WO2000054180A1 - System and method for creating formatted document on the basis of conversational speech recognition - Google Patents

Abstract

A formatted document creating system used for recording conversation comprises a speech recognition device, an information extracting section for extracting information to be recorded according to the results of speech recognition, a conversation record buffer for recording a sentence including the extracted information, a case database for storing therein a case which may occur as the content of conversation, a topicality control section for deducing an undefined information on the basis of one item of information included in the sentence recorded in the conversation record buffer section by referencing cases stored in the case database, a multimodal interaction control section for inquiring the undefined information of the document creator to define the undefined information, an input/output device control section for controlling the communication with the document creator, an input device for inputting the intention of the document creator into a computer system, and a conversation record archive for storing therein the record of the defined conversation. The system can be applied to conversation on the phone.

Description

 Specification

Typical document creation system and method based on speech recognition of conversation

Technical field

 The present invention generally relates to the documentation of conversations such as routine consultations, and more particularly to, for example, a routine document creation system and a routine document creation method for writing a dialogue via a telephone in real time. .

 Back

 If the contents of conversations such as consultations are documented, it will be useful for managing user information and consultation contents. In the past, recording conversations has relied on simply leaving a recording or making notes manually. The former method replays a recording and transcribes it manually, but it has the disadvantage that it requires a great deal of time and labor to document it. On the other hand, the latter method has a drawback that it is not possible to concentrate on the conversation due to the need to take notes, and the speed of taking notes is also limited. Therefore, instead of simply recording a recording of the conversation, it is conceivable to convert it into a sentence using the speech recognition function of a computer. Examples of such a device include Japanese Patent Application Laid-Open No. Hei 6-255307 and Japanese Patent Application Laid-Open No. Hei 5-169925. However, there is no proposal that stipulates a method for interactively creating a standardized record using the information in the transcription result by speech recognition.

In addition, the accuracy of the transcript described above is often not sufficient for practical use. However, transcripts from speech recognition contain the necessary information (information to be recorded) to create a routine record, even if it is not sufficient to document it as is. For example, in a conversation between a physician and a patient over the telephone, information about the physical condition, symptoms, and time is the content of the conversation that should be recorded because it may be a consultation about illness or injury. Consultations at counters in various service industries, such as counters at travel agencies, include the purpose of travel, number of people, timing, location, cost, and schedule. In such cases, the content of the spoken conversation is somewhat limited, and there is often a close relationship between one piece of information and the other information that follows it. On the other hand, with the advance of the graphic user interface technology, it has become common to display and input information to a computer by a combination of a graphic display device and a pointing device. In addition, the development of speech recognition and synthesis technologies has made it possible to send and receive information between computer systems and operators by voice.

 Therefore, an object of the present invention is to provide a standardized document creation system and a standardized document creation method for efficiently documenting the contents of a conversation such as a regular consultation.

DISCLOSURE OF THE INVENTION

In the present invention, as a means for solving the above-mentioned object, a standardized record is created interactively using information in a transcription result by voice recognition.

 More specifically, the present invention relates to a typical document creation system used when recording a conversation, a speech recognition device for recognizing speech of a conversation, and information for extracting information to be recorded from a speech recognition result. An extraction unit, a conversation record buffer that records sentences containing the extracted information as conversation contents, a case database that includes a plurality of pieces of information and stores cases that may occur as conversation contents, and a case. A topicality control unit that estimates other unconfirmed information from one sentence information recorded in the conversation recording buffer with reference to the cases stored in the database, and an unconfirmed recorded in the conversation recording buffer Communication between the multi-modal dialogue control unit that inquires the document creator of the information to determine the undecided information and the document creator using graphics , A GUI control unit for controlling the operation of the document, an input device for inputting the intention of the document creator to the computer system, and a conversation recording device for storing a record of the confirmed conversation. To provide a strategic document creation system.

