WO2022264434A1

WO2022264434A1 - Program code automatic generation system

Info

Publication number: WO2022264434A1
Application number: PCT/JP2021/023803
Authority: WO
Inventors: 基光白川
Original assignee: ソプラ株式会社
Priority date: 2021-06-15
Filing date: 2021-06-23
Publication date: 2022-12-22
Also published as: JP6958884B1; JP2022190776A

Abstract

[Problem] To fundamentally change a mechanism of developing programs to prevent an immense impact on society caused by a chronic shortage of system engineers in a near future. [Solution] The present invention includes: an accepting means that accepts input of text data through voice or an input terminal; an extracting means that subjects the text data to natural language analysis to extract an instructing operation and a case component consisting of object cases; a storing means that stores an execution operation table and an object case table; a searching means that searches an execution operation instruction corresponding to the instructing operation and an execution operation target corresponding to the object case by referring to the execution operation table and the object case table; and a code generating means that substitutes the execution operation instruction and the execution operation target searched by the searching means for basic syntax of program code to thereby generate program code for executing execution operation in which the execution operation target is prescribed for the execution operation instruction.

Description

Program code automatic generation system

The present invention relates to an automatic program code generation system that fundamentally changes the mechanism of program development.

In recent years, technology related to RPA (Robotic Process Automation) has attracted attention as a means of assisting tasks such as desk work (routine work). As such a technique, for example, an information processing apparatus disclosed in Patent Document 1 has been proposed.

The information processing apparatus of Patent Document 1 includes an acquisition unit that acquires user message information, an extraction unit that extracts task information related to the user from the acquired message information, and a task from the extracted task information. an identification unit that identifies a person in charge and a related person; a classification unit that classifies the task information into first task information in which the user is the person in charge or second task information in which the user is the related person; A registration unit for registering the classified first or second task information in a database, and an output unit for referring to the database and outputting the first or second task information.

JP 2019-16280 A

By the way, in order to automatically execute a new task by a program, it is necessary to create a program. Conventionally, the process of creating a program requires a process of defining the requirements for the new business, designing the system, coding the program code, and then testing and verifying it. Such program code was usually manually coded each time a new task was created.

However, with the rapid progress of IT in recent years, the demand for programming has continued to increase, and the shortage of IT engineers has become a serious social problem.

Also, depending on the situation, there are cases where even simple internal work must be changed at any time. For example, even if you create a program to automatically perform an internal task such as "Send a daily report to the PC of employee XX", if there is even the slightest change in the task, you have to recreate the program code. Gone.

Although there have been such problems in the past, a more serious problem is that in the near future, it is predicted that the chronic shortage of SE engineers will have a huge impact on society.

SUMMARY OF THE INVENTION Accordingly, the present invention has been devised in view of the above-described problems, and its object is to automatically execute business in a computer system, and to provide a method for executing the business. When generating program code, it is extremely easy, and even if you do not have the technical skills of an SE, you can easily give instructions to the person in charge who is familiar with the work, and intervene with the help of people with SE technology. It is possible to generate program code automatically without any need, fundamentally change the mechanism of program development, and in the near future, program code automation that can avoid the serious impact of the chronic shortage of SE engineers on society It is to provide a generation system.

A program code automatic generation system according to the present invention comprises a receiving means for receiving input of text data via voice or an input terminal; an extracting means for extracting case components consisting of at least objective cases that depend on, an execution action table in which actual execution action commands for referential actions are associated, and actual execution action objects for reference cases are associated. storage means for storing at least an object case table; an execution action instruction corresponding to the extracted pointing action; and an execution action object corresponding to the extracted object case are stored by the storage means. A search means for searching by referring to an execution action table and an object case table, and by substituting the execution action object and the execution action object searched by the search means into the basic syntax of the program code, the execution action object is determined. and code generation means for automatically generating a program code for executing the execution operation specified in the execution operation instruction for the execution operation destination.

According to the above-described invention, in order to automatically execute business in a computer-side system, it is extremely easy to generate program code for executing the business, and moreover, the technical capabilities as an SE are extremely easy. Even without it, if the person in charge of the work is familiar with the work, it is possible to easily give instructions and automatically generate the program code without human intervention with SE technology. will fundamentally change, and in the near future, it will be possible to avoid a chronic shortage of SE engineers having a huge impact on society.

FIG. 1 is a block diagram of a program code automatic generation system according to Embodiment 1. FIG. FIG. 2 is a flow chart of a program code automatic generation system to which the present invention is applied. FIG. 3 is a diagram showing an example of setting and registering an execution destination in advance for each company by a user who provides this system. FIG. 4 is a diagram showing an example of an execution action table. 5 is a diagram showing another example of the execution operation table shown in FIG. 4. FIG. FIG. 6 is a diagram showing an example of an object case table. FIG. 7 is a diagram showing another example of the object case table shown in FIG. FIG. 8 is a diagram showing an example of a final grade table. FIG. 9 is a diagram showing another example of the end grade table shown in FIG. FIG. 10 is a diagram for explaining a case where machine learning by artificial intelligence is used to search for the basic syntax of program code. FIG. 11 is a diagram for explaining a method of generating a confirmation sentence for confirming input contents to the user. FIG. 12 is a diagram illustrating an example of an execution operation table according to the second or fourth embodiment; FIG. 13 is a diagram showing an example of an object case table according to the third or fourth embodiment. FIG. 14 is a diagram showing another example of the end grade table.

Hereinafter, the program code automatic generation system according to Embodiments 1 to 4 of the present invention will be exemplified and explained with reference to the drawings.

[Embodiment 1]
(Program code automatic generation system 100)
An example of the configuration of a program code automatic generation system 100 according to this embodiment will be described with reference to FIG. FIG. 1 is a schematic diagram showing the overall configuration of a program code automatic generation system 100 according to this embodiment.

The automatic program code generation system 100 is not just a work assistance, but a fundamental change in the mechanism of program development, and in the near future, it will be possible to avoid a chronic shortage of SE engineers having a huge impact on society.

That is, the program code automatic generation system 100 automatically generates a program code for executing a business, thereby performing each business in the company (e.g., mail distribution, worker progress collection, file management, file transfer, etc.). etc.) can be automatically performed on a computer. The automatic program code generation system 100 can set the automatic generation of the program code particularly by voice or text. It is also possible for users who use the software to manage business operations, etc., to easily realize automatic generation of program codes for causing computers to automatically perform various business processing.

