JPH10143187A - Work support system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide the information, which is truly wanted by a user, in an interactive manner. SOLUTION: To provide the work support information to a user as a guide is a prerequisite for the work support system. The current condition of the user collected by condition collecting means 6 to 10 is recognized by a current condition recognition means. The user's speeches collected by a user speech collecting means 10 are recognized by the means 10 and the related speech with respect to the guiding information obtained by the user is recognized while relating the speech to the prior system speech or the user's speech. Then, based on these recognition results, the information required to support the work being conducted by the user is retrieved from databases 2 and 11, which systematically store and hold, and the retrieved necessary information is provided to a guiding section 13 by a guiding means. Since the required information necessary to support the work being conducted by the user is provided in an interactive manner in accordance with the current condition of the user, the truly necessary information is timely provided to the user.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a work support system such as a cooking guide for guiding a cooking procedure and a car navigation system for guiding an operation route.

[0002]

2. Description of the Related Art In recent years, various work support systems have been researched and developed with the progress of information processing technology and the like, and the progress and development of these systems are remarkable. As such a work support system, there is a cooking guide or the like for guiding a cooking procedure by voice, as represented by a car navigation technology which has rapidly spread with the advancement of GPS technology.

An example of a cooking guide is disclosed in, for example, Japanese Patent Application Laid-Open No. 63-201686. In this publication, cooking methods and ingredients are stored in an audio data memory,
There is disclosed an invention in which the amount of food and the cooking method for each dish are output as a sound by an advance command based on a voice input from a microphone or a key input by a push button. For example, when a certain dish is designated and guidance on the amount of ingredients is requested, the system inquires about the number of people, and when this is answered, the required amount of ingredients is output as voice. Then, when a guide for the cooking method is requested, the cooking method is output as voice.

As a car navigation system, for example, there is the following example. First, JP-A-62-2679
Patent Document No. 00 discloses an invention in which route guidance by voice is performed in addition to route guidance by image display. As a further advance, Japanese Patent Laid-Open Publication No.
Japanese Patent Application Publication No. JP-A-2002-209555 discloses an invention in which the arrival time to an intersection at which a voice announcement is made is calculated according to the speed of a vehicle, and the start timing of the guidance voice is controlled by calculating backward from the end of reading the voice announcement. I have. That is, an object of the present invention is to accurately start the voice announcement at the preceding intersection. Japanese Unexamined Patent Application Publication No. 7-27569 discloses a more accurate route guidance by changing the output timing of the guidance voice according to the distance between the intersection to be rerouted and the intersection in front of the intersection. The invention as described above is disclosed. According to the present invention, a distance between an intersection to be rerouted and an intersection in front of the intersection is determined, and when the distance is equal to or longer than a predetermined distance, guidance voice is output at a point of a predetermined distance, and the distance is equal to or shorter than a predetermined distance. In this case, a guidance voice is output after passing through the preceding intersection. As a result, the intersection before the intersection to be changed is not mistaken for the intersection to be changed.

[0005] In the car navigation speech recognition unit NVA-VRI manufactured by Sony Corporation, the system responds to user inquiries, for example, inquiries such as "current location", "next", "how long after". Voice guidance.

[0006]

If information on this work is required during any work, the person who has the information can be asked if it is nearby. It can be done very efficiently. On the other hand, if there is no person who has the necessary information nearby, it is necessary to search for an information source such as a manual, and in this case, it is difficult to perform an efficient operation. This is because there is a great gap in grasping the current situation on the information providing side unless there is a person who has the necessary information nearby. In other words, those who are close to the worker generally know whether they share the situation in which they are located and can easily understand what information the worker is requesting. Therefore, necessary information can be promptly provided. On the other hand, the information source itself such as a manual cannot naturally grasp the situation where the worker is located. For this reason, in order for workers to find necessary information from such information sources, much effort must be devoted to searching for a description part that fits their situation, which is extremely efficient. Will decrease.

[0007] Considering such a premise, a conventional work support system does not include any type of the conventional work support system which recognizes a situation where a worker is placed. For this reason, there is a problem that it is difficult to provide the information that the operator really needs.

For example, considering the inventions described in the respective publications introduced in the section of the prior art, Japanese Patent Laid-Open Publication No. 63-2016
The cooking guide described in Japanese Patent Publication No. 86 merely guides the amount of ingredients for cooking and the cooking method in response to a request from the user, and does not make any judgment on the situation where the cooking user is placed. For this reason, for example, when a situation change due to the use of a stove with a heating power different from the assumption of the system, a situation change due to a user mistake due to a mistake in listening, a situation change due to a variation in the size and quality of foodstuffs, etc. Can't respond. In addition, in actual cooking operations, there are many cases where a certain action such as turning off the fire when boiling occurs, and taking a certain action, but such a situation cannot be dealt with at all. Further, in an actual cooking operation, a plurality of operations generally proceed simultaneously, such as making a plurality of items at the same time or cutting vegetables while boiling the meat. However, since the order of the time when the contents of the pot boils and the time when the vegetables are completely cut cannot be predicted in advance, it is impossible to provide accurate guidance that can cope with such matters.