In the case database, cases that can occur as the content of a conversation and that include multiple pieces of information are stored. In the present invention, the cases that can occur as the contents of a conversation may vary greatly depending on the situation in which they are used. However, it is expected that the content of the conversation will be limited to some extent in individual usage situations. Then, in each saved case, there are multiple pieces of information (information to be recorded) that are important in recording the conversation. This information and its relationship are used to determine individual information in real conversations. This allows you to record conversations When creating a document, it is possible to present information that is a correct answer or information that has a high probability of being a correct answer as a candidate for information constituting a documented conversation record.

 Such a conversation record is recorded in a conversation record buffer. The conversation record including the undetermined information is presented by the GUI control unit on the GUI displayed on the computer monitor of the operator. The work up to this point is created in real time by a computer during a conversation. Regarding inquiries to the document creator about the indeterminate information recorded in the conversation recording buffer, the document creator inputs his intention to the computer system using various input devices. In other words, if the undecided information is correct, the decision is made as it is. If the information is incorrect, another candidate that is a correct answer is selected and decided, or the correct answer is entered by itself and decided. In most cases, the task is to click the undecided information, which is the correct answer, to confirm the task, so that a conversation record containing all necessary information can be created in a short time without interrupting the flow of the conversation. To The confirmed conversation record is recorded in the conversation record archive and organized and stored so that it can be used at any time.

 According to a second aspect of the present invention, in the fixed form document creation system according to the first aspect, the input device is any one of a speech synthesizer, a sound generator, a character input device, a pointing device, and a graphic display device. Or a combination thereof.

 The input device may be of any type as long as it can input the operator's intention into the computer system, but a voice synthesizer, a sound generator, etc., from the viewpoint of operability and accuracy. , A character input device, a pointing device, a graphic display device, or a combination thereof. According to a third aspect of the present invention, in the fixed form document creation system according to the first aspect, the topicality control unit stores information determined by the document creator in the conversation recording buffer as an example in the case database. It is characterized by having such a configuration. The learning effect obtained from the work to determine the unconfirmed information by the document creator will be effectively used for creating conversation records that will be performed thereafter.

According to a fourth aspect of the present invention, in the fixed form document creation system according to the first aspect, when information extraction for certain information is probable, the information extraction unit includes It is characterized in that it is configured to write a plurality of candidates relating to the contents together with likelihood information based on statistics and the like in a conversation recording buffer.

 When the information extraction unit extracts information to be recorded from the speech recognition result, if it can identify one piece of information with a probability of 100%, it is written as determined information. In other cases, that is, when it is probable that information extraction for a certain information is probable, a plurality of candidates related to the information content are written to the conversation recording buffer together with the likelihood information based on statistics or the like. To The document creator refers to this likelihood information when determining unconfirmed information.

 According to a fifth aspect of the present invention, in the fixed form document creation system according to the first aspect, when the estimation of the content to be input to the information of the conversation record buffer is not uniquely determined, the topicality control unit may be configured to use the topicality control unit. It is characterized in that it is configured to write a plurality of candidates in a conversation recording buffer together with the likelihood information based on statistics or the like.

 As described above, the topicality control unit estimates other undetermined information from one piece of information of a sentence recorded in the conversation recording buffer. If the content to be input to the information in the conversation recording buffer cannot be uniquely determined, the topicality control unit writes a plurality of candidates to the conversation recording buffer together with likelihood information based on statistics. As described above, by using the likelihood information based on statistics or the like by the speech control unit, the information to be selected by the user from the information in the transcription result by the speech recognition and the information input or determined by the user. Order or narrow down the content. This streamlines the creation of descriptions of routine information contents.

For such unconfirmed information recorded in the conversation record buffer, the document creator is inquired to determine the unconfirmed information. In response to an inquiry to the document creator about undecided information, the document creator enters his / her intention into the computer system using various input devices. In other words, if there is a correct answer in the undecided information, it is selected and confirmed, and if the correct answer does not exist in the candidates, the user himself inputs the correct answer and decides. The document creator refers to this likelihood information when performing these operations. According to a sixth aspect of the present invention, in the fixed form document creating system according to the first aspect, the multi-modal dialogue control unit creates a document by voice, GUI, or character input. It is characterized in that it is configured to determine the information of the sentence recorded in the conversation recording buffer by inquiring of the creator.

