design and execution method for constructing a data processing system by combining a plurality of modules on a computer, and in particular, by freely combining various modules in accordance with a predetermined purpose, The present invention relates to a system design apparatus and program capable of easily constructing a modular data processing system and easily operating to obtain a result.

  In the process of constructing a real-time speech recognition system or the like, it may be desired to confirm the performance such as the recognition rate by changing various settings for processing for speech recognition. For this purpose, for example, it is convenient not only to operate the same program by changing option values, but also to replace a certain program with a completely different program, or to experiment by freely changing the processing flow.

  One means for that purpose is a so-called modular system in which a plurality of modules each having a specific function are prepared in advance and can be combined in an appropriate manner according to the purpose.

  However, the conventional modular system is mainly employed in the design stage of a program or apparatus. In other words, the end product is manufactured by selecting existing modules according to the purpose and remodeling them according to the actual system conditions. The end user can freely combine various modules and assemble the system. No mechanism is provided to confirm the operation of such a system using test data.

  These problems are particularly important when a large amount of data needs to be processed as in speech recognition, and there are many varieties of data processing, but they occur not only in speech recognition but also in various data processing. Can do.

  SUMMARY OF THE INVENTION Therefore, an object of the present invention is to provide a data processing system design apparatus suitable for designing a modular system, in which modules can be freely combined and operation can be easily confirmed.

  A data processing system design apparatus according to a first aspect of the present invention is an apparatus for designing a data processing system by combining modules operable on a computer. This apparatus is based on module information storage means for storing module information including specifications of each module and option or parameter information, and specifications of each module and option or parameter information stored in the module information storage means. Assisting the user to define the connection relationship between multiple modules in the data processing system, and specifying the option or parameter value for each module in the data processing system, between the defined modules Flow definition generation means for generating flow definition information including connection relations and options or parameter values of each specified module, and flow definition information for storing flow definition information generated by the flow definition generation means Storage means.

  Preferably, the apparatus further includes a flow definition information selecting unit for causing the user to select any of the flow definition information generated by the flow definition generating unit, and the flow definition information selected by the flow definition information selecting unit. And an instance generation means for generating an instance of a module necessary for realizing a data processing system according to the flow definition information selected on the computer system according to the module information stored in the module information storage means; Data distribution means for distributing and processing data according to the selected flow definition information between the instances of the module generated by the instance generation means.

  More preferably, the apparatus further comprises means for distributing an option or parameter value for each module instance in the selected flow definition information to each module instance generated by the instance generation means. Including.

  The specification of a module may include a processing type of the module, information for specifying a program that realizes the module, and data type information for specifying a set of input data types and a set of output data types of the module. Good. The flow definition generation means includes a module specifying means for specifying a processing type module required in the data processing system according to a user's selection, and a connection between any two of the modules specified by the module specifying means. Data type belonging to the intersection set of connection specifying means for defining in accordance with the instructions of the above, a set of data types of one input of two modules connected to each other by the connection specifying means, and a set of data types of the other output A data type specifying means for causing the user to select one of the data types, specifying the type of data to be moved between the modules represented by the connection, and adding the attribute to the connection attribute specified by the connection specifying means; Based on the information about the module specified by the module specifying means and the connection specified by the data type specifying means. There are may include means for generating a flow definition.

  When executed by a computer, the computer program according to the second aspect of the present invention causes the computer to operate as one of the data processing systems described above.

−Configuration−
Hereinafter, a system according to an embodiment of the present invention will be described. This system selects the necessary modules from among the available program modules and connects the modules with lines (arrows) on the web browser screen according to the data flow between the modules. It is a system for performing speech recognition that can design and operate the system.

  In this system, for the purpose of processing the audio in real time, the audio data is divided into units of a predetermined length (referred to as frames) at a time interval of about 10 msec, and each module is divided between each module. Execute processing while passing data. Data in units of frames is input to a certain module, and processing for the frame is performed by the module, and then passed to the next module. This frame-unit data is referred to as “event message (hereinafter abbreviated as event)” in this specification.

  There are several types of events depending on the type of data. For example, in the case of raw audio data, it is WAVE type, and in the case of feature parameters, it is FEATURE type. Therefore, in the input / output of a module, it is necessary to clearly define the type of data input to or output from the module. In this specification, an event type determined according to a data type is referred to as an “event type”.

  When building a system, it is necessary to select a module according to the purpose and determine which output of which module is to be given to which input of which module in consideration of the data type. In this specification, the description of the data flow between modules is called a flow definition, and is typically stored in a storage device in a computer-readable format. In this system, the module definition is stored in the server, the module definition is read by the web browser, and the system design is performed on the web browser. What is obtained by the system design is data defining the connections between modules and the accompanying option / parameter values and the data flow. In this specification, this data is referred to as a flow definition, and the data obtained as a file is referred to as a flow definition file.

  FIG. 1 shows a functional block diagram of a speech recognition data processing system 50 according to an embodiment of the present invention. Referring to FIG. 1, the speech recognition data processing system 50 stores a module definition file 82 as described above in advance, and a flow for controlling the maintenance and execution of a flow definition by a user through a web server. Using the definition execution management device 60 and the module definition file 82 stored in the flow definition execution management device 60, a flow definition is created via a web browser, and the created flow definition is stored in the flow definition execution management device 60. And a flow definition creation / execution instruction device 62 for giving an instruction to cause the flow definition execution management device 60 to execute a system obtained by combining the modules according to the created flow definition. In the present embodiment, the flow definition execution management device 60 operates on the server device, and the flow definition creation / execution instruction device 62 operates on the client device, but details thereof are omitted in the drawings.

