WO2001003006A1

WO2001003006A1 - System and method for natural language dialog by computer and recorded medium

Info

Publication number: WO2001003006A1
Application number: PCT/JP2000/004526
Authority: WO
Inventors: Tadamitsu Ryu
Original assignee: Cai Co., Ltd.
Priority date: 1999-07-06
Filing date: 2000-07-06
Publication date: 2001-01-11
Also published as: TW528966B

Abstract

A system for natural language dialog by a computer comprises input means (101), analyzing means (102), switching means (103), chat mode dialog means (104), general mode dialog means (105), business mode dialog means (106), output means (107), ask-in-return means (108), and ask-in-return answering means (109). The sentence of a natural language inputted by a user is interpreted. The response is made by selecting one of dialog means having different properties and prepared preliminary. If information included in the input sentence is insufficient, asking about that is made to complement the information and retrieval is performed to present necessary information to the user.

Description

Specification

Natural language dialogue system, method and recording medium by Combu

Technology background

The present invention relates to a system that enables dialogue with input sentences expressed in a natural language, and is used as an interface of various information processing devices. Recently, input and response in natural language have been actively developed for interactive information processing devices such as information retrieval systems so that general users can easily operate them. Japanese Patent Application Laid-Open No. 9-169090 describes a process when an incomplete natural language sentence is input to an information processing device, and specifies a demonstrative pronoun included in the input sentence to respond according to the meaning of the sentence. (Japanese Unexamined Patent Application Publication No. 9-1446972).

In the distribution industry, attention is being paid to communication with customers using the Internet. There is a need for a method and system for acquiring the information necessary for business operations from customers as needed while having a natural dialogue with customers.

However, conventional dialog systems only specify the purpose of the dialogue from the beginning as a search request or the execution of some function, and only understand the meaning of the input sentence assuming the determined purpose. Human-to-human dialogue starts with no such assumptions.

SUMMARY OF THE INVENTION It is an object of the present invention to be able to judge an input sentence in a user's natural language, switch an interaction mode in accordance with the judgment, and have a dialogue, and furthermore, a person who can guide a user to a business dialogue. An object of the present invention is to provide a system, a method, and a recording medium that enable a dialogue close to a natural dialogue of a user.

Another object of the present invention is to judge the input sentence in the user's natural language in the state of the dialogue (state of the utterance such as whether the input sentence is an answer to a question asked by the system), and according to the state judgment. Input mode to understand input It is possible to analyze by switching the mode, determine the meaning of the input sentence, perform appropriate work, and return the appropriate response sentence to the user according to the execution. The aim is to provide a system that enables dialogue closer to human dialogue.

Ming Pm

According to a first aspect of the present invention, there is provided a natural language dialogue system using a computer having at least an input means for inputting information by a user. 1) At least two dialogue modes having different characteristics are provided. An interactive mode selection means for selecting either one;

2) morphological analysis of the input sentence and analysis means for extracting word strings as data; and

3) A natural language dialogue system using a computer and a natural language dialogue method between a computer and a computer, which includes a switching means for switching the dialogue mode by determining the result of the analysis.

This conversation mode is

1) A chat mode in which a response is output only according to a preset rule base and no data is acquired.

2) Judge the verb of the input sentence, obtain the case information of the verb by referring to the dictionary prepared in advance, extract the data that fits this case from the input sentence, and find the case where there is no applicable information. In some cases, a general mode in which a question sentence inquiring about the missing case is output to the user, the user is requested to input missing information, and the case is obtained with the case information as an item.

3) Obtain information matching the work item from the input sentence based on the work item set in advance from the input sentence. If there is an item that has not been applied, output a question sentence for asking the missing item to the user, There are at least two job modes that require the user to enter missing information and acquire data corresponding to the missing task item.

Or, at least provide a means for the user to enter information Computer-based natural language dialogue system,

A means for acquiring information that matches the business item from the input sentence by referring to the business item set in advance as necessary for the system administrator's business. A means for outputting to the user a question sentence inquiring about an item, and a means for requesting the user to input missing information and obtaining the data corresponding to an unassigned business item. It can be a natural language dialogue system using a computer including a business mode dialogue means provided with the above, and a natural language dialogue method between a computer and a user. Here, the business mode can be further divided into a plurality of business modes having different business items.