The car navigation systems disclosed in Japanese Patent Application Laid-Open Nos. 62-267900 and 7-27569 do not judge the situation where the user is located. This is the same as the cooking guide described in Japanese Patent Publication No. Therefore,
In some cases, the user may be erroneously recognized. For example, in the invention disclosed in Japanese Patent Application Laid-Open No. 7-27569, the output timing of the guidance voice is changed according to the distance between the intersection to be rerouted and the intersection in front of the intersection.
Although the idea of eliminating the user's misunderstanding is excellent, there is a possibility that a voice announcement may be made far before an intersection to be changed due to traffic congestion, and this may cause misunderstanding to the user . In other words, the invention described in Japanese Patent Application Laid-Open No. 7-27569 does not have the idea of grasping the situation where the user is located. In this regard,
Japanese Patent Laying-Open No. 4-1897 has grasped one of the situations where a user is placed, that is, the speed of a vehicle. However, as pointed out in Japanese Patent Application Laid-Open No. Hei 7-27569, there is a lack of correspondence with respect to an intersection just before an intersection to be rerouted. That is, the level of grasping the situation where the user is located has not reached the level required for the car navigation system. Therefore, it can be said that the invention described in Japanese Patent Application Laid-Open No. 4-1897 does not have the idea of grasping the situation where the user is located.

In that the system does not recognize the situation where the user is placed, the car navigation voice recognition unit NVA-VRI manufactured by Sony Corporation introduced above.
The same is true for In other words, the voice recognition unit NVA-VRI for car navigation has a structure in which voice guidance is provided in a one-answer-one-response manner in which one voice guidance is associated with one user utterance, and the context is understood. Since the voice guidance cannot be performed, there is a problem that the voice guidance deviates from the user's intention. For example, the car navigation voice recognition unit NVA-VRI is set to answer the distance to the destination set before traveling in response to the user's utterance "How long after". In the case of the utterance immediately after the voice guidance “Next is the x intersection”, it is natural to answer the distance to the x intersection.

As described above, the information that the user really wants at a certain point of time can be obtained only when the system recognizes the situation where the user is placed, and the system recognizes the situation where the user is placed. It is understandable that real-time communication between the user and the system is indispensable. In addition, it is necessary for the dialogue between the user and the system to handle a series of utterances, not a question-and-answer system. In view of the above, the present invention provides a work support system capable of grasping a situation where a user who is working is placed and providing an information guide suitable for the situation in an interactive manner in order to provide information that the user really wants. The purpose is to:

[0012]

According to a first aspect of the present invention, there is provided a work support system for performing a predetermined work in a work support system for guiding a user who is performing a predetermined work to information for supporting the work. In a work support system that guides information to a user who supports the work, the current state in which the user is located is collected from an information source other than the user's utterance, converted into an electric signal, and output. Means, current state recognizing means for recognizing a current state where the user is placed based on an output signal of the situation collecting means, user utterance collecting means for collecting user utterances, converting them into electric signals, and outputting the same, and user utterances Recognize the user's utterance based on the output signal of the collecting means, and recognize the related utterance for the guidance information performed by the user in association with the previous system utterance or the user's utterance. A user utterance recognizing means, a database for systematically storing and holding information supporting the user's work, and information necessary for supporting the work performed by the user according to the recognition results of the current state recognizing means and the user utterance recognizing means. The information processing apparatus includes an information search unit that searches from a database, and a guide unit that guides necessary information searched by the information search unit to a guide unit. Here, “related utterance” means
It means a user's utterance related to the guide information such as a user's question or return to the guide information, a request for related information, or the like.

Therefore, based on the output signal of the situation collecting means, the current state in which the user recognized by the current state recognizing means is located, and based on the output signal of the user utterance collecting means, the user utterance recognizing means sets the previous system utterance or According to the related utterance to the guide information performed by the user in association with the user's utterance, information necessary for supporting the work performed by the user is searched from the database by the information search means. You will be guided. As a result, information required by the user is retrieved in an interactive manner and guided appropriately.

Here, in the work support system according to the first aspect of the present invention, the search of the support information from the database by the user utterance recognizing means includes, for example, extracting a word contained in the immediately preceding system utterance and its synonymous word into the user utterance. It is performed as a recognition candidate (claim 2). In this case, the user utterance recognition means sets the priority of the word included in the immediately preceding system utterance higher than that of the synonymous word, and if the synonymous word includes a user utterance that has appeared previously, The priority of the synonyms is raised (claim 3).