  The speech recognition data processing system 50 is composed of one or a plurality of computer systems, and a group of process programs 66 for causing each module to be executed on a computer process is stored in a storage device provided in these systems. It is remembered. The flow definition execution management device 60 uses these computer hardware and the process program 66 to execute data processing by the data processing system 64 composed of modules 100 to 110 designed by the flow definition creation / execution instruction device 62. Let Details of data processing and control by the flow definition execution management device 60 will be described later.

  As described above, the flow definition execution management device 60 uses the module definition file 82 that defines the specifications of modules that can be executed by the flow definition execution management device 60, and the flow definition creation / execution instruction device 62 using the module definition file 82. A flow definition file 86 that stores the flow definition defined by the above, an input file 90 that stores input data for processing, and an output file 88 that stores output results. Both the input file 90 and the output file 88 are selected by the user. In particular, the output file 88 is secured when data processing according to the flow definition is executed.

  The flow definition execution management device 60 further stores information such as the module definition file 82 and the flow definition between the input file information storage unit 84 that stores information on a plurality of types of input files and the flow definition creation / execution instruction device 62. When the management program 80 for exchanging via the web and the execution of the operation of the system 64 are instructed by the flow definition creation / execution instruction device 62, the management program 80 is activated, and the data processing system 64 is realized using computer resources. And a module manager 92 for controlling a process of giving an input from the input file 90 and processing the data, and outputting the result to the output file 88.

  In FIG. 1, the data processing system 64 includes a plurality of modules 100, 102, 104, 106, 108 and 110. In practice, the module definitions are classed, and the modules 100, 102, 104, 106, 108 and 110 are created for each module generated in the process that implements the data processing system 64 under the control of the module manager 92. An instance of an object class. Therefore, instances of a plurality of modules may be generated and operated from the same module definition in the same process or in different processes.

  FIG. 2 shows the contents of the module definition file 82. Referring to FIG. 2, the module definition file 82 includes an internal module name, a module name (for display) to be displayed on the screen of the flow definition creation / execution instruction device 62, and a flow definition creation / execution instruction device. 62 includes a module description (for display) composed of text for displaying a description of the module on the screen, an author name, and version information.

  The module definition file 82 further includes a module type for internally distinguishing the module type, and a module type (for display) used to display the module type on the screen of the flow definition creation / execution instruction device 62. And security restriction information including user information that can use the module. The module type indicates a basic function of the module. In the case of speech recognition as in the present embodiment, speech segment extraction (hereinafter referred to as “EPD”), feature extraction, CMS ( Normalization processing using an average cepstrum), feature parameter editing, decoder, reevaluation, result output, and the like.

  The module definition file 82 further includes a reception event type, an output event type, input / output event related information, and a module branch number as events that can be processed by this module. Each of the reception event type and the output event type is a set of event types possible in the module.

  The module definition file 82 further includes option information for this module. Similar options are grouped together in the same section. However, some options do not belong to a section. Accordingly, the module definition file 82 can include zero, one, or multiple sections. Options that do not belong to a section are described outside the section instead of within the section, but the way of description is the same as that of the option within the section described later.

  Each section includes an internal section name and a section name for display to the user. Each section further includes one or more options.

  For each option, an option name for internally distinguishing the option, a display option name for displaying the option to the user, and a help for this option in the flow definition creation / execution instruction device 62 are displayed. Option help (for display).

  Options also correspond to default values of option values when no value is specified by the user, valid value fields that list internal values that the option value can take, and internal values that the option value can take. And a display valid value field for enumerating the display values. Actually, the display text is stored in a separate file, and what is stored in the module definition file 82 is the identification number of the display text.

  The module definition file 82 further includes an effective range field that defines an effective range when the option value is a numerical value. The effective range field includes the minimum value and the maximum value, and two flags indicating whether or not the minimum value and the maximum value are included in the effective range, respectively.

  In addition to these, the module definition file 82 includes an option value type (any one of string, integer, double, boolean, file, and model), a multiple-specified count field indicating the number of times an option can be specified, and whether the option is required. If the option field is a type that is selected from a plurality of predetermined candidates, a GUI (graphical interface) used for the selection in the flow definition creation process performed by the flow definition creation / execution instruction device 62 is used. User interface) display type field indicating the type of part.

  FIG. 3 shows a specific example of the module definition file 82. As shown in FIG. 3, the module definition file 82 is described in XML (eXtended Markup Language). The part enclosed by the start tag <module> and end tag </ module> is the module definition. The part enclosed by the start tag <module-info> and end tag </ module-info> indicates information other than the options related to this module. The part between the start tag <options> and end tag </ options> indicates information about the entire option. The part enclosed by the start tag <section> and end tag </ section> shows information about the section, and the part enclosed by the start tag <option> and end tag </ option> shows information about each option. Show.

  The flow definition creation / execution instruction device 62 uses the module definition information about each module to generate screen information when the user creates a flow definition, imposes connection restrictions, Assist users when making settings.

  FIG. 4 shows details of the configuration of the flow definition file 86. The flow definition file 86 is an XML format file for storing a flow definition and option values created by a user, and one flow definition file 86 defines one data process. This flow definition file is selected and designated by the user via the flow definition creation / execution instruction device 62 when the module manager 92 shown in FIG. The module manager 92 reads the flow definition file 86, starts a process described in the flow definition file 86, generates an instance of the module described in the flow definition file 86, and assigns the flow definition file 86 to each module instance. After delivering and setting the option values described in, start the data processing system. FIG. 5 shows a specific example of the flow definition file 86. Hereinafter, FIG. 4 will be described, but please refer to FIG. 5 as appropriate.

  Referring to FIG. 4, the flow definition file 86 includes information regarding a process that forms one unit for executing data processing on a computer. Information about processes is repeated for the number of processes included in the data processing system. The information about the process includes a process name, a host name that executes the process, a program name for the process, and information about a module to be executed in the process.