 The sentence information is determined using a means that is easy for the operator to use. Operability is improved. According to a seventh aspect of the present invention, in the fixed form document creating system according to the fifth aspect, the multi-modal dialogue control unit sequentially searches the information to be queried in the conversation record buffer from candidates having a high likelihood. It is configured to make an inquiry to the document creator in advance.

 By inquiring the document creator in order from the candidates having the highest possibility of being the correct answer, the time to reach the information that is the correct answer is shortened.

 The present invention according to claim 8 is the fixed form document creation system according to any one of claims 1 to 7, further comprising: a telephone that enables conversation between remote locations; and A caller telephone number obtaining device for obtaining a number and recording the number in a conversation recording buffer.

 The voice recognition device can also recognize voice of a conversation through a telephone. The voice on the transmitting side and / or the receiving side is recognized, and in response to this, the information extracting unit extracts information to be recorded from the voice recognition result. The caller telephone number acquisition device acquires the other party's telephone number and records it in the conversation recording buffer. Various information about the speaker of the other terminal of the telephone can be obtained from the other party's telephone number, and it is easy to guess other undetermined information from one piece of text recorded in the conversation recording buffer.

According to a second aspect of the present invention, there is provided a standard document creation method used when recording a conversation, comprising: a voice recognition step of recognizing a voice of the conversation; and an information extraction step of extracting information to be recorded from the voice recognition result. And a writing step of writing a sentence containing the extracted information into the conversation recording buffer as the contents of the conversation, and a case of storing in the case database a case that can be generated as the contents of the conversation and that includes a plurality of pieces of information. With reference to the storage step and the cases stored in the case database, guess about other unconfirmed information from one piece of sentence information recorded in the conversation recording buffer, and set it in the conversation recording buffer. The guess information set switch and the sentence containing the unconfirmed information recorded in the conversation record buffer are presented to the document creator graphically, and the unconfirmed information is displayed. Information inquiry update Sutetsu to confirm the information content, contact the Bungai author with And a conversation recording step of saving a record of the determined conversation as a sentence.

 According to a tenth aspect of the present invention, in the fixed form document creating method according to the ninth aspect, the information inquiry updating step includes: a speech synthesizer, a sound generator, a character input device, a pointing device, a graphic display device. It is performed using an input device configured to include any one of the above or a combination thereof.

 The present invention according to claim 11 is the method according to claim 9, wherein the information determined by the document creator in the conversation recording buffer is additionally stored in the database as a case. Characterized in that it is configured to include

 According to the present invention as set forth in claim 12, in the fixed form document creating method according to claim 9, when the information extraction for certain information is probable, a plurality of candidates for the information content are provided. Is written in the conversation recording buffer together with likelihood information based on statistics and the like.

 According to the present invention described in claim 13, in the fixed form document creation system according to claim 9, the guess information setting step is such that the estimation of the content to be input to the information of the conversation record buffer is performed uniquely. If it is not determined, it is characterized in that it is configured to write multiple candidates to the conversation recording buffer together with likelihood information based on statistics and the like.

 The present invention described in claim 14 is the fixed form document creation method according to claim 9, wherein the information inquiry update step includes the step of inquiring the document creator by voice, GUI, or character input and in the conversation recording buffer. It is characterized in that it is configured to determine the information of the sentence recorded in.

 According to a fifteenth aspect of the present invention, in the fixed form document creating method according to the thirteenth aspect, the guess information setting step is performed from candidates having a high likelihood with respect to information to be queried in the conversation recording buffer. It is characterized in that it is configured to inquire the document creator in order.

The present invention according to claim 16 is a method according to any one of claims 9 to 15, wherein the speech recognition step recognizes a conversation through a telephone. And a caller telephone number obtaining step of obtaining the other party's telephone number and recording the same in a conversation recording buffer.

 According to the present invention described in claim 17, in the fixed form document creation method according to claim 16, the guess information setting step is stored in the case database from the other party's telephone number recorded in the conversation recording buffer. With reference to the case, it is characterized in that it is configured to infer other undetermined information from one piece of information of the sentence recorded in the conversation recording buffer.

