JPH02207303A - Man-machine interface for plant - Google Patents

Abstract

PURPOSE:To cope with various types of instructions by means of input sentences in various expressions by preparing data to which the relation between a standard sentence pattern and instruction contents is applied. CONSTITUTION:In an input sentence analyzing device 3, a morpheme/syntax analyzing part 31 recognizes words included in an input sentence in a natural language by means of an operator/maintenance crew by means of a morpheme/ syntax analysis, and an instruction contents fixing part 32 identifies the instruction contents with the use of the data to which the relation between the standard pattern prepared beforehand and the instruction contents is applied. Next, a necessary information acquiring part 35 uses an information item necessary for the instruction execution prepared beforehand and the data concerning an acquiring method, and acquires information necessary for the execution instruction by referring to the input sentence, an interactive history, etc. Further a corresponding part 34 to function/argument outputs the instruction contents and the acquired information by making them correspond to the function and argument value of each type of a program prepared in a process computer. Thus the input sentence in the various expressions including the omission of a postpositional word, substantive finish, etc. is coped with.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a man-machine interface for a plant, and particularly to an interface that allows input and output in natural language.

[Conventional technology]

Conventionally, in the operation and maintenance of plants such as nuclear power plants, complex commands to process computers have been issued in the following ways: It relies on single menu selections and switches on the keyboard.

For this reason, it has been desired to develop a man-machine interface that simplifies command operations in order to reduce the burden on operators and prevent human errors during abnormal situations that require a large number of operations. As one measure against this, a method has been developed that captures keywords from, for example, the natural language utterances of operators and recognizes the meaning and content of input sentences based on them (Atomic Energy Society of Japan). 1986 meeting JE8).

(Problem to be solved by the invention) According to the above-mentioned conventional technology, it is possible to input in natural language,
The burden of inputting commands during plant operation is reduced. but,
Since this method is centered on keyword recognition, the expressions of input sentences and types of input commands that can be accepted are limited.

In order to improve man-machine efficiency for plant operation and maintenance, it is necessary to have a method that allows input of various commands regarding operation and maintenance using natural expressions normally used by operators and maintenance personnel. An object of the present invention is to provide an interface that can accommodate various command inputs using various expressions from operators and maintenance personnel.

[Means to solve the problem]

In order to achieve the above object, the man-machine interface of the present invention first recognizes the words contained in natural language input sentences from operators and maintenance personnel by morphological and syntactic analysis, and then uses the words prepared in advance. The contents of the command are identified using data that gives the relationship between the standard sentence pattern and the contents of the command.Next, the input sentence is identified using the data related to the information items necessary for command execution and the method for obtaining them, prepared in advance. ,
Obtain information necessary for command execution by referring to dialogue history, etc. Furthermore, the instruction content and the acquired information are output in association with the functions of various programs prepared in the process computer and their argument values.

[Effect]

Therefore, in the man-machine interface of the present invention, by preparing in advance data that provides the relationship between standard sentence patterns and command contents, it is possible to omit particles that are often seen in dialogue between operators and maintenance personnel. It is possible to respond to a wide variety of commands based on various input sentences, including formal stoppages. Also.

By using data related to information items necessary for command execution and their acquisition methods, it is possible to refer to dialogue history and supplement omitted information by making use of plant-related knowledge, simplifying input and making input more natural. can be realized.

(Example) Hereinafter, a man-machine interface for a plant that is an example of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of an apparatus according to an embodiment. In FIG. 0, 11 is a keyboard, and 3 is an input sentence analysis device. This device includes a morpheme/syntax analysis section 31. Instruction content identification unit 32. Necessary information acquisition unit 33. It is composed of a function/argument mapping unit 34. It also refers to a storage device 35 for data prepared in advance such as dictionaries/grammars, command identification rules, case frames, etc., and a dialogue history stack 36.