In order to analyze a user's input sentence, the system includes out-of-order processing means for determining a business item in which the input data should be stored by referring to a database storing the extracted data and the items to which the data belongs. Even if data over multiple business items can be accepted and the input data is applied to the missing business items and object data is generated, even if the input items are not specified in advance, the system mechanism will be improved. Even if a user who does not understand inputs the data in any order, it can flexibly respond.

Further, by accessing the database related to the business and acquiring the schema of the database as business items, the schema of the object database to be generated by the system administrator related to the business is obtained. The trouble of setting each time can be reduced.

In addition, if no business dialogue has been performed in a section that meets at least one condition selected from the lapse of a predetermined time and a predetermined number of dialogues, a business guidance means for performing a predetermined inquiry. Inclusion of as a component is effective for marketing.

Furthermore, in the dialogue of the business mode, even if another mode is started without the business item being allocated, the business item that has not been allocated is stored, and the missing item is inquired as necessary. Item monitoring means By including the as a component, it is possible to continue asking the user for necessary information until the end.

In a second aspect of the present invention,

1) Mode switching means for switching the input mode by determining the state of the dialogue;

2) word decomposition means for morphologically analyzing the input sentence,

3) role analysis means for synthesizing the result of the morphological analysis and generating a list of words to which labels indicating syntactic cases are attached;

4) When it is determined that an answer sentence to the question sentence output by the system is input, a hearing supplementing means for extracting a word whose part of speech is a noun from the result of the morphological analysis and extracting it as an answer,

5) Judgment of the input sentence, registration processing of storing the syntax expression data as knowledge in a database, search processing of searching for knowledge already stored in the database, lack of syntactically incomplete input sentence Input sentence determination means for selecting at least one input sentence process selected from a reflection process for prompting the user for input, and a function execution process for activating a preset function;

6) an input sentence processing means for processing the input sentence in accordance with the input sentence determination; and

7) responding means for generating and outputting an output sentence according to the result of the input sentence processing means;

We adopted a natural language dialogue system with convenience that is characterized by including

Also, the method performed by the system is:

1) determining the state of the dialogue and switching the mode for accepting input;

2) analyzing the input sentence in accordance with the mode, reconstructing the data into a list structure composed of words to which the analysis result is labeled, and understanding the input sentence;

3) Analyze the knowledge-based data, transfer knowledge, confirm knowledge, search for knowledge, Determining at least one sentence selected from a preset action request;

4) performing a task according to the result of the determination; and

5) a step of generating, outputting and responding to a response sentence according to the execution of the work.

In the invention according to the second aspect, the input mode includes at least two modes: a normal mode and an answer mode in which an answer sentence to the question sentence output by the system is input.

The input sentence is at least one selected from a text sentence by keyboard input and a sentence converted into a text via a speech recognition device.

The status of the dialogue can be determined based on whether or not the system has output a question sentence to the partner of the dialogue immediately before.

The labeling in the natural language dialogue method between the user and the user is performed by referring to at least one selected from a previously created dictionary and a result of parsing an input sentence.

No ^ ^

FIG. 1 is a block diagram of an entire embodiment of a natural language dialogue system using a combination according to the first embodiment of the present invention.

FIG. 2 is a table showing an example of a rule base used in the natural language dialogue system by the computer of FIG.

Fig. 3 is a table showing an example of a list of tasks and keywords used in the natural language dialogue system based on the combination shown in Fig. 1.

FIG. 4 is a detailed diagram of the chat mode dialogue means in the natural language dialogue system by computer of FIG.

FIG. 5 is a detailed diagram of the general mode dialogue means in the natural language dialogue system by the combination of FIG.

FIG. 6 is a detailed diagram of a business mode dialogue means in the natural language dialogue system by computer of FIG. FIG. 7 is a block diagram of one embodiment of the computer-aided natural language dialogue system according to the second aspect of the present invention.

FIG. 8 is a detailed diagram of the means for supplementing the feedback in the natural language dialogue system using the computer of FIG.

FIG. 9 is a detailed diagram of the input sentence processing means in the natural language dialogue system by the short-lived view of FIG.

FIG. 10 is a detailed diagram of the response means in the natural language dialogue system by the combination of FIG.

昍》 Sharing a longing bear

Hereinafter, embodiments of the present application will be described with reference to the drawings. FIG. 1 is a block diagram of the embodiment of the present invention. This method and system are realized by a program whose operation is controlled by reading a program stored in a storage medium.