  Information about modules is repeated for the number of modules included in one process. Information about each module includes a module name, a module name with a branch number, a module identifier, and a branch number. Since a plurality of the same modules may operate in one process, branch numbers are provided to distinguish them.

  The flow definition file 86 further includes information regarding the file. The information regarding the file includes information regarding the input file of the data processing system and information regarding the output file. Information about each file includes an input / output name, a file name, a format, an event type, a data type, and a size. These are repeated as many as there are multiple files.

  The flow definition file 86 further includes information regarding arrows indicating connection relationships between modules in the system. Information about the arrows is repeated as many times as there are arrows in the data processing system.

  The information regarding the arrow includes information regarding the origin of the arrow, information regarding the destination of the arrow, and information regarding the event indicated by the arrow. The information about the source includes the module name of the source of the arrow and a number indicating from which output of the module the arrow is. Similarly, the information regarding the destination includes a module name of the destination of the arrow and a number indicating which number of the destination module the destination of the arrow is. The information related to the event includes an event type indicating the data type of the event.

  The flow definition file 86 further includes information regarding the options of each module. Information about module options is repeated for the number of modules.

  The information about the module option includes information about the section if there is an option grouped in the section, and the information about each section includes information about the value of the option belonging to the section. Information about module options also includes information about option values that are not grouped into sections.

  In the present embodiment, regarding the information about modules with branch numbers, the amount of data is reduced by describing only the difference from the option value of the same module but without branch numbers. ing.

  FIG. 6 shows a block diagram of the flow definition creation / execution instruction device 62. Referring to FIG. 6, the flow definition creation / execution instruction device 62 creates a flow definition by GUI based on the contents of the module definition file 82 (see FIG. 1) received from the flow definition execution management device 60. In addition, the flow definition creation / execution instruction program 140 for specifying the created flow definition and instructing the flow definition execution management apparatus 60 to execute data processing, and the contents of the module definition file 82 are extracted. A module type table 146 for storing information on the module type, a module table 148 for storing the contents of the module definition file 82 received from the flow definition execution management device 60, and a flow definition creation / execution instruction device 62. A list of hosts belonging to the connected computer system And a host table 150 for storing. Details of these tables will be described later.

  The flow definition creation / execution instruction device 62 further includes a monitor 142 for realizing an interface with the user, an input device 144 including a keyboard and a pointing device, and a screen displayed by the user through a GUI using the monitor 142 and the input device 144. It includes a node table 152 and an arc table 154 for storing information related to the flow definition graph created above. Each module corresponds to a node in the flow definition graph, and an arrow corresponds to an arc. The arrows are hereinafter referred to as “arcs”. Details of these tables will also be described later.

  With reference to FIG. 7, the flow definition screen and its method by the flow definition creation / execution instruction program 140 will be described. Referring to FIG. 7, a flow definition screen 170 displayed on the monitor 142 by the flow definition creation / execution instruction program 140 includes a flow definition creation area 178 and a new button 180 that is operated when creating a new flow definition. , A save button 182 that is operated when saving a created flow definition, a delete button 184 that is operated when deleting a created flow definition, and a search that matches a predetermined condition from the created flow definitions A search button 186 that is operated, and an execution button 188 that is operated when instructing execution of the created flow definition.

  The screen 170 further includes a module palette 172 that displays, as icons, module types that can be used during flow definition, and an arc button 176 that is operated when creating an arc.

  Creation of a flow definition using the screen 170 is performed in the following procedure. First, as indicated by the mouse pointer 174, the icon of the module to be used is arranged by dragging and dropping from the module palette 172 onto the flow definition creation area 178. Next, for each module, determine which of the available modules to use and set options. Furthermore, an arc is arranged between the modules, and the input / output relationship between the modules is defined. Finally, set the event type for each arc.

  For example, when a module rectangle arranged on the screen is selected (clicked with the mouse), a selection screen for modules usable as the module of that module type is displayed. When a module is selected, a screen for setting options of the module is displayed. When an option is set, a check is performed according to option information described in the module definition file 82 of the module, and if there is an error, an error message is displayed and re-input is prompted.

  When an arc is selected (arc is double-clicked with a mouse), a screen for selecting the event type of the arc is displayed. The event types displayed here are common to the set consisting of those listed as the output event type of the module of origin of the arc and the set consisting of those listed as the input event type of the module to which the arc is destined. It is an element of a set (product set). When the event type of the arc is uniquely determined from the combination of modules, the event type is automatically set.

  For such an operation, information of the module type table 146, the module table 148, and the host table 150 is used, and information regarding the created flow definition is held in the node table 152 and the arc table 154.

  The operation using the screen 170 is essentially the same as that performed by, for example, an application creation tool that creates an application program skeleton by placing objects on the screen.

  FIG. 8 shows the configuration of the module type table 146. Referring to FIG. 8, each entry in module type table 146 includes a module type and a display character string for displaying the module type on screen 170. These pieces of information can be extracted from the module definition file 82 given from the flow definition execution management device 60.

  FIG. 9 shows the configuration of the module table 148. Referring to FIG. 9, each entry of module table 148 includes a module identification number (ID), a module type of the module, and a full path to a program of a process executing the module. Information of this module table 148 can also be extracted from the module definition file 82.

  FIG. 10 shows the configuration of the host table 150. As shown in FIG. 10, the host table 150 lists host names that can be used in the system. In this embodiment, the contents of the host table 150 are managed by the flow definition execution management device 60 and notified to the flow definition creation / execution instruction device 62.

  FIG. 11 shows the configuration of the node table 152. Referring to FIG. 11, each entry in the node table 152 executes a node ID, a module type constituting the node, an upper left coordinate and a lower right coordinate of a screen display rectangle of the module, and this module. It includes the full path name of the program, the execution module ID to be executed, and the execution machine name indicating the host on which this module is executed.