Reference numeral 4 denotes a response preparation device, in which a regular test consulting program for searching the knowledge base 41 is prepared. 5 is a response sentence generation device, and 6 is a CRT.

First, the regular test consulting program in the response generation device 4 will be briefly explained.

Figure 2 is part of the knowledge base used in the consulting program. FIG. 3 is a diagram illustrating knowledge of frame formats. As shown in the figure, in the frame representing the test name, T
EST-PURPO5E (test purpose),
Provide a slot such as T E S T -FREQUENC'/ (test frequency) and set its Value.
The corresponding knowledge is given as the value of Furad. To search for this knowledge, for example, to search for test frequency, the function TEST-FREQ can be executed by giving the test name as an argument value. As a result, a process of extracting a value from the test frequency slot of the frame with the corresponding test name is executed.

FIG. 3 is a diagram illustrating the remaining portion of the knowledge base, knowledge in Common Sense Algorithm (CSA) form. As shown in the figure, in the C8A format knowledge, an operation procedure or a change in equipment status caused by an operation is represented by an event and a link connecting the events. Events include an "Action" event that represents an operation or confirmation, and a "5Tate" event that represents the status of a device or the like. In addition, in the link, “
5equential” links, “Action” and “
“0ne-3hot” which represents the causal relationship between “5tate”
"Causal" link y "5tate" which represents an equivalent relationship or a causal relationship between "5tate Couple"
ng” link, etc. Also, a link can have a “Gate” that represents a condition under which the link is satisfied.

As with the frame format knowledge, this C8A knowledge can also be executed by giving a function and its argument values. For example, when searching for a test procedure, the test name can be given as an argument and the function TEST-PROCEDURE can be executed. When this function is executed, for the C8A knowledge corresponding to the test given as an argument, from the first “Action” event, “5equentia
A process of sequentially extracting 'Action' events connected by '1' links is performed.

As mentioned above, for example, an inquiry regarding a regular test can be carried out by executing a function of a consulting program by specifying argument values. Furthermore, in general,
Instructions to various application programs prepared on ordinary process computers can be similarly given by specifying the corresponding functions and argument values.

Next, a description will be given of processing performed by the input sentence analysis device for converting input sentences from operators and maintenance personnel into functions and argument values for the consulting program of the response creation section.

FIG. 4 is a flowchart showing an overview of processing within the input sentence analysis device. The input sentence analysis device analyzes an input sentence input from the keyboard 11, converts it into a processing instruction to the process computer, that is, a function and an argument value, and outputs it. Alternatively, consider the case where the input sentence is ``Genshiro Kuramuno Geninwo Shimese, J (101).

The input sentence analysis device first performs morphological and syntactic analysis (102) of the input sentence by referring to the part-of-speech category of words listed in the dictionary 202 and the grammar 201. As a result, the input sentence is separated and a meaning category and meaning are added to each word. Here, as a method for morpheme/syntax analysis, a commonly used method such as a predictive bottom-up method is used. Here, grammar concerns the structure of sentences. For example, S (sentence) → NP (noun phrase) S. (Sentence) expresses the rule that S on the left side can be composed of NP and S on the right side. The dictionary also contains information about each headword.
Part-of-speech categories, display words converted from headwords to kana-kanji,
It provides four types of information: semantic categories and meanings. However, for particles, only two categories are given: the part-of-speech category and the indicator word.

Note that words included in an input sentence can be similarly recognized by dictionary lookup using the longest match method commonly used in word processors and the like.

Next, a process (103) for identifying the contents of the instruction is performed. In this process, the result of syntax analysis is taken in, and the instruction content is identified using the instruction identification rule (203). The command is prepared in correspondence with a processing function in the regular test consulting program, such as, for example, EVENT-CAUSE (search for an event that causes a specified event).

The instruction identification rule provides a standard sentence pattern for identifying each instruction.