For example, as shown in FIG. 1, the automatic program code generation system 100 includes an automatic program code generation device 1, and the user may use the automatic program code generation device 1. The automatic program code generation system 100 may include, for example, a terminal 2 connected to the automatic program code generation device 1 via a communication network 4 , and the user may use the automatic program code generation device 1 via the terminal 2 . The automatic program code generation system 100 includes, for example, a server 3 connected to the automatic program code generation device 1 via a communication network 4, and a user can exchange various information with the server 3 via the automatic program code generation device 1 or the terminal 2. Each means may be realized by transmission and reception of .

<Program code automatic generation device 1>
Known electronic devices such as servers, personal computers, smart phones, tablet terminals, and wearable terminals are used as the program code automatic generation device 1 . In addition, in the program code automatic generation system 100 in this embodiment, the program code automatic generation device 1 serves as the core of the cloud server.

Although not shown, the program code automatic generation device 1 includes an acquisition unit and a calculation unit, and may include, for example, an execution unit, a storage unit, an output unit, and an intent storage unit. Each function of the automatic program code generation device 1 is implemented by the CPU using the RAM as a work area and executing a program stored in a storage unit or the like. Also, part of each function may be controlled by artificial intelligence. Here, "artificial intelligence" may be based on any known artificial intelligence technology.

The acquisition unit accepts input of text data via voice or an input terminal. The acquisition unit acquires text data input by the user via the terminal 2 or the input unit of the automatic program code generation device 1, for example. For example, when a conversational sentence is input by voice from the user via the terminal 2 or the input unit of the automatic program code generation device 1, the acquisition unit uses a known voice recognition technology (for example, voice recognition technology) to generate Get the text data that has been specified. Note that cloud-type speech recognition technology may be used as the speech recognition technology via the communication network 4, for example.

The calculation unit refers to the database and executes various processing operations and calculations based on the acquired text data. The calculation unit performs natural language analysis on the received conversational sentence to extract individual constituent elements of the sentence, including verbs, nouns, case components, and the like. The calculation unit performs a processing operation of allocating an execution action instruction corresponding to the instruction action extracted from the received text data, an execution action target corresponding to the object case, and an execution action destination corresponding to the end case. This computing unit performs a processing operation to determine an intent corresponding to the received text data. The calculation unit refers to the storage unit and extracts the basic syntax of the program code. Further, the arithmetic unit generates a program code by substituting the assigned execution operation instruction, execution operation target, and execution operation destination for the basic syntax of the extracted program code.

The execution unit executes business processing based on the program code generated by the calculation unit. Examples of business processing include routine work such as sending emails to the person in charge based on the content and deadline of the task, managing work hours, and updating the task progress history. be done.

The storage unit temporarily stores text data consisting of conversational sentences acquired via the acquisition unit. The text data stored in this storage unit may be read and updated under the control of the calculation unit, execution unit, and the like. The storage unit holds at least three tables: an execution action table, an object case table, and an end case table. Also, the storage unit may hold a code table.

The output unit outputs various information related to the operations executed by the program code. The display data is notified so that the user can recognize it via the notification unit of the program code automatic generation device 1, the terminal 2, or the like. The output unit outputs display data and the like to the terminal 2 and the like via the I/F, and outputs display data and the like to the notification unit via the I/F.

　One or more intents are stored in the intent storage unit. The intent may be stored in the intent storage section in association with information specifying the business process. The information specifying the business process is usually an action name, which will be described later, but the format is not limited to this. Corresponding also includes, for example, the case where an intent has information specifying business processing.

<Terminal 2>
As the terminal 2, for example, a known electronic device such as a server, personal computer, smart phone, tablet terminal, or wearable terminal is used. The terminal 2 may have at least part of the same configuration and functions as the program code automatic generation device 1, for example. For example, a plurality of terminals 2 may be provided, and each terminal 2 may be connected to the program code automatic generation device 1 via a communication network 4 .

<Server 3>
Various information is stored in the server 3, for example. The server 3 accumulates various kinds of information sent from the program code automatic generation device 1 or the like via the communication network 4, for example. The server 3 may store, for example, the same information as the storage unit of the automatic program code generation device 1 , and transmit/receive various information to/from the automatic program code generation device 1 and the like via the communication network 4 . That is, in the automatic program code generation system 100 , the server 3 may be used instead of the automatic program code generation device 1 or the storage unit of the automatic program code generation device 1 .

<Communication network 4>
The communication network 4 is an Internet network or the like to which the program code automatic generation device 1 is connected via a communication circuit. The communication network 4 may be composed of a so-called optical fiber communication network. Also, the communication network 4 may be realized by a known communication network such as a wireless communication network, in addition to the wired communication network.

Next, the operation of the program code automatic generation system 100 to which the present invention is applied will be explained.

As shown in FIG. 2, the dialogue is accepted in step S11. Specifically, when a conversational sentence is acquired as voice data by the acquisition unit of the automatic program code generation device 1, text data is generated using a known voice recognition technology (for example, phoneme recognition technology). Further, when text data converted into electronic data is obtained by the obtaining unit of the automatic program code generation device 1, it is used as it is. The text data converted into electronic data may be, for example, e-mails, information bulletin boards used inside or outside the company, and postings on various sites as they are.

The acquisition unit of the automatic program code generation device 1 acquires text data at each timing of input by the user, and may acquire, for example, multiple pieces of text data accumulated over a certain period of time at once. A known technique can be used to generate text data from non-text data input via the terminal 2 or the input unit of the automatic program code generation device 1 . The text data acquired in this manner is temporarily stored in the storage unit of the automatic program code generation device 1 .

Next, the process proceeds to step S12, where the conversation sentence as text data acquired in S11 and temporarily stored in the storage unit of the automatic program code generation device 1 is read out and subjected to natural language analysis. This natural language analysis is mainly performed by the computing unit of the automatic program code generation device 1 . This natural language analysis may use any known natural language analysis technique, including morphological analysis and syntactic analysis.

As a result of the natural language analysis in step S12, for example, for the text data "Register the product name A5-7853K", an instruction consisting of "| It can be broken down into actions. In addition, the text ``Taro Yamada's daily report for this month to Tokugawa section manager'' has the object case of ``Taro Yamada | | That is, text data can be extracted through a combination of case components consisting of object cases and terminal cases that are dependent on verbs consisting of pointing actions.