  FIG. 12 shows the configuration of the arc table 154. Referring to FIG. 12, each entry of arc table 154 includes a start node ID, an end node ID, and an event type of the arc.

  In the creation of a new flow definition, initially, the node table 152 in FIG. 11 and the arc table 154 in FIG. 12 are empty, but modules and arcs are arranged on the flow definition creation area 178 shown in FIG. As the event types of arcs and arcs are determined, the contents of these tables are specified little by little. When the flow is finally completed, the contents of the node table 152 and the arc table 154 are determined, and the contents of the flow definition file can be created based on them.

  In the modification of the flow definition, an arbitrary flow definition is received from the flow definition execution management device 60 shown in FIG. 1, and the contents of the node table 152 and the arc table 154 are created based on them. The subsequent operation is the same as that for creating a new flow definition.

  An example of a graph between modules defined by the flow definition is shown in FIG. These correspond to the data processing system 64 shown in FIG. A data processing system 200 shown in FIG. 13 includes a first process 222 and a second process 224. The first process 222 includes a first module 230, a second module 232, and a third module 234. The second process 224 includes a fourth module 236, a fifth module 238, and a sixth module 240.

  For example, the overall process speed can be improved by executing the first process 222 on one computer and executing the second process 224 on another computer. Of course, there is no problem even if the first process 222 and the second process 224 are operated on the same computer. However, there is a problem of how to configure these systems based on the flow definition to operate efficiently. In the present embodiment, in order to configure such a data processing system 200 based on the flow definition and operate efficiently, a module execution mechanism as described below is used.

  FIG. 14 is a block diagram of a data processing system 250 realized by the flow definition execution management apparatus 60 according to this embodiment. This system 250 is implemented according to, for example, one of the flow definition files 86 shown in FIG. 1, and includes a first process 252 and a second process 254 as in the system 200 shown in FIG. In this example, the flow definition file 86 defines the flow of data between modules constituting the first process 252 and the second process 254, respectively.

  The system 250 is further connected to the first process 252 and the second process 254 and to the outside of the system 250, between the modules in the first process 252 and the second process 254, and between these modules and the book. A module manager 92 for controlling the flow of events to and from the outside of the system 250 according to the description of the flow definition file 86.

  The first process 252 includes a first module 260, a second module 262, a third module 264, between these three modules 260, 262 and 264, and between these modules and the module manager 92. A message manager 266 for controlling the flow of events according to the description of the flow definition file 86 and managing message communication within the process.

  As with the first process 252, the second process 254 is similar to the fourth module 270, the fifth module 272, the sixth module 274, and between these three modules 270, 272 and 274, and these modules. And a message manager 276 for controlling the flow of events between the module manager 92 and the module manager 92 according to the description of the flow definition file 86.

  Each of the modules 260 to 274 has one or more inputs and outputs. Each module has a function of internally processing input data according to a function specific to the module and outputting a result. For example, the voice feature extraction module receives voice data as an input and outputs its feature parameters.

  The module manager 92 includes an input module 282 for receiving an input given from outside the system 250, an output module 284 for giving an output from the system 250 to the outside, an input module 282, an output module 284, a message manager. 266 and message manager 276, and a message manager 280 for controlling the flow of events between each process 252 and 254 and the input module 282 and output module 284 based on the description of the flow definition file 86, 1 includes a start / end control unit 286 for receiving a start instruction of the data processing system 250 from the module manager 92 shown in FIG. 1 and executing processing necessary for the execution of the data processing system 250, start processing, and end processing. The message manager 280 can be said to manage message communication between processes by controlling the flow of events.

  FIG. 15 shows the configuration of the event 320 communicated between the modules. Referring to FIG. 15, an event 320 includes a transmission source module ID, a transmission destination module ID, an event type, and data contents. The event type is a type of data included in the event, and the above-mentioned WAVE type, FEATURE type, and the like are examples.

  FIG. 16 shows the configuration of the message manager 280 in the form of a block diagram. Referring to FIG. 16, message manager 280 determines the module of the transmission destination module using the event transmission source module ID and the event type as a key in order to determine the transmission destination module of the event input to message manager 280. Responding to a flow management table 292 having a configuration capable of searching, a module management table 294 having a configuration capable of searching which module is included in which process, and a start instruction given from the start / end control unit 286, A table creation processing unit 296 for creating a flow management table 292 and a module management table 294 based on the description of the flow definition file 86 is included.

  The message manager 280 is further connected to the first process 252, the second process 254, the input module 282 and the output module 284, and the event given from the first process 252, the second process 254 and the input module 282 is A message distribution processing unit 290 for outputting to any of the first process 252, the second process 254, and the output module 284 based on the descriptions of the flow management table 292 and the module management table 294 is included.

  FIG. 17 shows an example of the configuration of the flow management table 292, and FIG. 18 shows an example of the configuration of the module management table 294. Referring to FIG. 17, the flow management table 292 includes a plurality of entries, and each entry includes a transmission source module ID, an event type, and a transmission destination module ID. These correspond to descriptions of arcs in the flow definition file 86 shown in FIG. By searching the flow management table 292 using the transmission source module ID and event type (see FIG. 15) included in the received event as keys, it is determined which module the event should be transmitted to. This can be determined by looking at the column. Note that the flow management table 292 shown in FIG. 17 shows only a part of the information necessary for the configuration shown in FIG.

  Referring to FIG. 18, module management table 294 also includes a plurality of entries. Each entry corresponds to a module and includes the process name to which the module belongs and the module ID of the module. The module management table 294 shown in FIG. 18 also shows only a part of information necessary for the configuration of the system 250 shown in FIG.