In instruction identification, the input sentence and each sentence pattern are sequentially matched, and the instruction corresponding to the matched sentence pattern is determined to be the input instruction.In the example shown in Figure 0, the sentence pattern "
``Show cause % + do'' is matched with the input sentence, and an identification result is obtained that the input command is EVENT-CAUSE. Here, to give the sentence pattern,
A symbol is provided. "%" is a symbol representing a word used for matching only when there is a particle in the input sentence. This makes it possible to deal with the omission of particles, which is often seen in dialogue sentences between operators and maintenance personnel. ``+'' is a symbol that represents an optional word and is used to stop the formal expression.

Further, "?" is a symbol representing an arbitrary word, and this makes it easy to set a sentence pattern.

Furthermore, "<" indicates matching with the meaning category of the word.

Here, the flow of processing in the instruction content identification section will be explained. FIG. 5 is a flow chart showing the process. The instruction identification section first sets the sentence pattern number k to 1 as shown in the figure. Further, for that sentence pattern, i representing the number of words in the sentence pattern is set to the total number of words I in the sentence pattern. Further, j indicating the number of words in the input sentence is set to the total number J of words in the input sentence (step 401).

Next, the i-th word P1 of the k-th sentence pattern and the j-th word WJ of the input sentence are brought (step 402). The following matching process is performed for Ps and WJ. Note that since i is initially set to I and j is set to J, words are compared from the end of the input sentence and sentence pattern toward the beginning of the sentence.

The matching process differs depending on the form of the word P1 in the sentence pattern, that is, the presence or absence of the above-mentioned symbol (step 403).

First, we will explain the process when no symbol is attached. In this case, compare Pl and WJ (step 404).
, if they are the same, i and j are reduced by 1 (step 405). In other words, the word to be compared is the one before. If j is O as a result (step 406), this means that all the words of the sentence pattern are matched with the words of the input sentence, so the command name corresponding to the sentence pattern k is assumed to be the input command content. Output (step 410
), on the other hand, if i is 1 or more (step 406), a determination is made regarding j (step 407),
As a result, if j is 1 or more, the process from step 402 is repeated.

Conversely, if j is O, matching with the input sentence has failed, so preparations are made for matching with the next sentence pattern (k+1) (step 408).
If a sentence pattern still remains (step 409), the processing from step 402 is executed for the new sentence pattern. On the other hand, when there are no more sentence patterns, a message indicating that the instruction cannot be identified is output and the process ends (step 412).
.

If Pi has the symbol rku, Pi and WJ
The semantic categories of are compared (step 413).

Depending on the result, the same processing as described above is performed.

If the symbol "%" is attached to Pl, it is determined whether the part of speech category of WJ is a particle (step 414
), and if the result is a particle, Pl and WJ are compared (step 415), and the same processing as described above is performed depending on the result. On the other hand, if W4 is not a particle, j is subtracted by 1, and if j is 1 or more (step 417), step 4
Return to 02. On the other hand, if j is 0, since there are no more words in the input sentence, it is assumed that the matching between the sentence pattern and the input sentence has failed, and the processing for the next sentence pattern, that is, the processing from step 408 is executed.

If Ps has a symbol “+”, Pt
and WJ are compared (step 418), and if they are the same, the processing from step 405 is executed.

On the other hand, if they are different, after subtracting i by 1, if i is 1 or more, the process returns to step 402. If i becomes O, all words in the sentence pattern (except those that can be omitted)
Since it was included in the input sentence, step 41
0, and the instruction identification ends.

If P is the symbol "?", the word Pi before "?"
Compare j and W- (step 421).

If one word matches as a result, subtract 1 from i and get 1.
Matching processing for the next word is executed from step 405. On the other hand, if they are not the same, subtract j by 1, and j becomes 1.
If this is the case (step 424), matching is performed from step 402 using the word immediately before the input sentence. If j is 0, matching with one sentence pattern has failed, so matching with +1 to the next sentence pattern,
That is, steps from step 408 are executed.