Simple B month of drawing

 FIG. 1 is a block diagram of an embodiment of a typical document creation system according to the present invention.

 FIG. 2 shows an embodiment of a typical document creation system in a conversation using a telephone according to the present invention.

It is a block diagram of a form.

 FIG. 3 is an explanatory diagram showing the contents of the conversation between the doctor and the patient. FIG. 4 is an explanatory diagram showing the contents of the conversation record.

Best shape bear for carrying out the invention

 Hereinafter, the present invention will be described in detail with reference to the illustrated preferred embodiments.

 FIG. 1 is a block diagram showing an overall configuration of an embodiment of a typical document creation system based on speech recognition of a conversation according to the present invention.

 As shown in FIG. 1, the typical document creation system based on speech recognition of a conversation according to the present embodiment includes a microphone 1, a speech recognition device 2, an information extraction unit 3, a topic control unit 4, a database 5, , A multi-modal dialogue control unit 6, a conversation recording buffer 7, a speech synthesizer 8, a sound generator 9, a character input device 10, a GUI control unit 11, a pointing device 12, a graphic It comprises a display device 13 and a conversation record archive 14.

The speech recognition device 2 interprets the speech signal from the microphone 1 as linguistic speech, and outputs a character string or a morphological analysis result (text that is decomposed for each word and to which information such as part of speech is added). In this case, if the speech signal can be identified as one piece of information with a probability of approximately 100%, the speech recognition device 2 converts the speech signal into text as determined information. Is written to the conversation record buffer. In other cases, that is, when it is probable that information extraction for certain information is possible, write multiple candidates for the information content based on statistics etc. To When deciding unconfirmed information, the document creator will be able to make more appropriate decisions by referring to this likelihood information.

 In addition, if a microphone or a communication line is inserted between the input to the speech recognition device 2 and the microphone mouth phone 1, it is not always necessary to perform the speech recognition at the time or place where the conversation takes place.

 Also, when the multimodal interaction control unit 6 collects information interactively from the document creator, the speech recognition device 2 converts the information provided by the document creator by voice into a character string, etc. Send to dialogue control unit 6.

 The information extraction unit 3 receives the transcription mode output of the speech recognition device 2 as input, extracts information to be recorded, and writes the information to the conversation recording buffer 7. The information to be extracted, that is, the information to be recorded, is selected in advance according to the type of conversation that is assumed. For example, if it is intended to be used at a travel agency counter, conversations including the purpose of the trip, number of people, name, contact place, itinerary, destination, fee, accommodation, payment method, etc. are recorded. The information should be left as

 Taking the “purpose of travel” as an example, conversations can include honeymoons, company comfort meetings, graduation trips, individual trips, group trips, business trips, business trips, inspection trips, parent and child trips, summer vacation, sea bathing, skiing, golfing If it contains words (information) such as, scuba diving, watching a theater, watching a game, or participating in an event, the information extraction unit 3 extracts this as information J to be recorded and records it in a conversation recording buffer. Record in number 7. If the number is "number of people", a combination of a word that represents a number and "nin", or a word that means two people such as "married couple" or "honeymoon" is included. The information extraction unit 3 extracts this as “information to be recorded” and records it in the conversation recording buffer 7.

The topicality control unit 4 estimates contents to be input as other information while referring to the database 5 from one piece of information shown in the conversation record in the conversation record buffer 7, and harms the conversation record buffer 7. I'm sorry. At this time, if the estimation is not uniquely determined, a plurality of candidates for the information are written into the conversation recording buffer 7 together with the likelihood value. The likelihood value is Based on the information described in the database 5, it is calculated by, for example, Bayesian method.

 The case database 5 stores a case for estimating the content of other information from one information given to the conversation record buffer 7. Actually, for the information determined by the document creator in the conversation recording buffer 7, the relationship between the contents of each information is stored as an example. In the example of the travel agency counter mentioned above, for one piece of information such as “honeymoon”, “number of people” is 2 people, “travel itinerary” is 3 days to 2 weeks, and “destination” is Overseas travel is major, Hawaii, the Mediterranean, the west coast of the United States, "airlines" use business classes with Japanese airlines such as JAL or ANA, and "expenses" are somewhat high. Many cases have been saved, such as staying in a room and paying in advance with cash.