Further, in the work support system according to the first aspect, the user utterance recognizing means estimates whether or not the user's attention target is in the current guidance information, and determines that the user's attention target has shifted from the current guidance information. When it is determined, preparations are made for a previous request for guidance information (claim 4).
As a result, guidance is performed according to the user's natural behavior pattern when the user's attention target is shifted from the current guidance information. Here, it is inferred whether or not the user's attention target has shifted from the current guidance information, for example, whether the information search unit has searched for information that is the answer to the relevant utterance of the user recognized by the user utterance recognition unit. (Claim 5) Alternatively, after a lapse of a predetermined time after the information is guided from the guidance section by the guidance means, whether or not a sufficient progress of the work can be recognized from the recognition results of the current state recognition means and the user utterance recognition means. 6). In other words, when the information search unit searches for information that is the answer to the user's related utterance recognized by the user utterance recognition unit, it is determined that the user's attention has shifted from the current guide information, and the immediately preceding user-related utterance The priority of an item that has not been targeted is raised (claim 5). In addition, after a certain period of time has passed since the information was guided from the guidance unit by the guidance means, if the progress of the work cannot be recognized from the recognition results of the current state recognition means and the user utterance recognition means, the attention target of the user is the current guidance information. , And raises the priority of an item that was not the target of the immediately preceding user-related utterance (claim 6).

In the work support system according to the first aspect, the guide means sets a priority order for the guide information to be guided by the guide unit, and guides the guide information to the guide unit sequentially according to the information request utterance of the user ( Claim 7). This allows
Guidance information with a low priority is provided only when there is an information request utterance of the user. In this case, when configured as a work support system that supports the cooking work, the guide unit, for example, guides the guide unit to the essential definite information,
The guide unit guides the reference information in response to the user's information request utterance (claim 8). Here, the “essentially definitive information” means information that is essential for supporting a cooking operation irrespective of knowledge or preference such as how to process a certain material. The “reference information” refers to reference information that may not be necessary for supporting the cooking operation. Further, when the vehicle is configured as a work support system for supporting a driving operation of a vehicle, for example, the guide unit guides the urgent necessary information to the guide unit, and additional information to the guide unit according to the information request utterance of the user. Guide (claim 9). Here, the “emergency necessary information” means essential information that is always required for supporting driving operation such as a direction and a distance. Further, “additional information” means additional information that may not be necessary for supporting driving work such as an intersection name or a landmark.

Further, in the work support system according to the first aspect, the user utterance recognizing means seeks semantic resolution of anaphora and omission included in the user utterance based on the recognition result of the current state recognizing means (claim 10). Therefore, the meaning of anaphor and omission included in the user utterance are recognized with higher accuracy. In this case, when configured as a work support system that supports the driving work of the vehicle, for example, current position information is used as a recognition result of the current state recognition means (claim 11).

[0018]

FIG. 1 shows a first embodiment of the present invention.
This will be described with reference to FIG. This embodiment is an example of application to a dish navigation system.

FIG. 1 is a schematic diagram showing an outline of a cooking navigation system. This system is based on a dialog management unit 1 that centrally controls each unit and plays a central role in exchanging information between the units.
The database includes a cooking knowledge database 2 as a database, an utterance storage unit 3 for temporarily storing various information, and an input / output unit for various information.

As input units connected to the dialog management unit 1, a situation recognition unit 4 and a voice recognition unit 5 are provided. The situation recognizing unit 4 transmits various information collected from the temperature sensor 6, the gas sensor 7, the weight sensor 8, and the timer 9 as the situation collecting means and output in the form of an electric signal to the dialog managing unit 1. The temperature sensor 6 is a sensor that measures the temperature in and around the cooking utensil, and the gas sensor 7 is a sensor that measures the flow rate and the flow velocity of the gas used.
The weight sensor 8 is a sensor for measuring weight incorporated in a kitchen scale for measuring weight or a sensor for measuring the weight of a material being cooked, and the timer 8 is a kitchen timer for measuring time. Therefore, the temperature sensor 6, the gas sensor 7, the weight sensor 8, and the timer 9 function as sensors for sensing a physical or chemical change of the object to be cooked. Further, the voice recognition unit 5 stores voice information which is taken in by a plurality of microphones 10 functioning as situation collecting means and user utterance collecting means and output in the form of electric signals in a voice language database 11 as a database. Analysis is performed based on the information, and the result is transmitted to the dialog management unit 1. The microphone 10 functioning as a situation collecting unit is installed at any position in the kitchen, such as a position where a cooking operation is performed or a position where a cooking tool such as a pot is used, and picks up a sound generated during cooking. In addition, the microphone 10 functioning as a user utterance collecting unit is installed near the user, and picks up the user's utterance during the cooking operation.

