JPH0154736B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Application of the Invention) The present invention relates to an information terminal processing control device for computers and the like, and particularly to an information terminal processing control device suitable for improving a man-machine interface.

(Background of the Invention) As shown in FIG. 1, a conventional information terminal uses an input device 1 such as a keyboard, and receives input contents from the input device 1.
It consists of a processing device 2 that performs processing, and a display device 3 that displays the processing contents.

In addition, as for the commands of the information terminal, for example, as shown in FIG. Buttons 5 of the key input device 1 corresponding to the buttons 5 are arranged.

With such a configuration, when the user of the information terminal wants to execute the command 4 "transcribe", he or she can simply press the button 5 of the key input device 1 corresponding to the "said" command 4 to execute the desired process. can be executed.

Also, if the user does not understand the meaning of this "transcribe" command and wants to know how to use it, press button 7 corresponding to "HELP" command 6, and then press button 5 of the "transcribe" command. The guidance content written in the manual of the information terminal device is displayed.

By the way, consider the case where, for example, in a screen displaying a city map as shown in Figure 3a, you try to transfer Figure 8 at the lower left corner of the city map to the upper right part 9 of the broken line on the customer card. look.

If the above-mentioned "transcription" command has a condition that "the character string cannot be transcribed to the place where it is written," the user of this terminal must perform the following steps in order to realize the target state shown in Figure 3b. , (1) enter various commands and perform trial and error, (2) read the manual carefully and learn the solution, (3) ask an experienced user, or (4) consult the manufacturer. I asked the person in charge how to input the commands, but there was no way. As a result, for users who rarely use this terminal, handling it is extremely complicated and has a high threshold.

(Objective of the Invention) The object of the present invention is to provide a man-machine interface and an easy-to-operate information terminal processing control device that allows a user to execute desired processing simply by inputting data in natural language (for example, Japanese). The goal is to provide the following.

(Summary of the Invention) The present invention is directed to a user (user) of an information terminal device.
However, if you simply give the content of the job you want to execute in natural language that is normally used in conversation, rather than in the form of a so-called command string, the control device will automatically analyze and understand the meaning and perform the job. The feature is that it can display the command string necessary for the user to the user.

More specifically, the present invention recognizes the target state of the information terminal desired by the user by analyzing input natural language, and by applying a predetermined command to the target state. An intermediate target state to be obtained is determined, and once it is confirmed that the intermediate target state is equal to the current state of the information terminal, a command is executed to reversely execute the command used when transitioning from the target state to the intermediate target state. column,
The feature is that it is configured to be output and displayed to the user.

Another feature of the present invention is that the history of the information terminal - that is, the operation history of the information terminal can be stored by associating and storing changes in the status of the information terminal and commands executed at that time. The point is that it is configured so that it can be displayed and output.

(Embodiment of the Invention) An embodiment of the present invention will be described below with reference to FIG. The input device 1, the processing device 2, and the display device 3 are the same as the conventional information terminal 20, but in this embodiment, the operating state of the processing device 2 and the user's input sentences and questions in natural language are transmitted to the knowledge processing device. 100 is input.

The knowledge processing means 100 generally has the following configuration.

(b) a natural language analysis means 110 that analyzes the input natural language and converts the meaning into an internal representation of the knowledge processing means 100 (for example, Maasin language); (b) upon receiving the analysis result of the analysis means 110; , an inference means 120 that determines what kind of input or question the user is asking; (c) a response sentence creation means 13 that uses the results of the inference means 120 to generate answers and instructions for the user;
0. (d) Rule base means 140 for organizing and classifying and storing operation commands of the information terminal 20 necessary for the inference means to carry out inference; Knowledge processing means 1
00, and at the same time as the operating state of the information terminal 20 changes, the terminal state memory means 150 is configured to change the internally represented state accordingly; (f) the terminal state A history memory means 160 that stores the state before the change every time the contents of the memory means 150 change; and (g) an information terminal 2 that is the output of the inference means 120.
The output of the plan memory means 170 for storing the command operation sequence of 0 and the response sentence creation means 130 is shown to the user via the output display means 180.

The natural language analysis means 110 is composed of a microcomputer or an electronic computer, and executes natural language analysis according to the procedure shown in FIG. 5, as will be described later.

Step 111: The natural language input string is divided character by character, and when the input is completed, control is passed to the next step.