A case component is a noun phrase required to form a sentence with a verb (indicative action). A case in case components is a function of selectively requesting a combination of nouns or noun phrases necessary for forming a sentence according to the lexical meaning that a verb has. The case grammar, which is the rationale for these case components, is based on the fact that a simple sentence consists of a deep case (objective case, conditional case, instrumental case, starting case, ending case, temporal case, etc.) and a verb connected to these cases. It is a theory that analyzes sentences as things that become. Then, the noun (phrase) requested by the verb as the deep case of the sentence can be said to be the necessary case component to supplement the realization of the proposition by the verb. A case component is not a mere noun phrase, but a noun phrase to which a verb is dependent in order to realize its own proposition. In other words, a case component is an element of nouns (including noun phrases) dependent on a verb that can serve as an action start condition for realizing a proposition by a verb that defines a referential action.

The object case here means the execution object of the proposition by the verb that defines the referent action. In the case of the phrase "transfer file A", "file A" is subject to the verb "transfer". In addition, the terminal case defines the place where the execution object defined by the object case is reflected, the recipient of the benefit of the execution object, and the destination to which the execution object is sent. In the case of the sentence ``Send File A to Taro Yamada'', the verb ``Send'' is subject to ``File A'' as an object case, and since this is what is sent with Taro Yamada as the end point, This final case is "Taro Yamada".

Next, the process proceeds to step S13 to determine the meaning and action of the natural language-analyzed conversational sentence. In this step S13, a semantic action consisting of an execution action instruction, an execution action target, an execution action destination, etc. is determined from the instruction action and the case component analyzed through the natural language analysis.

First, in step S12, the arithmetic unit of the automatic program code generation device 1 performs the natural language analysis of the conversational sentence, the execution action command corresponding to the extracted pointing action, the execution action object corresponding to the extracted object case, and the extracted The destination of the execution action corresponding to the given end case is searched by referring to the degrees of association of the execution action table, the object case table, and the end case table.

The execution operation command here is a word for commanding the execution operation in the actual program code, and corresponds to actual database registration, file writing, transfer, deletion, monitor display, etc.

The execution operation target indicates the object for executing this execution operation command, and may be invoices, daily reports, specification files, data, emails, etc., product names, various events, and information. (eg, working hours, inventory, etc.).

Various program codes can be completed by combining this execution operation target and execution operation instructions.

For example, if the execution command is "transfer" and the execution command is "specification file", it is possible to create the program code "transfer of specification file" by combining them.

The execution destination indicates the execution destination for executing the execution action instruction. For example, a user who installs this system may set and register an execution destination in advance for each company. For example, as shown in FIG. 3, "Taro Yamada's PC", "Manager Tokugawa's PC", "development department's shared PC", etc. may be registered as execution destinations in the company LAN. In addition, it is also possible to register, for example, "Person in charge B of company A", etc., through an external public communication network with respect to business partners outside the company, even though the in-house LAN is crossed.

That is, before implementing the present invention, an end-point case table is created. "Shared PC of the development department") may be registered in advance as each referencing execution destination of the terminal case table.

Next, the program moves to step S14 and analyzes the program code for executing the conversation sentence for which the semantic action has been determined. This program code association analysis operates to extract the basic syntax of the program code.

The specific method of steps S13 and S14 will be detailed later.

Next, move to step S15 to create a program code. In step S14, as described above, the basic syntax of the program code is merely extracted, and the target of the actual processing operation and the text that defines each condition necessary to complete the processing operation are substituted. The program code is complete. For this reason, in step S15, by substituting words defining an execution target, an execution target, and an execution action command into the basic syntax of the extracted program code, the execution target is assigned to the execution target. to generate a program code for executing the execution action specified in the execution action instruction.

In the process of steps S11 to S15, a program can be automatically generated based on the user's intention included in the conversation received in step S11.

After completing the program code in this way, it may be provided to the user, or may be displayed via the notification unit of the automatic program code generation device 1, or may be displayed by the execution unit of the automatic program code generation device 1. , the completed program code may be executed. That is, according to the present invention, it is possible to execute the automatically generated program code as it is. Therefore, when the process is included from step S11, when the user utters a conversation sentence, program code incorporating the intention can be automatically generated, and the generated program code can be executed as it is. be able to. When the user speaks the action he wants the system to perform in conversational sentences or inputs text data according to this, the action incorporating the intention is executed as it is based on the automatically generated program code. becomes possible.

Next, a detailed method of program code analysis for determining the semantic action of the conversational sentence subjected to natural language analysis in step S13 and executing the semantic action-determined conversational sentence in step S14 will be described.

In carrying out these methods, an execution action table in which actual execution action commands are associated with referential actions with three or more degrees of association, and three actual execution action objects with respect to reference objects are provided. Use the object case table associated with the degree of association of a grade or higher. Each of these tables is not limited to being arranged in a visually visible table or matrix, but is simply composed of learned data in which machine learning models of artificial intelligence are stored and their collections. may be

4 and 5 show examples of execution operation tables among these. In this execution motion table, it is premised that three or more degrees of association between reference instruction motions and execution motion commands are set in advance. Assume that the input data are reference instruction operations P01 to P03, for example. For example, the reference instruction action P01 is "send", the reference instruction action P02 is "up", and the reference instruction action P03 is "erase". Such reference instruction operations P01 to P03 as input data are linked to execution operation commands as outputs.

Such a referencing instruction action has three or more levels of association with an execution action command (for example, "open file", "transfer", "print", "delete", etc.) as an output solution. They are related to each other through degrees. Reference instruction actions are arranged on the left side through this degree of association, and each execution action instruction is arranged on the right side through the degree of association. The degree of relevance indicates the degree of relevance to which of the execution motion commands the referential motions arranged on the left side are highly related. In other words, the degree of association is an index indicating the high possibility that each instructed action is linked to any execution action command, and is an indicator of the accuracy in selecting the most probable execution action command from the instructed actions. is shown. In the example of FIG. 4, w13 to w19 are shown as association degrees. These w13 to w19 are shown in 10 stages as shown in Table 1 below, and the closer to 10 points, the higher the degree of correlation between each combination as an intermediate node and the execution operation command as an output. Conversely, the closer to one point, the lower the degree of correlation between each combination as an intermediate node and the execution action command as an output.

Such three or more levels of association w13 to w19 shown in FIG. 4 are obtained in advance. In other words, in determining the actual search solution, which of the processing operation and the execution operation command was adopted and evaluated is accumulated in the past data sets, and these are analyzed and analyzed, as shown in FIG. Build up the relevance.