The system in this embodiment includes input means 101, analysis means 102, switching means 103, chat mode interaction means 104, general mode interaction means 105, business mode interaction means. It consists of 106, output means 107, listening means 108, and answering means 109.

The input means 101 recognizes the input sentence uttered by the user's natural language, converts the text, and inputs the text to the analysis means 102.

Here, the input means 101 employs a speech recognition device.However, the input device is not limited to this, and even if the user directly inputs an input sentence with the keyboard, the written sentence is converted into an OCR. It may be read by a recognition device.

The analysis unit 102 performs a morphological analysis on the input sentence, generates a word string to which part-of-speech information is added, and sends it to the switching unit 103. The switching means discriminates the input sentence, and switches between chat modes having different characteristics such as chat mode interaction means 104, general mode interaction means 105, and business mode interaction means 106. The output means 107 is a conventional speech synthesizer, The answer sentence generated by the interactive means is output as speech. If there is a shortage in the interactive means, if there is an evening, the listening means 108 is activated as necessary to output an inquiry to the user. The input sentence from the user after the reflection is returned and the answering means is used to supplement the missing data and proceed with the dialogue. Here, a speech synthesizer is applied to the output, but any device that can express words, such as a screen display, can be used.

The switching means 103 examines the word string of the input, refers to a prepared rule base, and detects a greeting, a call, a lack of sufficiency, and a time question. The declarative words (nouns, adjectives, verbs) excluding particles and endings are sent to the chat-mode dialogue means as a parameter together with the discrimination result such as “sufficient”.

Figure 2 shows an example of a rule base to be referenced.

For example, in the case of an input sentence “It's hot,” a sentence of the word “Hot” is sent to the chat mode dialogue means 104 together with the result of the judgment “Lack of tongue”.

In the case of “What time is it now?”, The determination result of “Question of time” is sent to the chat mode dialogue means 104 together with the parameter “Now”. In the switching means 103, when a business keyword is detected in the sent word string, the "business name" determined from the keyword is sent to the business mode dialog means 106 to transmit the business mode. Activate the conversation. It is good to prepare a business keyword list in advance and refer to it when a keyword is detected.If this is the case, set multiple tasks and prepare a keyword list for each task to handle multiple tasks. It is even better because it is possible. Figure 3 is an example of a list of business keys and business items.

If neither the rule base nor the key list is used, the analyzed word string data is sent to the general mode interaction means 105.

FIG. 4 shows the chat mode interaction means 104 in detail. Rule selection means 401, Rule base 402, Response generation means 403, Timer Consisting of one 404.

The discrimination result and the parameter information transmitted from the switching means are input to the rule selection means 401. In the previous example, if the input sentence is "It's hot," the discrimination result is "not enough," and the parameter is "Hot."

If it is "What time is it now?", The discrimination result: "Question of time" and the parameter information: "Now" are input to the rule selection means 401.

The rule selection means 401 refers to the rule base 402 and sends output sentence generation information to the response generation means 404. Figure 2 shows an example of a rule base. In response generation means 404, if the input is "hot", the rule shown in Fig. 2 is received and the response is generated as it is. If the input is "What time is it now?", The input time corresponding to the parameter value "Now" is obtained from the timer 404 and a response sentence "Time is X X minutes" is generated. The response sentence generated by the response generation means 404 is sent to the output means 1 7.

FIG. 5 shows the general mode interaction means 105 in detail. Syntax information generation means 501, verb case list database 502, question sentence discrimination means 503, search means 504, general knowledge data base 505, knowledge generation means 506, Aiuchi It consists of generating means 507 and answer generating means 508. The syntactic information generating means 501 extracts a verb from the word string data analyzed from the input sentence, calls the verb case list from the verb case list database 502, and calls the case information and the verb case list. It generates an object consisting of words and words that apply to the case (hereinafter referred to as “syntax information”). At this time, in the called case list, the case that is not applied by the word data, that is, the insufficient case information in which the insufficient case is detected is sent to the listening means 108. The recall unit 108 refers to the rejection output sentence template (not shown) to generate and output a repetition sentence to the user, and prompts the user to input a reply. For example, if the underserved is a nominative, a return of "Who is it?" Power. When the reflection sentence is output, the input sentence from the user is directly input to the reflection answering means 109, and the word data that satisfies the insufficient case is extracted from the input sentence of the response.