Step 112...Morphological analysis processing in which the input character string divided into individual characters is divided into words by searching for headwords in the dictionary as shown in Fig. 6b.Step 113...Results of the morphological analysis. After receiving the
Syntactic and semantic analysis processing that analyzes the meaning of the input sentence by referring to a frame dictionary as shown in Fig. 7 Step 114...According to the result, the input sentence is converted into a semantic structure, which is used for the following processing. Output processing for. When the process of step 114 is completed, the process returns to step 111, where the input of a natural language character string is awaited again.

The above processing will be explained in more detail. As a specific example, when the input character string ``Zu 1 wo Migiue ni Tenshiya Shitai'' shown in a in FIG. The word is found from the entry word in the dictionary shown in FIG.

When it is found that the entry word "zu" (diagram) exists, it searches for the word "zu1" together with the character "1" that follows it, and finds "diagram 1." In the same way, ``wo'', ``upper right'', ``ni'', ``transfer'', and ``tai'' are first produced.

The extracted characters "Fig. 1" and "wo" are then compared with the connection matrix of Fig. 6c. Then, it is confirmed that the noun "Fig. 1" in the first term and the particle "wo" in the second term satisfy the connection relationship. In this way,
Morphological analysis processing is performed to extract words.

Based on this result, when we examine the semantic conditions of the dictionary shown in FIG. 6b for the predicate "transcribe", we find that the pointers (arrows) 18 and 19 point to the frame dictionary shown in FIG. 7. Therefore, next, the results of the morphological analysis described above are substituted into the frame dictionary.

As can be seen from the item indicated by pointer 19 in the frame dictionary in Figure 7, the main action requires the grammatical condition of (noun + particle "ga"), but since there is no corresponding condition here, the frame Select the default value "Operator" in the dictionary.

Also, from FIG. 7a, the object of the action ``transcribe'' is (noun + particle ``wo''), so if you look into it, you will find that ``FIG. 1'' applies.

Also, when we look up (noun + particle ``ni'') as a word that indicates location, we find ``in the upper right''. Therefore, for the input sentence in Figure 6a, the input sentence in Figure 7b
The analysis results shown in are obtained.

Natural language analysis means 11 obtained as above
The output of 0 is supplied to the inference means 120 and subjected to processing as shown in detail in FIG.

That is, first, the "other information" in the output shown in FIG. 7b is referred to, and it is determined whether there is an item representing a desire or command (step 121).

If there is a term representing the above-mentioned desire or command, this is determined to be plan generation, and the process proceeds to step 123.
If the above term does not exist, proceed to the next step 122.

In step 122, it is determined whether or not there is an item representing a question about past operation history in the above-mentioned "other information". As a result, if the above-mentioned question exists, the process proceeds to a history verification step 124; otherwise, the process proceeds to a step 130 for creating a response sentence.

The plan generation step 123, as shown in FIG. 9, consists of the following processing procedure.

(b) Step 230 for generating the target state (details thereof are shown in Figure 10), (b) Verification assignment step 231 (details thereof are shown in Figure 13), (c) Verification assignment step 231 (details thereof are shown in Figure 13). The state (intermediate or sub-target state) of the terminal before the command is applied that created the target state using the command rule base shown in FIG. 12 is determined by the procedure described later in FIG. Update step
232 (d) Verification step 233 (233) in which it is determined whether the intermediate or sub-target state is equal to the contents of the terminal state memory 150 representing the current state of the terminal, and if they are equal, the processing means 130 is activated and executed to generate a response sentence. e) Processing step 234 of updating the search range so as not to search the same place twice if the result of the judgment is that they are not equal 234 FIG. This is a flowchart showing the processing procedure for creating a target state, and consists of the following steps.

(B) Step 2301 of cutting out the predicate of the input sentence (in the previous example sentence, cutting out "to transcribe") (B) The cut out predicate is extracted from the pre-memorized input as shown in Figure 11. A step 2302 in which the predicates of the input sentence semantic structure are found to be equivalent to the predicate "transcribe" extracted in step 2301 by comparing the correspondence table between the sentence semantic structure and the target state. Assign or replace constants to variables such as X and Y in the terms of the figure), depending on the correspondence from the terms in the input sentence (in the example, ``Figure 1'' is assigned to X and ``upper right'' is assigned to Y) Step 2303 (D) Extract the result of the assignment in the previous step from the target state in Figure 11C (in this case, extract "Existence (Figure 1, upper right") from the target state column in Figure 11)
Step 2304 A and b in FIG. 12 each show an example of a rule base for terminal commands. Generally, the states in which a terminal command can be executed are limited, and a precondition expression defines the conditions under which the command can be executed.