  Next, each part constituting the first process 252 and the second process 254 in FIG. 14 will be described. Since the configuration of each module is not directly related to the present invention, a detailed description thereof will be omitted. Since the message manager 266 and the message manager 276 have the same configuration, the message manager 266 will be described below.

  Referring to FIG. 19, the message manager 266 determines the module ID of the transmission destination module using the event transmission source module ID and the event type as a key in order to determine the transmission destination module of the event input to the message manager 266. Which module is included in the flow management table 302 of the same format as that shown in FIG. 16 and the process managed by the message manager 266 (the first process 252 in the present embodiment) having a configuration capable of searching. The module management table 300 describing whether or not, the flow management table creation processing unit 306 for creating the flow management table 302 by copying the flow management table 292 shown in FIG. 1 process 252 And a module management table creation processing section 304 for creating a module management table 300 including Yuru ID.

  The message manager 266 further receives an event from the first to third modules 260, 262 and 264 and the module manager 92, and refers to the flow management table 302 and the module management table 300, and identifies the event as the first to first modules. Message distribution processing unit 308 for determining which of the three modules 260, 262 and 264 and module manager 92 should be transmitted and distributing the event according to the determination result, and message distribution processing by the message distribution processing unit 308 An event queue 310 for temporarily storing an event determined to be transmitted to a process managed by the unit 308 (first process 252 in this embodiment) in the FIFO format, and one event from the event queue 310 Read one by one, Based on the destination module ID included in the cement, and a process in the message delivery unit 312 for delivering the event to one of the first to third module 260, 262 and 264.

  FIG. 20 schematically shows the format of the module management table 300. In the case of the first process 252 managed by the message manager 266, the modules included therein are the first module 260, the second module 262, and the third module 264. Accordingly, the contents of the module management table 300 in this case are as shown in FIG.

  FIG. 21 shows a flowchart of control processing of the data processing system 250 by the module manager 92. Referring to FIG. 21, when the start / end control unit 286 of the module manager 92 receives an execution instruction accompanying the specification of the flow definition file 86 from the flow definition creation / execution instructing device 62, in step 340, the specified flow Read the definition file. Then, the processes in steps 342 to 346 are repeated for all processes described in the flow definition file.

  That is, in step 342, the process described in the flow definition file is activated on the host described for the process in the flow definition file. Then, the processes in steps 344 and 346 are repeated for all modules described in the process definition.

  In step 344, an instance of the module described in the process definition is generated. In step 346, the option / parameter value described in the process definition for the same module is distributed to the generated module instance.

  By repeating the above processing for all processes and all modules, a process including each module constituting the data processing system 250 is started on the designated host.

  In step 348, instructions are given to the table creation processing unit 296 shown in FIG. 16 and the flow management table creation processing unit 306 shown in FIG. 19, and the flow management tables 292 and 302 are read in accordance with the contents of the flow management file read in step 340. Create In step 350, an instruction is given to the table creation processing unit 296 shown in FIG. 16 and the module management table creation processing unit 304 shown in FIG. 19, and the module management tables 294 and 300 are created according to the contents of the flow definition file. In step 352, a start instruction is given to the first module (input module if there is an input module), and the first event is generated and given to the input of this data processing system. Thereafter, the system operates by event distribution by the message manager 280 and the message managers 266 and 276 in each process, and data processed by each module in the system is output to an output file.

  In step 354, it is determined whether or not there is an unprocessed event in the data processing system 250. The process waits until there are no more events. At the timing when there are no unprocessed events, all programs started in step 342 are processed in step 356. End with, and end yourself.

  The outline of the start / end control processing of the data processing system 250 by the start / end control unit 286 of the module manager 92 has been described above.

  The message distribution processing unit 290 of the message manager 280 and the message distribution processing unit 308 of the message manager 266 are also realized by programs executed on computer hardware as will be described later. In the case of the present embodiment, different processing is performed depending on whether the transmission source of the event received by the message distribution processing units 290 and 308 is within the process or outside the process. The control structure of the computer program that implements each will be described with reference to FIGS. 22 and 23, particularly the message distribution processing unit 308. FIG.

  FIG. 22 is a flowchart of a processing program executed by the message distribution processing unit 308 of the message manager 266 shown in FIG. 19 when an event is received from a module in the process. The same applies to the message distribution processing unit 290 of the message manager 280 shown in FIG. Referring to FIG. 22, when an event is received from an in-process module, in step 430, the flow management table 302 is searched using the transmission source module ID and event type of the event as a key, and all corresponding arcs are extracted.

  Subsequently, the following processing is performed for all arcs found as a result of the processing in step 430. That is, in step 432, it is determined whether the destination module ID of the arc is that of the module in the process. If the destination module ID is that of the module in the process, the process proceeds to step 434. Otherwise, the process proceeds to step 440.

  In step 434, the destination module ID included in the searched arc is set as the event destination. Since the destination module is in the same process, this event is output to the event queue 310 in step 436.

  On the other hand, in step 440, the destination module ID included in the arc searched for is set as the event destination. Since the destination module exists outside the process, this event is output to the module manager 92 in step 442.

  By performing the processing in steps 434, 436, 440 and 442 for all arcs found in step 430, the number of events found for one event is generated, and the location of the destination module is determined. In response to the event queue 310 or the module manager 92, respectively.

  FIG. 23 is a flowchart of a processing program executed by the message distribution processing unit 290 of the message manager 280 shown in FIG. 16 when an event is received from outside the process. Referring to FIG. 23, when an event is received from outside the process, in step 460, the transmission destination module ID in the event is searched in the module management table 294, and the process including the transmission destination module is determined. In step 462, it is determined whether the determined process is itself. If it is the transmission destination, the event is stored in its own event queue in step 464. If it is not the transmission destination, in step 466, the event is transmitted to the process including the transmission destination module.