 In such a situation, despite the fact that “Honeymoon” is recorded as one piece of information in the conversation record buffer 7, it is recognized that “number of people” is three and “travel destination” is Atami. Assume that In this case, since “honeymoon” and “number” contradict each other, the multimodal dialogue control unit 6 considers “two” as the first candidate and “three” as the second candidate. Display with the degree information and contact the document creator. In this case, since the number of persons was “2” in all cases, the likelihood of the former is 1.0000 and the likelihood of the latter is 0.0000. After confirming with the customer, the document creator clicks and confirms if “two” is correct, and selects and confirms if “three” is correct. For example, if one of the couples has a child, even if they are newlywed, it may be possible to go on a “honeymoon” with three people.

Or, the recognition of “honeymoon” is actually wrong, and the customer shortens the “new employee training trip” and utters “shinken trip” in a conversation, and the voice recognition device 2 mistakenly recognizes this as “honeymoon travel”. Such a case may occur. In this case, for example, the document creator corrects the part of “Honeymoon” to “Shinken Travel” using an input device 10 such as a keyboard, and then determines the part. A “new employee training trip” does not contradict that the travel destination is Atami. In the initial stage of use of the system, there is no accumulation of cases by learning, so the assumed interrelationship between information is written by the designer.

 The topicality control unit 4 stores the information determined by the document creator in the conversation record / buffer as a case in the database 5 in order to provide the likelihood information to the database 5. The multi-modal dialog control unit 6 reads the conversation recording buffer 7 and collects information whose contents have not been determined by inquiring of the document creator. At this time, if it is judged that the spoken dialogue is appropriate, a spoken inquiry is made. Otherwise, the information is collected by the GUI or the character input device 10.

 The multi-modal dialogue controller 6 presents the contents of the conversation recording buffer 7 to the document creator via the graphic display device 13 or the speech synthesizer 8 and the sound generator 9 at the request of the document creator. can do. In particular, the multimodal dialogue control unit 6 can present the transcription result of the speech recognition device 2 held in the conversation record buffer 7 to the graphic display device 13 at the request of the document creator. When collecting information interactively from a document creator, the multi-modal dialogue control unit 6 sends the options to the speech recognition device 2 in advance if the user's utterance options are limited. The accuracy of recognition can be improved. For example, when a customer wants to travel to Oki Ref, the place name of Oki Ref and its reading are transmitted to the speech recognition device 2 in advance. By doing so, the probability of correctly recognizing the accommodation place as “Ogimi-son, Kunigami-gun” even when the customer utters “Nigamigun Ogimimura” is raised. The conversation record buffer 7 holds the contents of each information of the typical dialog. If the content is not finalized (in the case of unconfirmed information), a candidate list is given for the content. Each candidate may be given a likelihood. The conversation recording buffer 7 holds the transcription result of the voice recognition device 2.

 The speech synthesizer 8 can convert a message from the multi-modal interaction control unit 6 to the document creator from text to a speech signal.

 The sound generator 9 converts the sound signal from the sound synthesizer 8 into sound.

 The character input device 10 such as a keyboard converts a type input from a document creator into a signal to the multi-modal interaction control unit 6.

The GUI control unit 11 receives the option input request from the multi-modal dialog control unit 6. The information is displayed on the graphic display device 13, and the selection by the pointing device 12 of the document creator is transmitted to the multi-modal interaction control unit 6.

 The pointing device 12 transmits the selection of the document creator (for example, a mouse click at a certain position) to the GUI control unit 11.

 The graphic display device 13 <displays the contents of the conversation recording buffer 7 based on an input request from the multi-modal dialog control unit 6 and a request from the document creator.

 The conversation record archive 14 stores the contents of the conversation record buffer 7 at the request of the document creator.