Also, by running that command,
A state that disappears from the precondition expression is an elimination expression, while a state that newly appears as a result of the execution of a command is an addition expression.

In the previous sentence example, change X (for example, "Figure 1") to Z.
(For example, the "copy" command to copy to "top right") says "X (Figure 1) exists at Y (any position on the screen), and a blank area (φ) exists at Z (top right)." The content 151 that says "is being executed" is a prerequisite for copy execution.

Furthermore, as a result of executing the copying process, the deletion formula 152 in FIG.
, and the additional expression 153 in a of the same figure shows the new state that "X exists in Z".

This rule base is based on step 231 in Figure 9.
232, etc. For example, if you assign "Figure 1" to the variable X and "upper right" to the variable Z, the command name in Figure 12 will be "Copy (Figure 1, upper right)". means "copying Figure 1 to the upper right".

In order to execute such a copy command, the condition that "Figure 1" exists somewhere on the screen (this can be anywhere, so leave the Y variable as is) is required. (Figure 1, Y)
The two prerequisites are that the destination for copying this figure is blank, that is, "existence (φ, upper right)."

After the command is executed, the state of "Existence (φ, upper right)" disappears, and instead, the state of "Existence (Fig. 1, upper right)" is generated. The additional formula is memorized.

As is clear from the above explanation, in order to create a state equal to the additional expression, the prerequisite expression must hold. The erasure formula is stored in the history memory 160 each time the state of the information terminal changes, as will be described later.

The processing procedure of the verification substitution step 231 utilizing this fact is shown in FIG. As is clear from the figure, the verification assignment step consists of the following processing procedure.

(B) Step to extract “existence (X, Y)” of the predicate part [C in Fig. 11] of the goal state found in the goal generation processing step (2304 in Fig. 10)
250. Note that this process can be performed using, for example, LISP (LIST
In the Processor language, this can be achieved by using ``CAR'' (an instruction to extract only the first part of a list) or ``CDR'' (an instruction to extract parts other than the first part of a certain list).

(b) Determine whether the predicate extracted in the previous step 250 is equal to the predicate of any additional expression in the rule base 140 (FIG. 12), and if they are equal, execute step 252 described later, and determine which rule Processing step 251 that executes step 255, which will be described later, if it is not equal to the predicate in the base.

(c) When the predicates are equal, step 252 of assigning the target parameter to the variable of the corresponding predicate in the rule base.

(d) As a result of assigning the target parameters, it is determined whether there is any logical contradiction between the variables. If there is a contradiction, the process of step 254 described later is executed, and if there is no contradiction, the process is executed. , step 253 to execute step 256 to return to main processing.

(E) If a contradiction occurs as a result of the determination in step 253, step 254 temporarily removes the rule from the rule base and passes control to the previous step 251.

(f) Step 255, which branches from step 251, sets an error code, and then returns to the main process.

Figure 14 shows the state update processing step 232 in Figure 9.
This is a flowchart showing the details.

In step 2321, the additional expression is removed from the target state C in FIG. That is, a state before one step of processing to reach the target state is temporarily created.

In step 2322, a precondition expression is added to the target state, and the resulting state is set as a new intermediate target state.

Thereafter, in the verification step 233 of FIG. 9, the intermediate target state is compared and verified with the current state of the terminal, as described above, and if the two do not match, when the processing procedure returns to the state update step 232 again. , create an intermediate target state that goes back one step further.

The intermediate goal state created in this way is
When the current state of the terminal matches, the target state can be reached from the current state of the terminal by tracing the intermediate target states created so far and sequentially executing the additional expressions.

The operation of the present invention having the above-described configuration will be further explained by a specific example with reference to FIGS. 15 and 16.

When the input sentence 190 shown in FIG. 15 - that is, "I want to transcribe Figure 1 on the lower left to the upper right" is input,
The natural language analysis means 110 divides this into words,
In the goal generation step 230 of the inference means 120, a goal state 191--"Existence (upper right corner of FIG. 1)" is generated.