Next, a process (104 in FIG. 4) for acquiring information necessary for executing the command is executed.

Some instructions require some information in order to execute. As a method to control this information acquisition, necessary information items and their acquisition methods were prepared as frame-format data. This is called a case frame. A case frame is defined for each instruction, and each slot of the instruction name frame is given an item of information necessary for executing that instruction. In the illustrated example, it is shown that two pieces of information, "test name" and "event number", are required for the command EVENT-CAUSE. The farad and value of each slot are given the data necessary to obtain that information. The VALUE facit is given the semantic category name of the word to be searched from the input sentence, and the IF-NEEDED facit is given the name of the function to be activated when the required word is not found in the input sentence. Here, the function 5TACK
is a function that refers to the dialogue history stack, and function ASK is a function that requests additional input of information from the questioner. Further, the function GET-EVENT-No is a function for obtaining an event number from an input sentence. This function expresses the event name, ACT (act), GOMP (equipment name),
Search the input sentence for words that have ATTRIB (equipment attribute) and 5TATE (equipment status) as semantic categories,
The event number is determined by comparing the result with the knowledge event name expressed in C8A format.

The flow of this process will be explained using FIG. FIG. 6 is a flowchart showing the process. As shown in the figure, this process first extracts the case frame that has the instruction name as the frame name (
Step 501). Next, set i to 1 (step 5
02). After this initial processing, the VAL of the i-th slot
Read the value of UE facit and set it as the value of vi (
Step 503), if the value of Vi is nil (
Step 504), from the input sentences, the semantic category Vi
Search for words with (step 505). If the word is found (step 506), the obtained word is set as the value of the information item of the i-th slot (step 507), and if there are still slots remaining (step 510), i is incremented by 1 (step 507). 511), the process moves to the next slot.

On the other hand, if the word is not found in step 506.

Alternatively, if the value of Vi is nil in step 504, IF-Nil! HD Read the HD Facit value N
Set as the value of i (step 508). Then, N
Execute functions of i sequentially. As a result, if a value is obtained, it is set as the value of the information item of the i-th slot (step 509), and processing moves on to the next slot.

An overview of the processing will be explained for the example shown in FIG. Command EVE
The case frame corresponding to NT-CAUSH has two slots: "r test name" and "event number."

Regarding the "test name", first of all, the semantic category TEST-NAM given as the value of the VALUE facit.
Search for words with E in the input sentence. Currently, there are no words with this semantic category in the input sentence. Therefore, next I
The value of F-NEEDED is referred to, and the first given function 5TACK is executed. As a result, in the example in Figure 0, where the "Test Name" part of the dialogue history stack is referenced, the MSIVIO% Closed Test (MSIV-TEST)
), and it is assumed that the "Test Name" stack contains MSIV-TEST as a value. Therefore, by executing the function 5TACK, MSIV-TEST is obtained and set as the value of "test name".

Regarding "event number", there is no value in the VALUE facit (nil), so the value of the IF-NEEDED facit is referenced and the function GET-EVENT-N
O is executed. This function starts with the input sentence, ACT,
Search for words with a semantic category such as COMP. As a result, it is discovered that the input sentence includes "nuclear reactor" as a word with the semantic category COMP, and "scrum" as a word with 5TATE elements. Compare this result with the event name in C and SA format shown in Fig.
, to get 204 as the number of the event that has "reactor" and "scrum" as the value of 5TATE,
We considered the case where both "nuclear reactor" and "scrum" were input, but if only "scram" was input,
Similar processing is performed by referring to the plant-related knowledge that "scrum" is the state of a "nuclear reactor."

Finally, processing for mapping to functions and arguments (105
). In this process, the values of commands and information items obtained in the previous process are output to the corresponding program in association with program functions and argument values. In the example shown in FIG. 4, the function name and argument values in the response generation section and the commands of the input sentence analysis section.