 As described above, according to the present invention, it is possible to interactively create a fixed record by using information in a transcription result by voice recognition. The present invention uses the likelihood information based on statistics and the like to order or narrow down the information to be selected by the user, and to streamline the creation of descriptions of typical information. FIG. 2 is a block diagram of an embodiment of a typical document creation system in a conversation using a telephone according to the present invention.

 Telephone conversation and consultation continue to be an important means of communication even in the modern times of the Internet. For example, it is difficult for an elderly person who does not stay out to learn to operate a computer, and often lacks the financial resources to buy the necessary equipment. Therefore, it is expected that medical consultations will be conducted using telephones in the future. In this case, it is natural for the physician to have a proper telephone conversation, and it is necessary to keep a record of the consultation, the prescription of the medicine, and other various instructions. For face-to-face treatment, it does not matter if the doctor fills in the medical record as needed and takes the time. However, in medical treatment through the telephone, if the conversation is interrupted for a long time, patients tend to be anxious. Therefore, there has been a demand for a system that can record conversations in real time.

This embodiment responds to such a demand. Basically, a telephone 15 is installed in place of the microphone 1 in the system shown in FIG. 1 described above, and the telephone number of the other party is obtained and the conversation record buffer 7 is obtained. A caller telephone number acquisition device 16 to be recorded in the system is added. In the present embodiment, the voice recognition device 2 can be configured to be able to switch so as to recognize only the transmission side, only the reception side, or both of the transmission and reception voices in the telephone conversation. In general, the smaller the number of sound sources, the higher the recognition rate of voice recognition. Therefore, even if it takes some time, the doctor can repeat the patient's remarks and leave only the doctor's remarks as a conversation record. In addition, in a usage situation where it is not necessary to make a conversation record in real time and only the content of the conversation of the other party includes information to be recorded, it is also possible to configure so that only the speech content of the other party is voice-recognized. .

 The caller telephone number acquisition device 16 acquires the other party's telephone number and records it in the conversation record buffer 7. From the other party's telephone number, various information about the speaker of the other terminal of the telephone can be obtained. For example, it is possible to obtain the name, gender, age, presence / absence of health insurance card, illness record, treatment history, etc., and to easily guess other undecided information from one piece of text recorded in the conversation record buffer. I do. Of course, it is also possible to retrieve the other party's telephone number during a conversation and use it to obtain the same information from a database stored in the document creator's computer.

 (Example 1)

 Next, a typical document creation system that assumes telephone medical care will be described. First, in the case database 5, the interrelationship between the assumed information is written by the designer. This is because there is no accumulation of cases by learning in the early stage of using the system.

 The information to be recorded is "body temperature", "blood pressure", "worry", "symptoms", "what time", "pain", "nausea", "chill", "cough", "breath" , "Diarrhea", "where", "how", "who", "what", "insurance card", "payment", "diagnosis" and so on. When the speech recognition device 2 outputs words relating to these, the information extracting unit 3 records the words to be recorded in the conversation recording buffer 7.

On the other hand, the case database 5 shows, for example, as a case related to a cold, “The fever is close to 40 degrees and coughing and sneezing occur.” J, “There is no fever, but the joints of the limbs are painful. , “I feel cold and my runny nose does not stop.”, “I have a headache, coughing and sometimes yellow sputum.”, “The throat is irritating. No. And so on. In the case of a cold, there is no association with “blood pressure”, and the association with various physical abnormalities that appear as “symptoms” increases as the likelihood. Similarly, for other diseases, various cases containing multiple pieces of information to be recorded are accumulated.

 This system will be used to record telephone consultations between doctors and patients. Figure 3 shows the conversation between the doctor and the patient. In this embodiment, it is assumed that both telephone transmission and reception are documented and recorded.

 The voice recognition device 2 converts all the transmission / reception signals of the telephone 15 into voice signals and outputs them to the information extracting unit 3. The words described above are registered in the memory of the information extraction unit 3 as information to be recorded, and in the transcribed sentences by the speech recognition device 2 of A1 to A9, the words should be recorded in A3 to A8. Information is included. On the other hand, the transcribed sentence of the voice recognition device 2 for B1 to B9 includes information that should be recorded in B2 to B4 and B6. Information that should be recorded is underlined in the table.