This target state 191 is compared with the predicate of the additional expression of the rule base 140, and an equal predicate (for example, the 12th
If the additional expression 153 ("existence (X, Z)") in the figure is found, in a verification assignment step 231, assignment is made to each variable to check whether there is a contradiction in each variable.

In this example, by substituting "FIG. 1" into the variable X and "upper right" into the other variable Z of the additional equation 153 in FIG. 12, equation 192 in FIG. 15 is obtained. It can be seen that the additional expression 192 obtained as a result of the above-mentioned substitution is equal to the target state expression 191.

As detailed above, since the state after executing the copy command is an addition expression, the prerequisite expression and elimination expression 192A, which are the states before execution of this command, are the next target state, that is, the intermediate goal 193- "Existence (Figure 1, lower left); Existence (φ, upper right)".

This process corresponds to the status update steps 232, 2321, and 2322 in FIGS. 9 and 14.

Next, according to the procedure of verification step 233 in Fig. 9,
The intermediate goal 193 and the current status 196-
In other words, "Existence (Figure 1, lower left); Existence (text, upper right)" is compared. As a result, intermediate goal 1
Out of 93, "Existence" marked with * (Figure 1, bottom left)
It can be seen that the states match and the conditions are satisfied.

However, the other state “existence (φ, upper right)”
does not match and the condition is not satisfied. Therefore, the 9th
After executing the search range update step 233 in the figure,
The verification assignment step 231 is executed again.

As a result, it is matched with the predicate of the additional expression "Existence (φ, V)" of the deletion command shown in FIG. Existence (φ, upper right)" is obtained.

In this case, the precondition and elimination formula 194A - that is, the variable W of "existence (W, upper right)" cannot be matched with any, so it is left as is and intermediate goal 1
95--that is, create "Existence (Figure 1, bottom left); Existence (W, top right)."

By comparing and substituting the intermediate goal 195 and the current state 196, a post-statement comparison can be achieved. Therefore, when "sentence" is substituted for the variable W, it becomes "Existence (lower left in FIG. 1); Existence (sentence, upper right)", which is found to match the current state. Therefore, the inference step 120 of FIG. 4 ends.

In response to the result of the inference step, in the response sentence creation step 130, the previous intermediate goal is traced backwards, and ``After executing deletion (statement, upper right), execute copying (Figure 1, upper right). ” is displayed on the output display device 180.

Through the process described above, when the user inputs a desired operation or process content in natural language, a series of necessary command strings are displayed on the display device 180. Therefore, if the user simply operates according to the display, he or she can perform the target task accurately and most efficiently.

Note that the (terminal) history memory 160 in FIG.
As shown in Figure 16, for each terminal,
A pointer 19 indicating the position of the most recently applied rule is stored while storing the rules that have been applied sequentially and the state that has disappeared due to the application of the rules.
It has 8.

With this configuration, for example, in step 122 of FIG. 8, when the user asks the question "Why did the upper right sentence disappear?", the inference means 120 (FIG. 4) In the history matching step 124 of the figure, the history memory 1
60 and searches for a state in which "existence (sentence, upper right)" disappears.

As a result, in the example of FIG. 16, it is found that the previous extinction state is equal to state 197. and,
From the above state 197, the applied rule is "Erase"
, so as an answer, ``That is,
It is possible to output (answer) that "This is because the data was deleted."

In the manner described above, the answer to the user's question can be displayed on the output display means 180.