  On the other hand, when the message manager 266 shown in FIG. 19 receives an event from outside the process, the transmission destination of the event is guaranteed to be within the process. Therefore, the event is stored in the message distribution processing unit 308 shown in FIG.

-Operation-
The system shown in FIGS. 1 to 23 operates as follows. The operation is roughly divided into two aspects, that is, creation of a flow definition and processing for operating the data processing system 250 based on the created flow definition. First, the flow definition will be described.

  When the user starts the flow definition creation / execution instruction program 140 shown in FIG. 6, the contents of the usable module definition file and the list of usable hosts are transferred from the flow definition execution management device 60 to the flow definition creation / execution instruction device 62. Sent to. The flow definition creation / execution instruction device 62 creates a module type table 146, a module table 148, and a host table 150 according to the received information. A screen as shown in FIG. 7 is displayed on the screen. In the following, the operation of the flow definition creation / execution instruction device 62 will be described by taking as an example the case of newly creating a flow definition.

  Referring to FIG. 7, when the user presses new button 180 on screen 170, flow definition creation area 178 becomes blank. The module palette 172 displays usable module types by icons and module type names. When the user selects any icon on the module palette 172 and arranges it on the flow definition creation area 178 by drag and drop, an entry corresponding to the module is displayed in the node table 152 shown in FIG. Created. Each time a user places a new module, a new entry is created in the node table 152. Referring to FIG. 11, at this time, only the node ID, module type, and coordinates are set in each entry, and the other items are blank.

  The user further connects each module with an arrow (arc). Each time an arc is placed, a corresponding entry is created in the arc table 154. However, at this time point, only the start node ID and the end node ID among the information shown in FIG. 12 are stored, and the event type is undefined.

  Clicking on one of the module rectangles will display the module settings frame. In the module setting frame, a list of usable modules among the module types is displayed. This list can be created from the module table 148 shown in FIG. By selecting one of the displayed modules, it is determined which module will be used. It also specifies on which host this module should be run.

  Further, when a module to be used is determined, an option / parameter input frame that can be specified by the module is displayed at the lower left of the screen. When each option / parameter is set, a check according to the module definition information is performed. If there is an error, an error is displayed and a re-input is prompted.

  When the user double-clicks one of the arcs, a list box for specifying the event type is displayed. The list box lists elements of a common set (product set) of the set of output event types of the source module and the set of input event types of the destination module. By selecting one of them, the event type of the arc is determined, and information indicating the selected event type is stored in the event area of the arc table 154 shown in FIG. If there is only one type of event that can be used, the arc event type is automatically set.

  Thus, the usage module, host, and options / parameters for all modules are determined, and once the event type for each arc is determined, a flow definition can be created based on them. The flow definition is transferred from the flow definition creation / execution instruction device 62 to the flow definition execution management device 60 and stored in the flow definition execution management device 60 as one of the flow definition files 86.

  The modification of the flow definition is the same as in the case of creation except that one of the flow definition files 86 is first selected and the contents of the file are transferred to the flow definition creation / execution instruction device 62. is there.

  In order to operate the data processing system 64 in accordance with the created flow definition, the following operation is performed. Referring to FIG. 7, when an “execute” button 188 is pressed, a flow definition file selection dialog is displayed. By selecting one of them, the flow definition creation / execution instruction device 62 instructs the flow definition execution management device 60 to start data processing based on the flow file.

  The management program 80 shown in FIG. 1 gives information for specifying the selected flow definition file to the module manager 92 according to the instruction and instructs the start of data processing.

  Referring to FIG. 14, in response to this instruction, activation / termination control unit 286 reads the designated flow definition file (step 340 in FIG. 21), and designates each process of data processing system 250 as the designated host. It starts up (step 342). Further, for each module described in the flow definition file, an instance is generated in the designated process (step 344), and the options / parameters described in the flow definition file are distributed (step 346). Thereafter, an instruction is given to the table creation processing unit 296 shown in FIG. 16 and the flow management table creation processing unit 306 and the module management table creation processing unit 304 shown in FIG. 19, and the flow management tables 292 and 302 and the module management table 294 are given. And 300 are created (steps 348 and 350 in FIG. 21). Thus, the setup of the data processing system 250 is completed.

  Next, the start / end control unit 286 gives a start instruction to the input module shown in FIG. 14 so that data is read from the input file and given to the message manager 280 as an event (step 352 in FIG. 21). The input module 282 creates frame data from the input file and provides it to the message manager 280 as an event.

  Referring to FIG. 16, the message distribution processing unit 290 of the message manager 280 refers to the flow management table 292 and the module management table 294 to determine a process including the event transmission destination module (see FIG. 23). Step 460), sending the event to the process (step 464 or 466). In this example, it is assumed that this event is transmitted to the first process 252, for example.

  Referring to FIG. 19, message distribution processing unit 308 of message manager 266 stores the event in event queue 310 because the event is from outside the process. The in-process message delivery unit 312 reads the event from the event queue 310 and delivers the event to the module specified by the destination module ID.

  The module that has received this event returns the resulting event to the message distribution processing unit 308 of the message manager 266. At this time, a value is set for each of the event transmission source module ID and the event type by the module. No destination module ID is set.

  The message distribution processing unit 308 searches the flow management table 302 using the transmission source module ID and event type of the received event as keys (step 430 in FIG. 22), and obtains a matching arc. The message distribution processing unit 308 further refers to the module management table 300 to determine whether or not the transmission destination module is in the first process 252 (step 432). If it is in the first process 252 (that is, if the module ID exists in the module management table 300), the destination module ID included in the arc is set as the event destination module ID (step 434). ) The message distribution processing unit 308 writes this event to the event queue 310 (step 436).