 The information extraction unit 3 writes the transcription result including such information in the conversation recording buffer 7 as "What happened. I have heat and I feel cold. It is almost 40 degrees when measured with a thermometer." "Do you have any other noticeable symptoms. Your joints are painful and you feel irritated.", "How about ¾, strength, sneezing. No, but the back of my lungs is a bit painful. "How about a runny nose?"

"When did the fever start? From yesterday's 晚. My joint began to hurt only this morning." "I have a cold." Please drink a commercially available antipyretic agent. " Here, in the transcribed sentence by the speech recognition device 2, portions that are different from the actual conversation content are bolded and underlined for easy understanding.

The topicality control unit 4 selects “throat” as a candidate for “i” from “throat is throbbing and coughing makes the back of the chest a little painful” in the case recorded in the case database 5. Replace and update the record in the conversation record buffer 7. In addition, in the case “Ripe near, jaundice, <sneezing.”, “Cough” is selected and replaced as a candidate for “&”, and the record in the conversation record buffer 7 is updated. Under the control of the GUI control unit 11, the multi-modal dialogue control unit 6 displays "What's going on?" And "It has heat and chills. It is close to 40 degrees when measured with a thermometer." , "Do you have any other noticeable symptoms?", "I have a pain in my joints and my throat is rough.""What about coughing or sneezing?", "No. The back is a bit painful. "," How about a runny nose? "," Not out. "," When did the fever come out? "It's only this morning. "," I have a cold. ", And" Warm up and take a rest. Please drink a commercially available antipyretic. " Of course, instead of GU I, a voice synthesizer 8 and a sound generator 9 can be used to inquire the document creator by voice.

 The document creator uses the pointing device 12 which is shared with the character input device 10 and the graphic display device 13 such as a monitor to select and confirm the correct information for the inquired information. In the illustrated preferred embodiment, the number of candidates for the undetermined information candidate is one. However, when the recognition accuracy in the voice recognition device 2 is low or when there are a plurality of candidates with similar names, the number of candidates is two or more. Become. In this case, it is preferable to add likelihood information to each candidate, as described above.

 The multi-modal dialogue control unit 6 further rewrites the determined conversation record in the conversation record buffer 7 as a fixed sentence containing each information and records it. In this example, "The fever is 40 degrees and I feel chills.", "The joints are painful, my throat is irritated.", "The back of my lungs is a little painful." , "The joint began to hurt this morning.", "I have a cold.", "Wake up and take a day off.", "Please take antipyretic." Is recorded as the confirmed conversation record. Then, this conversation record is recorded in the conversation record archive 14 and also stored in the case database 5 as a case (FIG. 4) including a plurality of pieces of information to be recorded. Here, the information in the patient remarks column is obtained by the caller's telephone number acquisition device 16 acquiring the other party's telephone number when the patient calls the doctor, but of course, the patient himself Can also be heard from.