(Effects of the Invention) According to the present invention, simply by inputting the task that the user wants the information terminal to perform in natural language, a command sequence for accomplishing the task can be automatically generated. Even if a mistake is made, the information terminal can be prevented from becoming inoperable due to a deadlock, and the desired work can be carried out efficiently with the minimum number of steps.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the configuration of an information terminal in a conventional computer system, and FIG. 2 is a front view showing an example of displaying commands in a conventional information terminal.
Figures a and b in Figure 3 are front views showing examples of screen displays in conventional information terminals, Figure 4 is a block diagram showing a schematic configuration of an embodiment of the present invention, and Figure 5 is the natural language in Figure 4. A flowchart showing the processing procedure of analysis, FIG. 6 is a diagram showing an example of a dictionary used in the morphological analysis process in FIG. 5, and FIG. 7 is a frame dictionary referred to for the syntactic and semantic analysis process in FIG. 5. 1
FIG. 8 is a flowchart showing the inference processing procedure in FIG. 4; FIG. 9 is a flowchart showing the plan generation processing procedure in FIG. 8;
Figure 0 is a flowchart showing the processing means for generating the target in Figure 9, Figure 11 is a diagram showing an example of memory contents showing the correspondence between the input sentence semantic structure and the target state, and Figure 12 is the rule for terminal commands. FIG. 13 is a flowchart showing an example of the processing procedure for matching assignment in FIG. 9;
The figure is a flowchart showing the processing procedure of the status update in Fig. 9, Fig. 15 is a state diagram for explaining a specific example of information terminal processing of the present invention, and Fig. 16 shows an example of the contents of the history memory. It is a format diagram. 1... Input device, 2... Processing device, 3... Display device,
4, 6... Command, 5, 7... Button, 20... Information terminal, 100... Knowledge processing means, 110... Natural language analysis means, 120... Inference means, 130... Response sentence creation means, 140... Rule base, 150... Terminal Status memory, 160...history memory, 170...plan memory, 180...output display means.

Claims (1)

[Scope of Claims] 1. An information terminal comprising an input device, a processing device that performs scheduled processing in response to input from the input device, and a display device that displays at least one of the content of the processing and the processing result. a processing control device, which divides natural language input from the input device into words, analyzes the meaning of the input sentence, and converts it into an internal expression that can be understood by the processing device; a rule base that stores in advance the state of the information terminal before and after execution of various commands; a terminal state memory means that stores the operating state of the information terminal; and an analysis result of the natural language analysis means. The target state of the information terminal supplied and input in the natural language is changed to an intermediate target with reference to the rule base, and when the obtained intermediate target does not match the current state of the information terminal, the intermediate target repeating the process of changing the target, and when the obtained intermediate target matches the current state of the information terminal, an inference means for outputting a command sequence for changing to each of the intermediate targets; An information terminal processing control device comprising: plan memory means for storing a command sequence for changing to an intermediate target; and means for displaying a command sequence output from the inference means. 2. The inference means includes goal generation means for creating a goal state of the information terminal from the semantic structure of the input sentence;
state updating means for creating an intermediate target state by adding or subtracting the rule-based additional formula and precondition formula to the target state; Compare the status and when the two match,
The additional expression used for the state change is output as a command string, and on the other hand, when the two do not match, the intermediate target state is regarded as a new target state and the state change means is activated. An information terminal processing control device according to claim 1, characterized in that: 3. A processing control device for an information terminal, comprising an input device, a processing device that performs scheduled processing in response to input from the input device, and a display device that displays at least one of the processing contents and processing results. , a natural language analysis means that divides the natural language input from the input device into words, analyzes the meaning of the input sentence, and converts it into an internal expression that can be understood by the processing device; and before executing various commands. and a rule base for storing the state of the information terminal after execution, a terminal state memory means for storing the operating state of the information terminal, and an analysis result of the natural language analysis means, which is inputted in the natural language. The target state of the information terminal is changed to an intermediate target with reference to the rule base, and if the obtained intermediate target does not match the current state of the information terminal, the process of changing to the intermediate target is repeated, and the target state is changed to an intermediate target with reference to the rule base. an inference means for outputting a command string for changing to each of the intermediate goals when the determined intermediate goals match the current state of the information terminal, and a reasoning means for outputting a command string for changing the target state to the intermediate goal plan memory means for storing; means for displaying a command string output from the inference means; and each time the contents of the terminal state memory means change due to execution of the command string, the state of the information terminal before the change is displayed. an information terminal processing control apparatus, comprising: a history memory means for sequentially storing the command sequence; and the plan memory means stores the command sequence in the order of operations. 4. The inference means includes goal generation means for creating a goal state of the information terminal from the semantic structure of the input sentence;
state updating means for creating an intermediate target state by adding or subtracting the rule-based additional formula and precondition formula to the target state; Compare the status and when the two match,
The apparatus is characterized by a collation means that outputs the additional expression used for the state change as a command string, and on the other hand, when the two do not match, the intermediate target state is regarded as the target state and the state change means is operated. An information terminal processing control device according to claim 3. 5. The information terminal according to claim 3 or 4, wherein the history memory means sequentially stores changes in the state of the information terminal in association with commands operated for that purpose. Processing control device.