When all the insufficient cases have been applied, the generated sentence information is determined by the question sentence determining means 503 as a question sentence or a positive sentence. If it is a question sentence, the search means 504 performs an object search on the general knowledge database 505 using this syntax information as a key, and the hit object is output to the answer generation means 508. send. In the answer generation means, if the question is a question with a question, the word data corresponding to the case of the question is cut out from the hit object and an answer is generated by referring to a response template (not shown). If there is an object that hits as a result of the object search if the input sentence asks for the truth of the input sentence without question words, it is judged as true, and if not hit, it is judged as false. According to the answer template that is not shown, if it is true, the answer sentence "Yes" is generated, and if it is false, the answer sentence "No, not correct" is generated and sent to the output means 107.

The answer generation for a question with a question is explained in some detail using the question "When did Yoshida go to Shibuya?"

This question sentence becomes syntax information (object) of {action (line), nominative (Yoshida), destination (Shibuya), time (when)} by the syntax information generating means 501. The search means 504 searches the general knowledge database 505 using {action (line), nominative (Yoshida), destination (Shibuya)} as a key. If there is no object to be hit, an answer sentence "I do not know about it" is generated by referring to an answer sentence template (not shown) and sent to the output means 107.

If the object hits and it is {action (line), nominative (Yoshida), destination (Shibuya), time (yesterday)}, the answer template (not shown) ) Is ", the word with the temporal case that matches the case of the interjection, that is," Yesterday "is inserted into (answer) to generate an answer sentence" It is yesterday "and output means 10 Send to 7. FIG. 6 shows the business mode interaction means 106 in detail. Business mode monitoring means 600, mode transition means 600, business specification means 603, item monitoring means 604, search condition collection means 605, case list out-of-order processing means 60 6, Search condition buffer 607, Search replies generation means 608, Case list buffer 609, Case list extraction means 610, Business knowledge database 611 Be composed.

The business keyword and the parameter information sent from the switching means 103 are sent to the business specifying means 603. The business mode may be specified as necessary according to instructions from the business mode monitoring means. The specified business mode is sent to the item monitoring means 604, where the parameter information is organized by business items extracted from the business keyword / business item list and used as business information objects. Generate. At this time, the fact that the parameter information corresponding to the item in the item list is insufficient is referred to by the case list extraction means 6110 by referring to the case list buffer—609. If detected, the missing item is sent to the reflection means 108. Then, a reflection sentence is generated and output which prompts the user to input information corresponding to the insufficiency. Answers from the user are collected in the search condition collecting means 605 through the return answer means 109.

Even if the answer from the user is information that does not match the case requested in the previous hearing, if the case matches the shortage of the currently running case list, the case list is processed in random order processing means 606 Add case information to supplement the case list. An object that satisfies all the case lists is temporarily stored in the search condition buffer 607, and the search / answer sentence generation means 608 searches the business knowledge database—the object base 611 for an object search. Generate answers from the business knowledge obtained. The generated answer is sent to the output unit 107 and displayed to the user using the speech synthesizer and the monitor.

Further explanation will be given with the input sentence “I want to buy a red car” as an example. The example sentence input to the input means 101 is sent to the analysis means 1102 as {adjective (red), Ending (i), noun (car), particle (o), verb (buy), ending (tai)}. This analysis means is determined by the switching means 103 as the business mode, and {business mode (car, buy), parameter (red)} is sent to the business specifying means 603. The task identification means refers to the task keywords and task item list (Fig. 3), identifies the task as “used car sales”, and uses the item monitoring means to identify the specific task “used car sales” as the item monitoring means. Send to 4.

The item list extraction means 610 obtains an item list (manufacturer, price, mission, color) by referring to the business keyword / business item list shown in FIG. Here, the word “color” in the item list fits the word “red”, but other items are missing items. The item monitoring means 604 sends the missing items to the feedback means 108. The feedback means 108 outputs a feedback sentence "Which manufacturer is it?" In accordance with a not-shown feedback sentence template. Here, if the answer from the user matches the question item in the form of “^ product”, the missing item is complemented as it is.