For example, in the past, it is assumed that "open file" was judged to be most suitable as an execution action command for reference instruction action P01 and was evaluated. By collecting and analyzing such a data set, the degree of association with the reference pointing action is strengthened.

This analysis may be performed by artificial intelligence. In such a case, for example, in the case of instruction action P01, the analysis is performed based on the past data set. In the case of the instructed action P01, if there are many cases of "open file", the degree of association leading to this execution action command is set higher, and if there are many cases of "print", this execution action is set. Set higher relevance leading to orders. For example, in the example of processing operation P01, "open file" is linked to "print", but from the previous example, the relevance of w13 leading to "open file" is set to 7 points, and "print" is linked. is set to 2 points.

In addition, this degree of association may be composed of nodes of a neural network in artificial intelligence. That is, the weighting coefficients for the outputs of the nodes of this neural network correspond to the degrees of association described above. Moreover, it is not limited to a neural network, and may be composed of all decision-making factors that constitute artificial intelligence.

In such a case, as shown in FIG. 5, a processing operation is input as input data, an execution operation instruction is output as output data, at least one or more hidden layers are provided between the input node and the output node, and the machine You may make it learn. Either or both of the input nodes and the hidden layer nodes are set with the above-mentioned degree of relevance, which serves as a weighting for each node, based on which the output is selected. Then, when the degree of association exceeds a certain threshold, the output may be selected.

Such a degree of association becomes learned data in terms of artificial intelligence. After such learned data is created, a search for an execution motion command is actually newly performed from the instructed motion. In such a case, the instruction action extracted in step S12 is newly obtained. Based on the newly acquired instructed action, a search is made for an execution action command that matches it. In such a case, reference is made to the degrees of association shown in FIG. 4 (Table 1) that have been acquired in advance. For example, if the newly acquired instruction action is the same as or similar to P02, "transfer" is associated with w15 and "print" with w16. In such a case, "transfer" with the highest degree of association is selected as the optimum solution. However, it is not essential to select the one with the highest degree of association as the optimum solution, and "print", which has a low degree of association but still recognizes the association itself, may be selected as the optimum solution. In addition, it is of course possible to select an output solution that is not connected by an arrow, and any other priority may be used as long as it is based on the degree of association.

At this time, a synonym dictionary may be stored in the storage unit of the automatic program code generation device 1. A synonym dictionary is a dictionary about synonyms. The synonym dictionary contains, for each of one or more reference instruction actions stored in the storage unit of the automatic program code generation device 1, a word possessed by the reference instruction action and one or two or more synonyms of the word. is registered. Specifically, for example, "send", "distribute", etc. may be registered in association with the instruction action for reference "transfer". Through this synonym dictionary, it is possible to determine whether or not the newly acquired pointing action is similar to the referring pointing action. If the instruction action for reference is "transfer" and the newly acquired instruction action is "send", it corresponds to "distribute" because it is registered in advance as a similar action in the synonym dictionary. It is possible to judge that

By referring to such a degree of association, it is possible to accurately determine to which execution action command the instruction action extracted from the text data corresponds.

　Figs. 6 and 7 show an example of using an object case table in which an object case for reference and an object for an actual execution action are associated with a degree of association of three or more levels.

In this object case table, it is assumed that three or more degrees of association between the reference case and the execution action object are set in advance. Assume that the input data are reference cases P01 to P03, for example. For example, the reference case P04 is "mail at 12 o'clock on December 3", the reference case P05 is "statement of delivery", and the reference case P06 is "weekly report". Such reference objects P04 to P06 as input data are linked to execution action objects as outputs.

Such a reference case is applied to the execution action object (for example, "mail received at 14:00 on December 7", "invoice", "daily report", "specification", etc.) as this output solution. They are related to each other through three or more levels of relatedness. Objects for reference are arranged on the left side through this degree of association, and each execution action object is arranged on the right side through the degree of association. The degree of relevance indicates the degree of relevance to which execution action object the reference case arranged on the left side has a high degree of relevance to. In other words, the degree of association is an index indicating whether each object case is highly likely to be associated with an execution action object, and is an indicator of the accuracy in selecting the most probable execution action object from the object cases. is shown. In the example of FIG. 6, w13 to w19 are shown as association degrees. Of course, the examples of association degrees in FIG. 7 are not necessarily the same as those in FIG.

Such three or more levels of association w13 to w19 shown in FIG. 6 are obtained in advance. In other words, in determining the actual search solution, which of the target case and the target of the execution action was adopted and evaluated is accumulated in past data sets, and these are analyzed and analyzed, as shown in FIG. Build up the relevance.

For example, in the past, it is assumed that "bill" was judged to be most suitable as an object to be executed for reference object P04 and was evaluated. By collecting and analyzing such data sets, the degree of association with the reference case becomes stronger.

This analysis may be performed by artificial intelligence. In such a case, for example, the target P04 is analyzed based on the past data set. In the case of the target case P04, if there are many cases of "mail received at 14:00 on December 7", the degree of association leading to this execution action target is set higher, and if there are many cases of "invoice" , the degree of association leading to this execution action target is set higher. For example, in the example of reference object P04, "email received at 14:00 on December 7" is linked to "daily report", but from the previous example, it leads to "email received at 14:00 on December 7". The degree of relevance of w13 is set to 7 points, and the degree of relevance of w14 connected to "daily report" is set to 2 points.

In addition, this degree of association may be composed of nodes of a neural network in artificial intelligence.

In such a case, as shown in FIG. 7, a processing operation is input as input data, an execution operation instruction is output as output data, at least one or more hidden layers are provided between the input node and the output node, and the machine You may make it learn. Either or both of the input nodes and the hidden layer nodes are set with the above-mentioned degree of relevance, which serves as a weighting for each node, based on which the output is selected. Then, when the degree of association exceeds a certain threshold, the output may be selected.

Such a degree of association becomes learned data in terms of artificial intelligence. After such learned data is created, a new search for an execution action target is actually performed from the object case. In such a case, the object cases extracted in step S12 are newly obtained. Based on the newly acquired object case, a search is made for an execution action object that matches the object case. In such a case, reference is made to the degrees of association shown in FIG. 6 (Table 1) that have been acquired in advance. For example, when the newly acquired object case is the same as or similar to P05, "invoice" is associated with w15 and "daily report" is associated with w16. In such a case, the "bill" with the highest degree of association is selected as the optimum solution. However, it is not essential to select the one with the highest degree of association as the optimum solution, and the "daily report", which has a low degree of association but is recognized for its association itself, may be selected as the optimum solution. In addition, it is of course possible to select an output solution that is not connected by an arrow, and any other priority may be used as long as it is based on the degree of association.