  If the transmission destination is outside the first process 252, the transmission destination module ID included in the arc is set as the transmission destination module ID of the event (step 440), and this event is transmitted to the module manager 92 (step 442). ).

  The message manager 280 of the module manager 92 determines the destination process of this event based on the flow management table 292 and the module management table 294 as in the case of receiving the first input data, and sends the event to the message manager of that process. Send. When the flow management table 292 indicates that the message is to be output outside the system, the message distribution processing unit 290 outputs an event to the output module 284. The output module 284 further outputs this event to the outside.

  Thereafter, the message distribution processing unit 290 of the message manager 280 and the message managers 266 and 276 of the first process 252 and the second process 254 repeat the same processing as described above. As a result, data input to the input module 282 one after another from the outside is processed by each predetermined module, and the resulting event is output from the output module 284.

  In the system 250 described above, each module does not need to be aware of which module the event resulting from its processing is sent to. In any process, the message manager, not the module, decides the destination module, and if the destination module is in the same process, the event queue is used to deliver the event. Deliver the event. The module manager that has received the event distribution determines a transmission destination process based on the transmission destination module ID, and transmits it.

  Therefore, the independence of the module becomes extremely high, and it is not necessary to change the module according to the system configuration. Furthermore, regarding inter-process communication, all processes need only be transmitted to the module manager without regard to the destination process. Therefore, the overhead for communication between processes is reduced.

  As described above, according to the system of the present embodiment, a flow definition is created according to a module definition, a system composed of a plurality of modules is efficiently designed using the flow definition, and actually operated with a simple operation. be able to. It is easy to change the connection relationship between modules, or to replace the module to be used with another module of the same module type and compare the system performance.

  In addition, communication between each module designed and operated as described above can be managed by a message manager provided in each process and in the module manager. Each module is a module for inputting an event and a module for a destination of the event. There is no need to be conscious of such. Therefore, the independence of each module can be increased and the system can be freely assembled. Further, data is processed in each module while being transferred between processes, and each process can perform data processing in the correct order while adjusting the processing time by an event queue. As a result, a large amount of data processing such as voice recognition can be performed in real time. Further, since the modules can be freely combined, it is possible to easily perform performance experiments of systems having different configurations in a short time.

-Realization by computer-
The system of this embodiment is realized by computer hardware, a program executed by the computer hardware, and data stored in the computer hardware. FIG. 24 shows the external appearance of the computer system 530, and FIG. 25 shows the internal configuration of the computer system 530. As described above, this system can also be realized by a combination of a plurality of computers via a network or the like.

  Referring to FIG. 24, this computer system 530 includes a computer 540 having an FD (flexible disk) drive 552 and a CD-ROM (compact disk read only memory) drive 550, a keyboard 546, a mouse 548, a monitor 542, including.

  Referring to FIG. 25, in addition to the FD drive 552 and the CD-ROM drive 550, the computer 540 includes a CPU (Central Processing Unit) 556 and a bus 566 connected to the CPU 556, the FD drive 552, and the CD-ROM drive 550. And a read only memory (ROM) 558 that stores a boot-up program and the like, and a random access memory (RAM) 560 that is connected to the bus 566 and stores a program command, a system program, work data, and the like. Computer system 530 further includes a printer 544.

  Computer 540 further includes a network adapter board 568 that provides connection to a local area network (LAN).

  The computer system 530 includes a flow definition creation / execution instruction device 62, a flow definition execution management device 60, or a system 250 (or a part of the first process 252, the second process 254, and a module manager) according to this embodiment. 92, etc.) is stored in the CD-ROM 562 or FD 564 inserted into the CD-ROM drive 550 or the FD drive 552 and further transferred to the hard disk 554. Alternatively, the program may be transmitted to the computer 540 through a network and stored in the hard disk 554. The program is loaded into the RAM 560 when executed. The program may be loaded directly into the RAM 560 from the CD-ROM 562, the FD 564, or via a network.

  This program is stored in the computer 540 (or another computer that can communicate with the computer 540), the flow definition creation / execution instruction device 62, the flow definition execution management device 60, and each process 252 included in the system 250 according to this embodiment. 254, a module manager 92, and a plurality of instructions for performing operations as modules executed in each process. Some of the basic functions required for this are provided by operating system (OS) or third party programs running on the computer 540 or various toolkit modules installed on the computer 540. Therefore, this program does not necessarily include all functions necessary for realizing the system of this embodiment. This program need only include instructions that implement the above-described devices and systems by calling appropriate functions or “tools” in a controlled manner to obtain the desired result. The operation of computer system 530 is well known and will not be repeated here.

  In the above-described embodiment, the module management table is prepared for each process. However, the present invention is not limited to such an embodiment, and the same module management table may be used for all modules in the system. In the above embodiment, the module identifier is unique within the system. However, the present invention is not limited to such an embodiment, and the module identifier may be unique only within the process. However, in that case, in event communication, it is necessary to specify a transmission source and a transmission destination by a process name and a module identifier.

  In the above-described embodiment, it is assumed that the flow definition creation / execution instruction device 62 is realized by a program operating on a browser program. However, the present invention is not limited to such an embodiment, and the flow definition creation / execution instruction device 62 may be implemented by a unique application program different from the browser. Also, the flow definition execution management device 60 has all the substantial parts of the function as the flow definition creation / execution instruction device 62. The flow definition creation / execution instruction device 62 is used for display on the browser program and for flow definition. Only the operation may be performed.

  Furthermore, the flow definition file 86, the module definition file 82, the input file 90, etc. need not be limited to those existing on the same computer as the flow definition execution management device 60, and other types that can be accessed via the network. What exists in the system may be designed as an available system.