For example, even if an answer that does not match the return sentence is input, such as “a five-seater manual car is good”, the item list is currently out of order in the out-of-order processing means 606 (manufacturer, price, ), And obtains data called “manuals” that apply to the item “missions”. The above questions / replies are repeated, and all the data fits in the business items in the form of {manufacturer-(production), price (300,000 yen), mission (manual), color (red)}. In the search / answer generation means 608, an object search is performed. The knowledge of the used car that has been hit is output to the user's monitor (output means 107) as image data or video data.

Next, a natural language conversation system and method by a combination user who can judge the state of the conversation will be described with reference to the drawings.

FIG. 7 is a block diagram of a computer-based natural language dialogue method / system that can determine the status of a dialogue according to the second embodiment of the present invention.

R printed paper (Rule The system according to the present embodiment includes mode determination means 1 1 1, word decomposition 'mouth analysis means 1 1 2, word decomposition' noun extraction means 1 1 3, input sentence judgment means 1 1 4, case list confirmation means It is composed of 1 15, recalling means 1 16, input sentence processing means 1 17, and response means 1 18.

The mode determining means 111 switches the input mode in response to the notification from the input sentence processing means 117. By default, the mode is the normal mode, and an input sentence from an input device (keyboard, voice recognition device) not shown is sent to the word decomposition / oral analysis means 112. The input sentence processing means 117 performs the question processing, and when there is a notice, the input sentence is sent to the word decomposition / noun extraction means 113 when the answer mode is set.

The word decomposition 'role analysis means 1 1 2 performs morphological analysis of the input sentence, and further performs syntax analysis. An object consisting of a list of words to which syntactic information (hereinafter referred to as “roles”) is generated with reference to a syntactic rule and a verb case list database (not shown).

For example, the input sentence "A of B is C." is, according to the syntax rules, {subject mouth (a), data (A of B): complement role (none), de Ichiyu (C) :} Object. Also, in the case of “D goes to E”, referring to the case list of the detected verb (line), the object {verb (line), verbal (is), data (D), role ( ), And generate data (E)}.

The word decomposition / noun extraction means 113 extracts a word that is a noun from the words cut out with the part of speech information added as a result of the morphological analysis, and sends the word to the reflection replenishment means 116.

The input sentence judging means 1 14 detects the presence of “?”, “Ka” and question words at the end of the sentence, and the positive sentence goes to the case list checking means 115. The question sentence is the input sentence processing means 1 17 And sends the input literary object.

The case list checking means 1 15 calls the case list from the detected verb from a database (not shown) and checks it against the syntax information of the words in the input sentence. If there is, recall the missing information and replenishment means 1 Send to 16 If there is no shortage, the input sentence object is sent to the input sentence processing means 117.

The recall supplementary means 1 16 will be described with reference to FIG. 8 showing the details. If an unsatisfied case is detected as a result of analyzing the input sentence accepted from the normal mode, the unsatisfied case information is sent from the case list checking means. The insufficient case temporary holding means 2 11 1 stores the insufficient case information in the buffer 2 12 and sends the same information to the input sentence processing means 1 17. When input from the answer mode, a word (noun) as a response to the inquiry of a deficiency comes from the word decomposition / noun extraction means 113, and the replenishment means 213 is the simplest one. As shown in Fig. 9, the input sentence processing means 117, the registration processing means 311, the search processing means 312, It consists of confirmation processing means 3 13, question processing means 3 1 4, action execution processing means 3 15, and knowledge database 3 16.

The registration processing means 311 is judged as an affirmative sentence by the input sentence judging means 1 14, and passes through the case list checking means 1 15 and the recall supplementary means 1 16 to obtain a complete object with no shortage. Receive and register a new one by comparing it with the already registered knowledge. To register, enter “n e w”.

"Old" is notified to the response means 1 18.

The search processing means 3 1 2 is turned into a question sentence by the input sentence judging means 1 14, and is activated when the presence of the question is detected by a question detection means (not shown). , "(Role (), data ()} in knowledge database 3

16 is searched and the subject () of the hit knowledge (object) is used as the answer. In other cases, the knowledge is searched for in the subject (), and the content that matches the case of the question word from the hit knowledge is used as the answer. This answer is a response 1

Sent to 8

The confirmation processing means 3 13 is activated when the input sentence determination means 1 14 determines that the question is a question sentence and the presence of a question word is not detected by a question word detection means (not shown). You. Match the input sentence object with the Knowledge Base 3 16 and notify the response means 1 18 of “yes” if they are the same and “no” otherwise. O