At this time, a synonym dictionary may be stored in the storage unit of the automatic program code generation device 1. In the synonym dictionary, for each of the one or more reference object cases stored in the storage unit of the automatic program code generation device 1, a word possessed by the object case and one or two or more synonyms of the word are registered. It is Specifically, for example, "business report", "business diary", etc. may be registered in association with the object for reference "daily report". Through this synonym dictionary, it is possible to determine whether or not the newly acquired object case is similar to the reference object case. If the object case for reference is "daily report" and the newly acquired object case is "business diary", it corresponds to "daily report" because it is registered in advance as a similar word in the synonym dictionary. can be determined.

　By referring to such a degree of association, it is possible to accurately determine to what execution action object the object case extracted from the text data corresponds.

　Figures 8 and 9 show an example of using a terminal case table in which actual execution action targets are associated with reference terminal cases with a degree of association of three or more levels.

In this terminal case table, it is assumed that three or more degrees of association between the reference terminal case and the execution action destination are set in advance. It is assumed that the input data are, for example, reference end points P07 to P09. For example, the reference ending point P07 is "Taro Yamada's PC", the reference ending point P08 is "Doi's PC", and the reference ending point P09 is "person in charge of customer A company C". and The reference end points P07 to P09 as such input data are linked to the execution action destinations as outputs.

Such an end point case for reference is the destination of the execution action as this output solution (for example, "Taro Yamada's PC", "Tokugawa's PC", "Person in charge B of company A", "Shared development department PC", etc.) are related to each other through a degree of association of three or more levels. The end point case for reference is arranged on the left side through this degree of association, and each execution action destination is arranged on the right side through the degree of association. The degree of relevance indicates the degree of relevance with which execution action destination the reference end case arranged on the left side is highly relevant. In other words, the degree of association is an index that indicates the high possibility that each end case is associated with an execution action destination, and is an indicator of the accuracy in selecting the most probable execution action destination from the end case. is shown. In the example of FIG. 8, w13 to w19 are shown as association degrees. Of course, the examples of association degrees in FIG. 8 are not necessarily the same as those in FIGS.

Such three or more levels of association w13 to w19 shown in FIG. 8 are obtained in advance. In other words, in discriminating the actual search solution, past data sets are accumulated to determine which of the end point case and the execution action destination was adopted and evaluated, and these are analyzed and analyzed, as shown in FIG. Build up the relevance.

For example, in the past, it is assumed that "Person in charge B of customer A company" was judged to be the most suitable as an execution action destination for the reference end point P07 and was evaluated. By collecting and analyzing such a data set, the degree of association with this reference ending case P07 becomes stronger.

This analysis may be performed by artificial intelligence. In such a case, for example, the target P07 is analyzed based on the past data set. If there are many cases of "Taro Yamada's PC" when the end point is P07, the degree of association leading to this execution action destination is set higher, and there are many cases of "Person in charge B of company A". In this case, the degree of association leading to this execution action destination is set higher. For example, in the example of the reference end point P07, "Taro Yamada's PC" and "Person in charge B of company A" are linked. is set to 7 points, and the degree of association of w14 connected to "Person in charge B of company A" is set to 2 points.

In such a case, as shown in FIG. 9, a processing operation is input as input data, an execution operation instruction is output as output data, at least one or more hidden layers are provided between the input node and the output node, and the machine You may make it learn. Either or both of the input nodes and the hidden layer nodes are set with the above-mentioned degree of relevance, which serves as a weighting for each node, based on which the output is selected. Then, when the degree of association exceeds a certain threshold, the output may be selected.

Such a degree of association becomes learned data in terms of artificial intelligence. After creating such learned data, a new search for the destination of the execution action is actually performed from the end point case. In such a case, the final case extracted in step S12 is newly obtained. Based on the newly acquired end point case, a search is made for an execution action destination that matches this. In such a case, reference is made to the degrees of association shown in FIG. 9 (Table 1) that have been acquired in advance. For example, if the newly acquired end point is the same as or similar to P08, "Tokugawa's PC" is w15 and "Person in charge B of company A" is associated with degree w16. In such a case, "Tokugawa's PC" with the highest degree of association is selected as the optimum solution. However, it is not essential to select the one with the highest degree of correlation as the optimal solution, and instead select "Person in charge B of company A", who has a low degree of correlation but is recognized for the correlation itself, as the optimal solution. good too. In addition, it is of course possible to select an output solution that is not connected by an arrow, and any other priority may be used as long as it is based on the degree of association.

At this time, a synonym dictionary may be stored in the storage unit of the automatic program code generation device 1. In the synonym dictionary, for each of one or more reference terminal cases stored in the storage unit of the automatic program code generation device 1, a word possessed by the terminal case and one or more synonyms of the word are registered. It is Via this synonym dictionary, it may be determined whether or not the newly acquired terminal case is similar to the reference terminal case.

　By referring to such a degree of association, it is possible to accurately determine to which execution action destination the terminal case extracted from the text data corresponds.

By performing the association analysis of the program code in steps S13 and S14 in this manner, the execution action instruction corresponding to the extracted instruction action, the execution action object corresponding to the extracted object case, and the extracted end case The execution action destinations corresponding to are narrowed down.

Next, the details of the method of creating the program code in step S15 will be explained. The creation of the program code extracts the basic syntax of the highly relevant program code based on the execution action target, execution action target, and execution action destination searched through the degree of association. Then, the object of execution action, the object of execution action, and the destination of execution action searched through the degree of association are substituted for the basic syntax of the extracted program code.

Table 2 shows an example of searching for the basic code of the program code.

The basic syntax of program code is the prototype of program code for actually executing various processing operations on a PC or the like. The program code can be completed by substituting necessary wordings into the basic syntax of the program code. An example of the basic syntax of the program code is "INSERT INTO registration destination VALUES {param1};" ({param1}) as the execution target is already completed in the syntax. The program code can be completed by substituting the execution action target, the execution action target, and the execution action destination actually searched through the degree of association into such a syntax. For example, by substituting the searched execution destination "Taro Yamada's PC" as the "registration destination", and substituting the searched sales data as ({param1}), "Sales data to Taro Yamada's PC The program code "Register" can be completed.