  The embodiment disclosed herein is merely an example, and the present invention is not limited to the above-described embodiment. The scope of the present invention is indicated by each claim in the claims after taking into account the description of the detailed description of the invention, and all modifications within the meaning and scope equivalent to the wording described therein are intended. Including.

1 is a block diagram of a speech recognition data processing system 50 according to an embodiment of the present invention. It is a figure which shows the structure of a module definition file. It is a figure which shows the example of a module definition file. It is a figure which shows the structure of a flow definition file It is a figure which shows the example of a flow definition file. 4 is a block diagram of a flow definition creation / execution instruction device 62. FIG. 6 is a schematic diagram of a display screen of a flow definition creation / execution instruction device 62. FIG. 5 is a schematic diagram showing a configuration of a module type table 146. FIG. 4 is a schematic diagram illustrating a configuration of a module table 148. FIG. 3 is a schematic diagram illustrating a configuration of a host table 150. FIG. 3 is a schematic diagram showing a configuration of a node table 152. FIG. 3 is a schematic diagram showing a configuration of an arc table 154. FIG. It is a figure which shows the structural example of the system which consists of a some process each containing several modules. FIG. 3 is a block diagram of a system 250 according to an embodiment of the present invention. 6 is a diagram showing a configuration of an event 320. FIG. It is a block diagram of the message manager 280 of the module manager 92 shown in FIG. It is a figure which shows the structure of the flow management table 292. It is a figure which shows the structure of the module management table 294. FIG. 15 is a block diagram of the message manager 266 of the first process 252 shown in FIG. 3 is a diagram illustrating a configuration of a module management table 300. FIG. It is a flowchart which shows the control structure of the process which the starting / ending control part 286 performs. It is a flowchart which shows the process performed by each message distribution process part when the event in a process is received. It is a flowchart which shows the process performed when the message distribution process part 290 of the module manager 92 receives an out-of-process event. It is an external view of the computer system for implement | achieving the process of the system 250 which concerns on one embodiment of this invention, or its one part. It is a block diagram which shows the internal structure of the computer system shown in FIG.

Explanation of symbols

  50 speech recognition data processing system, 60 flow definition execution management device, 62 flow definition creation / execution instruction device, 64, 200, 250 data processing system, 66 process program, 80 management program, 82 module definition file, 84 input file information 86 Flow definition file, 92 Module manager, 140 Flow definition creation / execution instruction program, 146 Module type table, 148 Module table, 150 Host table, 152 Node table, 154 Arc table, 172 Module palette, 178 Flow definition creation area, 222,252 First process, 224,254 Second process, 230,232,234,236,238,240,260,262,264,270,2 72,274 Module, 266, 276, 280 Message Manager, 286 Start / End Control Unit

Claims (5)

A data processing system design device for designing a data processing system by combining modules operable on a computer,
Module information storage means for storing module information including specifications of each module and option or parameter information;
Based on the specifications of each module and the option or parameter information stored in the module information storage means, a user defines a connection relationship between a plurality of modules in the data processing system, and Helps the user specify options or parameter values for each module, supports the definition of one or more processes, the connection relationship between the inputs and outputs of the defined modules, and the options or Flow definition generation means for generating flow definition information including parameter values,
Each of the one or more process definitions includes a host name that executes the process, a program name that realizes the process, and information about a module to be executed in the process,
Furthermore, the data processing system design apparatus contains the flow definition information storage means for memorize | stored the flow definition information produced | generated by the said flow definition production | generation means. further,
A flow definition information selection function Ru allowed a user to select any of the flow definition information generated by the flow definition generating unit,
To realize a data processing system according to the selected flow definition information on a computer system according to the flow definition information selected by the flow definition information selection function and the module information stored in the module information storage means. and Louis instance generation function to generate an instance of the required modules,
Between instances of modules generated by the instantiation function, within the selected flow definition information, and a Lud over data distribution function to distribute the data according to the connection relationship between the modules, according to claim 1 Data processing system design equipment. Furthermore, an instance of each module generated by the instantiation function, within the selected flow definition information has the ability to deliver the option or parameter value for the instances of each module, in claim 2 The data processing system design device described. The specification of the module includes a processing type of the module, information for specifying a program that realizes the module, and data type information for specifying a set of input data types and a set of output data types of the module. ,
The flow definition generation means includes
And makes the chromophore at the distal end Joule specific function be identified according to the selection of the user module processing type required in the data processing system,
A connection specific functions that define in accordance with an instruction from the user a connection between any between the two inputs and outputs of the identified modules by the module specifying unit,
Let the user select one of the data types belonging to the intersection of the set of data types of one input of the two modules connected to each other by the connection specifying function and the set of data types of the other output; and added to Lud over data type specific functions to the attribute of the identified connected by the connection specific function to identify the type of data being transferred between modules, represented by the connection,
Information about the module identified by the module specific function, based on the connection identified by the data types specific function has a function and that generates the flow definition, any one of claims 1 to 3 The data processing system design device described in 1. A computer program for operating a computer as a data processing system design device for designing a data processing system by combining modules operable on the computer,
The computer
Module information storage means for storing module information including specifications of each module and option or parameter information;
Flow definition information storage for storing flow definition information including one or more process definitions, connection relationships between defined module inputs and outputs, and options or parameter values for each specified module Means,
Each of the one or more process definitions includes a host name that executes the process, a program name that realizes the process, and information about a module to be executed in the process,
The computer program stores the computer,
Based on the specifications of each module and the option or parameter information stored in the module information storage means, a user defines a connection relationship between a plurality of modules in the data processing system, and assist in specifying user the option or parameter values for each module, the generated flow definition information the a flow definition generating means for storing the flow definition information storage means to function by a computer program.

Images