Since the values of the information items are the same, it is only necessary to output the obtained command and information item value as a list.On the other hand,
If the input sentence analysis section and the response creation section have different command names, etc., they are associated using a conversion table.

Processing for knowledge retrieval is performed in the response creation unit according to processing instructions from the input sentence processing device. The results are sent to the response sentence generation device as needed.

The response sentence generation device generates an output sentence based on the result and outputs it on a CRT. An overview of the processing in this output sentence generation device will be explained using the results of a consulting program as an example.

FIG. 7 is a diagram showing an example of processing in the response sentence generation device. For example, event names are expressed in the knowledge base used in consulting programs as shown in the diagram. That is, the "Action" event is expressed by an action (ACT), a device name (Comp), an attribute (ATTRIB), a word representing a state (STATIE), and a comment (301).

The output sentence generation device has rules (30
2) is prepared and is used to generate the output sentence.In the example in Figure 1, if the event is "Action",
In addition, the output sentence ``Confirm plant stability [before testing]'' is generated using the rule that should be used when the word representing an action includes ``confirm''. Here, the word order is changed and particles are inserted.

As another example, we have shown the process of generating an output sentence when there is a question about the possibility of an erroneous operation. At this time, if event A is obtained as a result of a correct operation and event B is obtained as a result of an incorrect operation (304), the rule 305 is applied and the output sentence "Event A when it should occur, Event B" (306) is generated. .

FIG. 8 is a diagram showing the operation of the interface according to the above-described embodiment. In response to input requests from the computer, rHP
CS Pump Shikennomoku Tekiha. '' is entered from the keyboard, the resulting sentence is morphologically and syntactically analyzed and the separated sentence is displayed on the CRT. The consulting program's knowledge base search function (search for testing purposes) is then executed and the results are displayed on the CRT. The user then enters "Hindwo Shimese," and the response is similarly presented. Here, since the input sentence does not have a test name, the test name necessary for executing the command is obtained by referring to the dialogue history stack.

Next, from the keyboard, rMsIV Heishi Kenmoku Teki Hikouji. ” is entered and the purpose of the MSIV closed test is presented. Next, "Shikenshuukiha," is input,
A response is presented. The above answers can be obtained by searching for frame-based knowledge.

Next, "Shikentejunwo Shiritai." is input,
The answer is obtained by searching for C8A format knowledge in the response creation section and generating a response sentence based on the result.

As described above, according to the apparatus according to the present embodiment, it is possible to deal with input sentences in various expressions including omitted particles, two-indication stops, and the like. In addition, when giving sentence patterns, not only the denotation words of words but also their meanings and semantic categories can be used, so the number of patterns that need to be prepared can be reduced compared to when giving only denotation words of words. Can be done.

Furthermore, by using symbols, it is possible to easily express the omission of particles as a sentence pattern.

Also, by referring to the dialogue history, it is possible to supplement omitted information in input sentences, so there is no need to input the same word each time.
Simplification of input can be measured. Furthermore, by using plant-related knowledge, it is possible to further reduce the amount of information that needs to be input.

Additionally, more natural dialogue is possible by supplementing information with the user.

Therefore, by using this device, operators and maintenance personnel can input commands more naturally and easily. Furthermore, with this device, it is possible to present responses from a computer in the form of natural language sentences. These improvements improve man-machine operation and maintenance. Furthermore, since the response sentence generation method can be given in the form of rules, maintenance of the response sentence generation section is easy.

Although the device of this embodiment uses a keyboard and a CRT as input/output devices, if a commercially available audio input/output device is used, it will be possible to output natural language audio using the same processing method as this device. It is possible to have a dialogue. Also 1
A liquid crystal display device or the like can also be used as a display device.

The man-machine interface described above can be applied not only to plant operation and maintenance, but also to general interactive systems such as consulting systems and design support systems.