The question processing means 3 1 4 detects the missing case as a result of analyzing the input sentence accepted from the normal mode, and receives the missing case information via the case list checking means and the short case temporary holding means. Fired when The case particle "ha", "ga", "he", "ni", etc. is converted into the terms such as subject, destination, departure place, etc. using the conversion rules prepared in advance, and the response means. To send to. The action execution processing means 3 15 is activated when a keyword (such as a verb) that specifies a preset action execution and the ending of a request such as “please” are detected as a result of analyzing the input sentence. I do. For example, if "Please give Mr. A f x a B" is input, the f a X program which has been set in advance is started and the document B is sent. Then, after the operation is completed, a keyword for designating the operation execution is notified to the response means 1 18.

As shown in Fig. 10, the response means 1 18 consist of a companion preparation means 4 1 1, answer preparation means 4 1 2, yes / no response means 4 1 3, question preparation means 4 1 4, and report preparation means 4 1 5. Output means 4 16

Upon receiving the notification from the registration processing means 3 1 1, the companion creation means 4 1 1 receives a response “yes, I understand” if it is “new” and “listen” if it is “o 1 d”. Has been sent to output means 4 16.

The answer creating means 4 12 applies the answer sent from the search processing means to a fixed format of “It is (answer)” to create a response sentence.

The yes / no response creation means 4 13 sets the response sentence to “Yes” if the notification from the confirmation processing means is “yes”, and “No” if the notification is “no”.

The question creation means 4 14 applies the terms, such as the subject indicating the insufficiency and the place of departure or departure, sent from the question processing means to the fixed phrase "What is (term)?" Sentence).

The report creation means 4 15 receives the keyword notified from the action execution processing means. A report (response sentence) is applied to the fixed sentence "(keyword)."

The output means 4 16 inputs the prepared response sentence to a speech synthesizer (not shown), and utters the answer as a voice from a speaker (not shown) to the user and answers. Of course, the response sentence may be displayed on a monitor or the like.

As is clear from the above description, according to the invention according to the first aspect of the present application, the input sentence in the user's natural language is determined, and the input sentence is made to correspond by different kinds of dialogue means, and the correspondence as needed by the user is It is possible to have a conversation closer to those who can. Furthermore, by setting specific business knowledge, it is possible to supplement the user's ambiguous questions and present appropriate information to the user.

As is clear from the above description, according to the invention according to the second aspect of the present application, the input sentence in the natural language of the user can be displayed in the state of the dialogue (for example, whether the input sentence is an answer to the question asked by the system). Utterance status), the input mode can be switched to understand the input according to the status judgment, and analysis can be performed by switching the input mode. Furthermore, the sentence of the input sentence is determined, and appropriate work is performed. However, it is possible to have a dialog that is closer to a human dialog that can return an appropriate response sentence according to the execution to the user.

Claims

The scope of the claims

1. In a natural language dialogue system with a convenience store with at least input means for inputting information by the user,

1) An interaction mode selection means capable of selecting at least one of at least two interaction modes having different characteristics.

2) analysis means for morphologically analyzing the input sentence and cutting out a word sequence as a whole; and

3) A switching means for determining the result of the analysis and switching the dialog mode, comprising: a natural language dialog system using a combination.

2. In the natural language dialogue system according to claim 1, wherein the dialogue mode is:

1) This mode is called a chat mode, in which a response is output only according to a preset rule base and no data is acquired (hereinafter, “chat mode”). ),

2) Judge the verb of the input sentence, obtain the case information of the verb by referring to the dictionary prepared in advance, extract the data that fits this case from the input sentence, and if there is a case that has no applicable information An interactive mode (hereinafter, “general mode”) in which a question sentence inquiring about the missing case is output to the user, the user is requested to input missing information, and the case information is acquired as data matching the item. O J \

3) Items that are set in advance according to the needs of the system administrator's business (hereinafter referred to as “business items”, information that matches the business item is obtained from the input sentence and is not allocated. If there is an item, a dialogue that outputs to the user a question sentence inquiring about the missing item, requests the user to enter missing information, and obtains data corresponding to the unassigned business item. (Hereinafter referred to as “business mode”),

A computer-based natural language dialogue system characterized by at least two selected from the above.

3. The natural language dialogue system according to claim 2 according to claim 2. hand,

The natural language dialogue system by a computer, wherein the business mode is further divided into a plurality of business modes having different business items.