Similarly, in the case of text data ``Display a presentation slide on the conference room display'', the object case is ``presentation slide'', the end case is ``conference room display'', and the instruction action is is extracted, and an execution action target, an execution action target, and an execution action destination corresponding to each are searched. Then, the basic syntax of the program code corresponding to this execution operation destination is similarly searched.

In practice, the basic syntax of the program code is searched via the code table shown in Table 2. In the code table, the basic syntaxes of the program code correspond to the reference execution action target, the reference execution action target, and the reference execution action destination in the same row. For example, if the reference execution operation target is "billing data" and the reference execution operation command is "DB reference", the basic syntax of the corresponding program code is "SELECT billing amount".

At this time, the combination of one reference execution-action object, reference execution-action target, and reference execution-action target may correspond to a plurality of basic syntaxes of program code, and conversely, one basic syntax may correspond to a plurality of references. In some cases, it corresponds to a combination of an execution action target for reference, an execution action target for reference, and an execution action destination for reference.

By storing such a code table in the storage unit of the automatic program code generation device 1, it can be read out and referred to in the process of calculation by the calculation unit of the automatic program code generation device 1.

Fig. 10 shows an example of searching for the basic syntax of a program code through machine learning using artificial intelligence for a code table. In the example of FIG. 10, a combination having a reference execution operation command, a reference execution operation target, and a reference execution operation destination as input variables in the code table is input, and the basic syntax of the program code is output. Specifically, it is assumed that the input data are, for example, a reference execution operation command ("printer output", "open file", etc.), a reference execution operation target, and a reference execution operation destination. An intermediate node 61 is a combination of a referencing execution action command as input data, a referencing execution action target, and a referencing execution action destination. Each intermediate node 61 is also connected to an output. In this output, basic syntaxes A to E of the program code are displayed as output solutions. The basic syntax A of the program code is, for example, "SELECT billing amount FROM billing data WHERE company = {param1} AND billing month = {param2};", and basic syntax B is, for example, "Write Line product list.dat {param1 } ;".

Each combination (intermediate node) of a reference execution action instruction, a reference execution action target, and a reference execution action target is mutually associated through three or more levels of association with the basic syntax of the program code as the output solution. They are related. Execution action instructions for reference, objects of execution action for reference, and execution action destinations for reference are arranged on the left side through the degree of association, and basic syntaxes are arranged on the right side through the degree of association. The degree of relevance indicates the degree of relevance to the basic syntax for the execution action instructions, etc., arranged on the left side. In other words, the degree of association is an index that indicates to what basic syntax of the program code each reference execution action instruction, reference execution action target, and reference execution action target are likely to be associated with each other. , execution operation instructions, etc., to select the most probable basic syntax of the program code.

Such three or more levels of association w13 to w22 shown in FIG. 10 are obtained in advance. In determining the actual search solution, past data is used to determine which of the reference execution operation instruction, the reference execution operation target, the reference execution operation destination, and the basic syntax of the program code in that case was suitable. By accumulating and analyzing these, the degree of association is created.

This analysis may be performed by artificial intelligence. In such a case, for example, when the reference execution operation command is "transfer" and the reference execution operation target and reference execution operation destination are "sales data, Taro Yamada's PC", the program code Analyze the basic syntax from past data. When there are many examples of basic syntax C, the degree of association leading to this basic syntax C is set higher. The degree of association leading to the basic syntax C is set low. For example, in the example of the intermediate node 61a, the outputs of basic syntax A and basic syntax B are linked. The degree is set to 2 points.

Also, the degree of association shown in FIG. 10 may be composed of nodes of a neural network in artificial intelligence. That is, the weighting coefficients for the outputs of the nodes of this neural network correspond to the degrees of association described above. Moreover, it is not limited to a neural network, and may be composed of all decision-making factors that constitute artificial intelligence.

In the example of the degree of relevance shown in FIG. 10, the degree of relevance of the node 61b for the basic syntax C is w15, and the degree of relevance for the basic syntax E is w16.

Such a degree of association becomes learned data in terms of artificial intelligence. After creating such learned data, the above-described learned data will be used when actually searching for the basic syntax. In such a case, the execution operation command, the execution operation target, and the execution operation destination searched in steps S13 and S14 are used.

　In this way, the basic syntax of the optimum program code is searched based on the newly obtained execution operation command, the execution operation target, and the execution operation destination. In such a case, reference is made to the degrees of association shown in FIG. 10 (Table 1) that have been acquired in advance. For example, when the newly acquired execution action command is "open file" and the execution action target and the execution action destination are P21, the node 61d is associated via the association degree. , this node 61d is associated with the basic syntax C with w19 and with the basic syntax D with the degree of association w20. In such a case, the basic syntax C with the highest degree of association is selected as the optimum solution. However, it is not essential to select the one with the highest degree of relevance as the optimum solution, and the basic syntax D, which has a low degree of relevance but is recognized as having relevance itself, may be selected as the optimum solution. In addition, it is of course possible to select an output solution that is not connected by an arrow, and any other priority may be used as long as it is based on the degree of association.

Table 3 below shows an example of the degree of association w1 to w12 extending from the input.

An intermediate node 61 may be selected based on the degree of association w1 to w12 extending from this input. In other words, the greater the degree of association w1 to w12, the heavier the weight in selecting the intermediate node 61 may be. However, the degrees of association w1 to w12 may all have the same value, and the weighting in selecting the intermediate nodes 61 may all be the same.

It should be noted that the present invention is not limited to the above-described embodiments. By searching for an execution action target, an execution action target, and an execution action destination, and substituting the execution action target, the execution action target, and the execution action destination into the basic syntax of the program code, the execution action target is assigned to the execution action destination. Although the case where the program code for executing the execution operation specified in the execution operation instruction is generated by the execution operation instruction has been described as an example, the present invention is not limited to this. It is also possible to omit the search for the target of the execution action and generate the program code for executing the execution action specified in the execution action instruction on the target of the execution action.

In addition, the case components to be extracted are not limited to the above-mentioned object case and end case, but may include start case and conditional case.

The starting case defines the place where the object to be executed defined by the object case is generated and the origin of the object to be executed. In the case of the sentence ``Please send file A from Taro Yamada'', the verb ``send'' has ``file A'' as an object case, and furthermore, this is what is sent starting from Taro Yamada. , the starting point case is "Taro Yamada".