〔Effect of the invention〕

As explained above, according to the present invention, input sentences related to various commands in various expressions including particle omission and declarative stop,
It is possible to convert it into processing instructions for various programs prepared in the process computer. Furthermore, it becomes possible to display the execution results of various programs in the form of sentences, by voice, or on a display device. This can reduce the burden on operators and maintenance personnel in inputting commands and recognizing output from the computer, and also reduces the possibility of human error. Therefore, the safety of using such a man-machine interface.

The effect of improving economic efficiency is significant.

[Brief Description of the Drawings] Figure 1 is a block diagram of a man-machine interface of a plant that is a preferred embodiment of the present invention, and Figures 2 and 3 are explanatory diagrams showing the contents of the knowledge base in Figure 1. , FIG. 4 is an explanatory diagram showing the processing in the input sentence analysis device of FIG. 1, FIG. 5 is an explanatory diagram showing the processing in the instruction content identification section of FIG. 1, and FIG. Explanatory diagram showing processing in the information acquisition unit, seventh
FIG. 8 is an explanatory diagram showing the processing in the response sentence generation device shown in FIG. 1, and FIG. 8 shows the operation of the interface of the embodiment shown in FIG. In particular, it is an explanatory diagram showing an example of a CRT screen. 3... Input sentence analysis device, 4... Response generation device, 5.
. . . Response sentence generation device, 6 . . CRT, 11 . Keyboard, 31 .・
Function/argument mapping unit, 35...Storage device, 36
...Dialogue history stack, 41...Knowledge base, 10
1... Input sentence, 102-105... Processing step, 201-204... Data used, 301, 304...
...Input, 302°305...Rules for sentence generation, 303,306...Output sentence, 401-424,50
1 to 511...Processing steps. 3rd year high school figure 8 figure
l ice response 1 figure ■ c > Input between 1 Ll < danoi GC > question demon input rimata no kudanoi ΦC > question 1 input and maita no i ■ c > Ir key input shimatari f: sai

Claims (1)

[Scope of Claims] 1. Means for morpheme/syntax analysis to identify words included in an input natural language sentence, and an input sentence using data that provides a relationship between a sentence pattern prepared in advance and instruction content. means for identifying the contents of the instruction; means for acquiring the information necessary for executing the identified instruction using previously prepared data giving information items necessary for executing the instruction and a method for acquiring the information; 1. A man-machine interface for a plant, comprising means for outputting instructions and information item values in correspondence with functions and argument values of a program on a computer. 2. The first step is to perform morphological and syntactic analysis to identify the words contained in the input natural language sentence, and the first step is to perform morphological and syntactic analysis to identify the words contained in the input natural language sentence. A second step of identifying the instruction content of the input sentence by matching with the result of the step, and executing the identified instruction using data that provides the information items necessary for executing the instruction and their acquisition method prepared in advance. The third step is to obtain the information necessary for the process, and output processing instructions to the program on the computer based on the instructions obtained in the second step and the values of the information items obtained in the third step. A man-machine interface for a plant, characterized in that it has a fourth step of: 3. The man-machine interface for a plant according to claim 1 or 2, wherein the sentence pattern used to identify the instruction content is prepared by a word, its meaning, one type of meaning category, or a combination thereof. 4. The man-machine interface for a plant according to claim 1 or 2, wherein the information necessary for instruction execution is obtained by referring to an input sentence or by referring to a dialogue history stack. 5. Claim 1 or 2, wherein the information necessary for executing the command is obtained by one or more of the following: referring to input sentences, referring to stored contents of dialogue history, referring to plant-related knowledge, and making inquiries to the user. Man-machine interface for the described plant. 6. The man-machine interface for a plant according to claim 1 or 2, further comprising means for generating an output sentence by permuting word order and inserting words based on the output from the computer. 7. The man-machine interface for a plant according to claim 1 or 2, further comprising means for generating an output sentence by rearranging the order of words and inserting words using rules prepared in advance based on the output from the computer.