4. In the natural language dialogue system using a computer according to claim 3,

By referring to a database that stores the word data extracted in the analysis of the input sentence by the user and the item to which the data belongs, the order is determined out of order to determine the business item in which the extracted data should be contained. Computer-based natural language dialogue system.

5. In the natural language dialogue system according to claim 2 described in claim 2,

A natural language dialogue system using a convenience, wherein the user accesses a database related to the business and obtains the schema based on the database as a business item.

6. In the computer-assisted natural language dialogue system according to claim 2,

In the above-mentioned business mode dialogue, even if another mode starts up while business items are not allocated, regression to store the business items that have not been allocated and query for missing items as necessary A natural language dialogue system with convenience that features features.

7. In a natural language dialogue system with a combination of at least a user and a means for inputting information,

A means for acquiring information matching the business item from the input sentence by referring to the business item set in advance according to the needs of the system administrator's business. A means for outputting a question sentence inquiring to the user, and

A business mode dialogue means including a means for acquiring data corresponding to business items that have not been allocated in response to a request for input of missing information to the user. A computer-generated natural language dialogue system characterized by including at least a component.

8. In the computer-assisted natural language dialogue system according to claim 7,

The business mode is further divided into a plurality of business modes having different business items.

9. In the natural language dialogue system according to claim 8 described in claim 8,

The order in which the cut-out data should be contained is determined by referencing the word data extracted in the analysis of the input sentence by the user and the database storing the items to which the data belongs. Computer-based natural language dialogue system that performs processing.

10. In the natural language dialogue system according to claim 7 according to claim 7,

A natural language dialogue system based on a computer, comprising accessing a database related to the business and acquiring a schema based on the database as a business item.

11. The natural language dialogue system according to claim 7, wherein:

In the above-mentioned business mode dialogue, even if another mode starts up while business items are not allocated, regression to store the business items that have not been allocated and query for missing items as necessary A natural language dialogue system with features that has features.

1 2. In a computer-based natural language dialogue system that has at least input means for the user to input information,

If business dialogue is not performed in a section that satisfies at least one condition selected from the lapse of a preset time and the preset number of dialogues, at least a component that guides the user to ask a preset question Include as This is a natural language dialogue system based on a convenience store.

13 3. In a natural language dialogue method between a computer and a user having at least an input means for inputting information by the user,

1) An interaction mode selection step of selecting at least one of at least two interaction modes having different properties.

2) a morphological analysis of the input sentence and an analysis step of cutting out a word string as data; and

3) a switching step of determining the result of the analysis and switching the interaction mode to a predetermined mode;

A natural language dialogue method between a user and a convenience store, characterized by including:

14. The natural language dialogue method between a computer and a user according to claim 13, wherein the dialogue mode is:

2) Judge the verb of the input sentence, obtain the case information of the verb by referring to the dictionary prepared in advance, extract the data that fits this case from the input sentence, and if there is a case that has no applicable information An interactive mode (hereinafter, “general mode”) in which a question sentence inquiring about the missing case is output to the user, the user is requested to input missing information, and the case information is acquired as data matching the item. ").),

A natural language dialogue method between a user and a combination, characterized in that there are at least two selected from the above.

15. The method of claim 14, wherein a natural language dialogue between the computer and the user is provided.

The business mode is further divided into a plurality of business modes having different business items, and a natural language dialogue method between a user and a console.

16. The natural language dialogue between the user and the user according to claim 15, further comprising:

An out-of-order process of determining a business item in which the extracted data should be stored by referring to a database storing the words extracted in the analysis of the input sentence by the user and the items to which the data belongs. A method of dialogue between a computer and a user in a natural language, characterized by including a step of performing.

17. The natural language dialogue method between a user and a user according to claim 14, further comprising:

A natural language dialogue method between a computer and a user, comprising a step of accessing a database related to the business and acquiring a schema of the database as a business item.

18. The natural language dialogue between a computer and a user according to claim 14, further comprising:

Even if another mode is started up while the business item is not allocated in the business mode dialogue, a regression step is provided for storing the business item not allocated and inquiring for a missing item as necessary. A natural language dialogue method between a computer and a user featuring octopus.