Even when using such a starting case, the reference ending case in the ending case table shown in FIG. In other words, a starting point case table is stored in which actual execution action sources are associated with reference starting point cases with a degree of association of three or more levels. Then, the text data is analyzed, and the source of the execution action for the extracted start case is searched by referring to the association degree of the stored start case table. Next, by substituting the searched execution action source into the basic syntax of the program code, the execution action specified in the execution action instruction is executed from the execution action source to the execution action destination. Generate program code for execution.

In addition, the conditional case defines the conditions for executing the action to be executed defined by the objective case. The conditions include when (time), by what method (method), when to start (operation start condition), by when (operation end condition), and the like.

For example, in the case of the phrase ``Send file A by tomorrow'', the verb ``send'' has ``file A'' as the subjective case, and ``by tomorrow'' as the conditional case. Therefore, the conditional case is ``by tomorrow''.

Even when using such a start case, the reference end case shown in FIG. That is, a conditional case table is stored in which actual execution action conditions are associated with reference conditional cases with a degree of association of three or more levels. Then, the text data is analyzed, and the execution action condition for the extracted conditional case is searched by referring to the degree of association of the stored conditional case table. Next, by substituting the searched execution operation condition into the basic syntax of the program code, the execution operation target is specified in the execution operation instruction based on the execution operation condition. will perform the action.

It should be noted that the present invention is not limited to the present embodiment described above. For example, as shown in FIG. 11, based on the degree of association defined in each of the execution action table, object case table, and end case table, a confirmation sentence for confirming the input content is generated for the user. may The generated confirmation sentence may be output as a voice or as an image. In other words, when any of the execution operation command, the execution operation target, and the execution operation destination cannot be suitably specified from the contents of the input from the user, a confirmation sentence is generated again to inquire of the user about the intention. Run a process to verify. While the number of execution action commands, execution action targets, and execution action destinations as search solutions is finite, the types of wording in text data generated by the user are nearly infinite. For this reason, it may not be possible to search for an execution action command, an execution action target, or an execution action destination depending on the pointing action or case component extracted from the text data generated by the user. Further, when the user uses a dialect, depending on the content of the dialect, it may not be possible to narrow down the execution action command, the execution action target, and the execution action destination through the degree of association. In such a case, a confirmation sentence is generated to inquire of the user again about the intention. The purpose of this process is to accept the re-input of the text data from the user in response to the generation and notification of the confirmation text, and this process may be performed in a so-called chat format.

The confirmation sentence may be generated based on the degree of association associated with each execution action command, execution action target, and execution action destination searched through the execution action table, object case table, and end case table. If an execution action command "open file" is searched for in the execution action table, but the degree of association w13 associated with it is 5 out of 10 evaluations, to be honest, the degree of association is not so strong. Therefore, it is possible that the search solution does not reflect the user's intention. In other words, this is the case where the execution action command, the execution action target, and the execution action destination cannot be narrowed down through the degree of association. In such a case, a confirmation sentence is generated, the user's intention is confirmed, and re-input is accepted. Then, based on the content of the received re-input, the solution search may be carried out without hesitation. In this way, when the degree of association associated with the execution action command, the execution action target, and the execution action destination is less than a certain threshold, it is determined that the search reflecting the user's intention is not performed. , may generate a confirmation sentence. This confirmation sentence may be prepared by preparing a plurality of types of example sentences in advance, and substituting the sentence corresponding to the search solution into the example sentence. For example, if you want to check whether or not the user really intends to execute the execution operation command "transfer", you can replace the ●● part of "Do you mean that you want to ●● file A?" A confirmation message may be generated by substituting "transfer (to)".

In addition, instead of setting such a threshold, a confirmation sentence is generated by comparing the degree of association associated with each execution action command, execution action target, and execution action destination searched using each table. You may make it Assuming that the degrees of association associated with the searched execution action command, the execution action target, and the execution action target are compared with each other, if the degree of association associated with the searched execution action target is the lowest, A confirmation sentence for confirming this execution operation target may be generated and output.

In addition, since noise is superimposed on the user's voice, when this is converted into text data, there is no such wording in the thesaurus in the first place, and the pointing action for reference, the object for reference, and the end point for reference There are cases where it is not possible to match with the case. In this way, even if it is not possible to match the referencing action, the referential object case, or the referential end case, a confirmation sentence may be similarly generated and output to the user. This confirmation sentence is generated by creating and storing example sentences for each referential action, reference object case, and reference end case that cannot be matched, and Example sentences may be read out and output in accordance with the object case and the final point case for reference.

If the degree of association is relatively low, it is possible to determine that the user's intent has not been reflected in the search, and generate a confirmation sentence to confirm that intent.

[Embodiment 2]
Next, a program code automatic generation system according to Embodiment 2 of the present invention will be described with reference to FIG. FIG. 12 is a diagram showing an example of an execution operation table according to the second embodiment. The main difference from the program code automatic generation system 100 in Embodiment 1 described above is that artificial intelligence is not used in the execution operation table. Description is omitted.

In the program code automatic generation system according to the second embodiment, in the execution operation table, for example, for the reference instruction operation P01-1 on the input side, the execution operation instruction on the output side is "open file". It differs from the program code automatic generation system 100 in the first embodiment in that the reference instruction operation data on the side and the execution operation command data on the output side are prepared as linked data sets. It should be noted that the execution operation table of FIG. 12 is an example that corresponds to the execution operation table of FIG. 4 in the first embodiment.

[Embodiment 3]
Next, a program code automatic generation system according to Embodiment 3 of the present invention will be described with reference to FIG. FIG. 13 is a diagram showing an example of an object case table according to the third embodiment. The main difference from the program code automatic generation system 100 in the above-described first embodiment is that artificial intelligence is not used in the object case table. Description is omitted.

In the automatic program code generation system according to the third embodiment, in the object case table, for example, if the input side reference case P04-1 is used, the output side execution action target is "received at 14:00 on December 7". Thus, the program code automatic generation system 100 according to the first embodiment prepares the input-side reference object data and the output-side execution action object data as linked data sets. different from Note that the object case table of FIG. 13 is shown as an example corresponding to the object case table of FIG. 6 in the first embodiment.