Claims

The scope of the claims  1. A standard document creation system used to record conversations,  A voice recognition device for recognizing the voice of the conversation;  An information extraction unit for extracting information to be recorded from the speech recognition result,  A conversation record buffer for recording sentences containing the extracted information as conversation contents, a case database composed of a plurality of pieces of information and storing cases that can occur as conversation contents,  A topicality control unit for estimating other unconfirmed information from one piece of sentence information recorded in the conversation record buffer with reference to the cases stored in the case database, and a conversation record A multimodal dialogue control unit for inquiring the document creator of the determined information to determine the undetermined information;  An input / output device controller that controls communication with the document creator using graphics;  An input device for inputting the author's will into the computer system; and a conversation record archive for storing a record of the confirmed conversation; A standard document creation system configured to include:  2. The fixed form document creation system according to claim 1, wherein the input / output device is controlled by a DUI, and the input device includes any one of a character input device and a pointing device or a combination thereof. In addition, a standard document creation system in which the output device includes any one of a speech synthesis device, a sound generation device, and a graphic display device, or a combination thereof.  3. The fixed form document creation system according to claim 1, wherein the topicality control unit is configured to store, in the conversation record buffer, information determined by a document creator as an example in the database. A standardized document creation system. 4. In the fixed form document creation system according to claim 1, when the information extraction for a certain information is probable, the information extraction unit determines a plurality of candidates for the information content based on statistics or the like. <A typical document creation system configured to be written in the conversation recording buffer together with likelihood information. 5. The fixed-form document creation system according to claim 1, wherein the topicality control unit determines a plurality of candidates as statistics if the estimation of the content to be input to the information in the conversation recording buffer is not uniquely determined. <A typical document creation system configured to write in the conversation record buffer together with the likelihood information.  6. The fixed form document creation system according to claim 1, wherein the multi-modal conversation control unit inquires of a document creator by voice, GUI, or character input to obtain information on a sentence recorded in the conversation record buffer. A fixed form document creation system configured to determine the  7. The fixed form document creation system according to claim 5, wherein the multi-modal conversation control unit inquires of a document creator in order from information having a high likelihood with respect to information to be inquired in the conversation recording buffer. A standard document creation system configured as follows.  8. The fixed form document creation system according to any one of claims 1 to 7, further comprising: a telephone that enables a conversation between distant places; and acquiring the other party's telephone number to perform the conversation. A fixed form document creation system comprising: a caller telephone number obtaining device that records the caller's telephone number in a recording buffer.  9. A standard document preparation method used when recording a conversation,  A voice recognition step for recognizing the voice of the conversation;  An information extraction step of extracting information to be recorded from the speech recognition result;  A writing step of writing a sentence containing the extracted information as a conversation content in a conversation recording buffer;  A case storing step of storing in the case database a case that can occur as the content of the conversation, which includes multiple pieces of information;  Referring to the cases stored in the case database, guessing other unconfirmed information from one piece of information of the text recorded in the conversation recording buffer, and setting the guess information in the conversation recording buffer. When, An information inquiry updating step of presenting the sentence containing the unconfirmed information recorded in the conversation recording buffer to the harm creator in a graphic form, inquiring the document creator about the unconfirmed information and confirming the information content, , A conversation recording step for storing a record of the confirmed conversation as a sentence; and a standard document creation method comprising:  10. The fixed form document creation method according to claim 9, wherein the information inquiry updating step is any of a speech synthesis device, a sound generation device, a character input device, a pointing device, and a graphic display device. A standard document creation method to be performed using an input device configured to include one or a combination thereof. 11. The fixed form document creation method according to claim 9, further comprising a case addition step for additionally storing information determined by the document creator in the conversation record buffer as a case in the database. A standardized document creation method composed of:  12. The fixed form document creation method according to claim 9, wherein, when information extraction for certain information is probable, the writing step is based on statistics or the like based on a plurality of candidates for the information content. A typical document creation method configured to be written in the conversation recording buffer together with likelihood information.  13. The fixed-form document creation system according to claim 9, wherein the guess information set step includes a plurality of guess information sets, when the estimation of the content to be input to the information of the conversation recording buffer is not uniquely determined. A typical document creation system configured to write candidates together with likelihood information based on statistics or the like in the conversation record buffer.  14. The fixed form document creation method according to claim 9, wherein the information inquiry update step is performed by inquiring a document creator by voice, GUI, or character input to send a sentence recorded in the conversation recording buffer. A standardized document creation method that is configured to determine the information of a document.  15. The fixed sentence damage creating method according to claim 13, wherein the guess information set step is to the document creator in order from the candidate having the highest likelihood with respect to the information to be queried in the conversation recording buffer. A canonical document creation method that is configured to make inquiries. 16. The fixed form document creation method according to any one of claims 9 to 15, wherein the voice recognition step is for recognizing a conversation through a telephone, and further, obtains the other party's telephone number. And a step of obtaining a caller telephone number to be recorded in the conversation recording buffer. 17. The fixed form document creation method according to claim 16, wherein the guess information set step refers to a case stored in the case database from a partner telephone number recorded in the conversation record buffer. And a method for making a guess for other undetermined information from one piece of information of a sentence recorded in a conversation recording buffer.