1 9. In a natural language dialogue method between a user and a user who has at least a means for inputting information by a user,

A step of obtaining information matching the business item from the input sentence by referring to the business item set in advance as required by the system administrator's business; and Outputting to the user a question sentence asking for

Unassigned for business due to request of user to enter missing information A method of interacting with a user in a natural language, comprising at least as a component a business mode dialogue including a step of acquiring data corresponding to an item.

20. The natural language dialogue method between a user and a user according to claim 19, wherein:

The business mode is further divided into a plurality of business modes having different business items, and a natural language dialogue method between a user and a user.

21. The natural language dialogue method between a computer and a user according to claim 19, further comprising:

By referring to the database storing the word data extracted in the analysis of the input sentence by the user and the item to which the data belongs, the order is determined out of order to determine the business item in which the extracted data should fit. A natural language dialogue method between a computer and a user, characterized by including steps.

22. The method according to claim 19, further comprising:

A method of accessing a database related to the business and acquiring a schema of the database as a business item, and a natural language dialogue method between the user and the computer.

23. In the natural language dialogue method between a combination user and a user according to claim 19,

24. In a natural language dialogue method between a user and a convenience store having at least input means for inputting information by a user,

If a business dialogue is not performed in a section that satisfies at least one condition selected from a predetermined time elapsed and a preset number of dialogues, A natural language dialogue method between a computer and a user, characterized by including at least a task guidance process for asking a set question as a component.

25. A computer-readable recording medium that records a program for executing the method of natural language conversation between a computer and a user according to claims 13 to 24.

2 6. In a natural language dialogue system with a conversation

2) word decomposition means for morphologically analyzing the input sentence,

3) a speech analysis means for synthesizing the result of the morphological analysis and generating a list of words (hereinafter, referred to as “syntax expression data”) to which a label indicating a syntactic case is attached;

5) Judgment of the input sentence, registration processing of storing the syntax expression data as knowledge in a database, search processing of searching for knowledge already stored in the database, lack of syntactically incomplete input sentence Input sentence determination means for selecting at least one input sentence process selected from a reflection process for prompting a user to input a function, and a function execution process for activating a preset function;

A computer-based natural language dialogue system characterized by the following.

27. In the natural language dialogue system according to claim 26 described in claim 26,

The input mode includes at least two modes: a normal mode and an answer mode when an answer to the question sentence output by the system is input. This is a natural language dialogue system using a combination evening.

28. In the natural language dialogue system according to claim 26 described in claim 26,

The input sentence is at least one selected from a text sentence by a keyboard input and a sentence converted into a text via a voice recognition device, and a natural language dialogue system based on a combo system.

29. The computer-implemented natural language dialogue system according to claim 26, wherein the status of the dialogue is determined by whether or not the system has output a question sentence to a partner of the dialogue immediately before. This is a natural language dialogue system based on convenience.

30. In a natural language dialogue method between a user and a convenience that has at least an input means for inputting information by a user,

2) Analyze the input sentence according to the mode and reconstruct it into a list of data consisting of words to which the analysis results are labeled (hereinafter referred to as “knowledge data”). The process of understanding the sentence,

3) analyzing the knowledge-based data, transmitting knowledge, confirming knowledge, searching for knowledge, and determining at least one sentence selected from a request for a preset function operation;

4) performing an operation according to the result of the determination; and

5) generating a response sentence according to the execution of the work, outputting and replying,

A natural language dialogue method between a computer and a user, characterized by including:

31. A method according to claim 30, wherein the computer and the user interact with each other in a natural language.

The input mode includes at least two modes: a normal mode and an answer mode when an answer to the question sentence output by the system is input. A natural language dialogue method between a computer and a user.

32. The method of claim 30, wherein the computer and the user interact with each other in a natural language.

The method according to claim 1, wherein the input sentence is at least one selected from a text sentence by a keyboard input and a sentence converted into a text via a voice recognition device.

33. The natural language dialogue method between a convenience and a user according to claim 30.

A natural language dialogue method between a user and a convenience store, wherein the status of the dialogue is determined based on whether or not the system outputs a question sentence to a partner of the dialogue immediately before.

34. In the natural language dialogue system according to claim 30 described in claim 30,

The labeling is performed by referring to at least one of a dictionary created in advance and a result of parsing an input sentence, and a natural language dialogue between the user and the computer. Method.

35. A recording medium readable by a convenience recording medium that records a program for executing the natural language conversation method between the convenience and the user according to claim 30 to 34.