[Embodiment 4]
Next, a program code automatic generation system according to Embodiment 4 of the present invention will be described with reference to FIGS. 12 and 13. FIG. FIG. 12 is a diagram showing an example of an execution operation table according to the fourth embodiment. FIG. 13 is a diagram showing an example of an object case table according to the fourth embodiment. The main difference from the program code automatic generation system 100 in the first embodiment is that the execution action table and the object case table do not use artificial intelligence. and the explanation is omitted.

In the program code automatic generation system according to the fourth embodiment, in the execution operation table, for example, for the reference instruction operation P01-1 on the input side, the execution operation command on the output side is "open file". The data of the instruction operation for reference on the side and the data of the execution operation instruction on the output side are prepared as data sets in which they are associated with each other.

Further, in the program code automatic generation system according to the fourth embodiment, in the object case table, for example, if the input side reference case P04-1 is used, the output side execution action object is "received at 14:00 on December 7". , the input-side reference object data and the output-side execution action object data are prepared as linked data sets.

The automatic program code generation system in the fourth embodiment differs from the automatic program code generation system 100 in the first embodiment in these respects. The execution action table of FIG. 12 corresponds to the execution action table of FIG. 4 in the first embodiment, and the object case table of FIG. An example corresponding to the table is shown.

In Embodiments 1 to 4, the final case table using artificial intelligence in FIG. is "Taro Yamada's PC", the data of the reference terminal case on the input side and the data of the execution action destination on the output side are replaced with the terminal case table prepared as a linked data set. may be implemented.

1 program code automatic generation device 2 terminal 3 server 4 communication network 61 intermediate node 100 program code automatic generation system

Claims

receiving means for receiving input of text data via voice or an input terminal;
extracting means for extracting a pointing action and a case component consisting of at least an object case related to the pointing action by performing a natural language analysis on the text data received by the receiving means;
storage means for storing at least an execution action table in which an actual execution action instruction for a referential action is associated and an object case table in which an actual execution action object for a reference object case is associated;
An execution action command corresponding to the extracted instruction action and an execution action object corresponding to the extracted object case are searched by referring to the execution action table and the object case table stored in the storage means. a search means for
By substituting the execution operation instruction and the execution operation object searched for by the search means into the basic syntax of the program code, the program code for executing the execution operation specified in the execution operation instruction for the execution operation object A program code automatic generation system comprising: code generation means for automatically generating the program code.
the execution motion table is stored in the storage means, in which actual execution motion commands are associated with reference motion instructions with a degree of association of three or more stages;
2. The method according to claim 1, wherein said search means searches for an execution action command corresponding to said extracted instructed action by referring to a degree of association in an execution action table stored in said storage means. Program code automatic generation system.
the object case table is stored in the storage means in which actual execution action objects are associated with reference object cases with a degree of association of three or more levels;
2. The searching means according to claim 1, wherein said searching means searches for an execution action object corresponding to said extracted object case by referring to the degree of association of the object case table stored in said storage means. Program code automatic generation system.
The extracting means extracts a case component consisting of an end case that depends on the pointing action,
the storage means stores an endpoint case table in which actual execution action destinations are associated with reference endpoint cases;
The searching means searches for an execution action destination for the extracted terminal case by referring to the terminal case table stored by the storage means,
The code generation means assigns the execution operation target searched by the search means to the basic syntax of the program code, thereby causing the execution operation target to be executed as specified in the execution operation instruction. 4. The program code automatic generation system according to any one of claims 1 to 3, wherein the program code for executing an operation is automatically generated.
The storage means stores an end case table in which actual execution action destinations are associated with reference end cases with a degree of association of three or more levels,
5. The program code according to claim 4, wherein said searching means searches for an execution action destination for said extracted terminal case by referring to the degree of association of the terminal case table stored in said storage means. Automatic generation system.
Based on the degree of association associated with each of the execution action command, the execution action target, and the execution action destination searched through the execution action table, the object case table, and the end case table, the search means asks the user a confirmation text generation means for generating a confirmation text for confirming the input contents;
an output means for outputting the confirmation sentence generated by the confirmation sentence generation means by voice or image,
6. The program code automatic generation system according to claim 5, wherein said accepting means accepts re-input of text data by a user.
The code generation means refers to a code table associated with a basic syntax of a program code for a combination of a reference execution action target, a reference execution action target, and a reference execution action destination, and the execution action searched by the search means. 7. The program according to claim 5 or 6, wherein a basic syntax of highly related program code is extracted based on the target, the target of the execution action, and the destination of the execution action, and the substitution is performed for the extracted basic syntax. Code auto-generation system.
The code generation means generates the code table in which combinations of reference execution action objects, reference execution action targets, and reference execution action targets are associated with the basic syntax of the program code at three or more degrees of association. 8. The program code automatic generation system according to claim 7, wherein the program code is referred to.
9. The program code automatic generation system according to claim 1, further comprising program execution means for executing the program code automatically generated by said code generation means.
9. The method according to any one of claims 5 to 8, further comprising operation destination registration means for registering in advance each terminal device in the communication network as each referential execution destination of the terminal case table. Program code automatic generation system.
The extracting means extracts a case component consisting of a starting case related to the pointing action,
The storage means stores a starting point case table in which actual execution action sources are associated with reference starting point cases with a degree of association of three or more levels,
The search means searches for an execution action source for the extracted start case by referring to the degree of association of the start case table stored by the storage means,
The code generation means substitutes the execution action source searched by the search means for the basic syntax of the program code, thereby transferring the execution action target from the execution action source to the execution action destination. 9. The program code automatic generation system according to any one of claims 5 to 8, wherein the program code for executing the execution operation specified in is automatically generated.
The extracting means extracts a case component consisting of a conditional case that depends on the pointing action,
the storage means stores a conditional case table in which actual execution action conditions are associated with reference conditional cases with a degree of association of three or more levels;
The search means searches for an execution action condition for the extracted conditional case by referring to the degree of association of the conditional case table stored by the storage means,
The code generating means assigns the execution operation target searched by the search means to the basic syntax of the program code, thereby causing the execution operation target to be executed based on the execution operation condition. 9. The program code automatic generation system according to any one of claims 5 to 8, wherein the program code for executing the execution operation defined by the operation instruction is automatically generated.
The program code automatic generation system according to any one of claims 2 to 12, wherein said storage means stores said degree of association corresponding to a weighting factor of each output of a neural network node in artificial intelligence. .
9. The program code automatic generation system according to claim 8, wherein said code generation means utilizes said degree of association corresponding to a weighting factor of each output of a neural network node in artificial intelligence.