As an output unit connected to the dialog management unit 1, a speech synthesis unit 12 is provided. The speech synthesis unit 12 generates necessary speech information based on information stored in the speech language database 11 under the control of the dialog management unit 1, and outputs the speech information from a speaker 13 as a guide unit.

FIG. 2 is a block diagram showing the electrical connection of each part. FIG. 1 is a diagram schematically illustrating an outline of a cooking navigation system according to the present embodiment.
Each unit is configured as shown in the block diagram of FIG.
That is, a storage device including the ROM 15 and the RAM 16 of the CPU 14 that executes various arithmetic processes and processes each unit intensively is connected via the bus line 17, and this is a main constituent element of the dialog management unit 1. ing. For the utterance storage unit 3, a resist area in the RAM 16 is used. In addition, sensors such as a temperature sensor 6, a gas sensor 7, a weight sensor 8, and a timer 9 as a situation collecting means, and a microphone 10 as a situation collecting means and a user utterance recognizing means include an A / D converter 1
The speaker 13 serving as a guide unit is connected to the bus line 17 via a D / A converter 19, and these units are collectively controlled by a CPU 14. Here, the A / D converter 18 has a multi-channel configuration, of which several channels are input sound or input voice from the microphone 10, and other channels are sensors such as a temperature sensor 6, a gas sensor 7, a weight sensor 8, and a timer 9. The collected information from each class is processed and output to the situation recognition unit 4 and the voice recognition unit 5, respectively. Further, the D / A converter 19 converts an output waveform from the voice synthesis unit 12 into a signal suitable for driving the speaker 13. A CD-ROM drive 20 is connected to the bus line 17 connected to the CPU 14, and the CD-ROM drive 20
The contents of M are read. The CD-ROM stores a cooking knowledge database 2 and a speech language database 11.

FIG. 3 is a schematic diagram showing a data structure of cooking knowledge stored in the cooking knowledge database 2. As shown in FIG. 3, the cooking knowledge database 2 stores, for each dish, a cooking procedure, a cooking object, a processing method, a quantity thereof, a processing degree, and an end condition in association with each other. . That is, the cooking knowledge database 2 systematically stores and holds information for assisting the cooking operation of the user.

Here, a CD-ROM whose contents are read by the ROM 15 or the CD-ROM drive 20 includes:
An operation program is fixedly stored, and various functions are executed by a general control function of the CPU 14 according to the operation program, that is, a microcomputer function. One of the functions executed in this way is a function performed by the situation recognition unit 4 and the voice recognition unit 5 under the overall control of the dialog management unit 1. This function is roughly described as a temperature sensor 6, a gas sensor 7, a weight sensor 8, a timer 9,
And a function as current state recognition means for recognizing the current situation where the user is located based on a signal collected by a part of the microphone 10 and output from the A / D converter 18. Among the current state recognition functions, the current state recognition function executed based on the voice picked up by the microphone 10 is performed based on the existing voice recognition technology in the voice recognition unit 5. Examples of existing speech recognition technologies include, for example, "Speech Recognition Using Sentence Templates" by the Information Processing Society of Japan / Spoken Language Information Processing Study Group (Masako Mochizuki, Tetsuya Muroi,
1995, 7) and "The Present and Future of Speech Recognition by Mathematical Statistical Models" on pages 26 to 32 of the Acoustical Society of Japan, Vol. 48, No. 1, etc.

Other functions performed by the microcomputer function include a function performed by the voice recognition unit 5 under the overall control of the dialog management unit 1. This function is generally based on a signal collected by some of the microphones 10 used as a user utterance collecting means and output from the A / D converter 18, and a related utterance performed by the user on the guidance information. Is a function as a user utterance recognizing means for recognizing in association with a previous system utterance or a user utterance. here,
The term “related utterance” means a question, a question returned, or a related information request made by the user to the guidance information. Further, “system utterance” means a voice announcement by the dish navigation system of the present embodiment, and “user utterance” means a user utterance. Such a user utterance recognition function is performed based on the existing voice recognition technology in the voice recognition unit 5. Examples of the existing speech recognition technology include, for example, "Speech Recognition Using Sentence Templates" by the Information Processing Society of Japan and Spoken Language Information Processing Study Group (Masako Mochizumi, Tetsuya Muroi, 1995, 7), Journal of the ASJ 48, 1 No. 26 to page 32, “Present and Future of Speech Recognition by Mathematical Statistical Model”.

Still other functions executed by the microcomputer function include an information search function and a guidance function. The information search function is a function as an information search means for searching the cooking knowledge database 2 for information necessary for supporting the work performed by the user. Information retrieval from the cooking knowledge database 2 and the speech language database 11 in the information retrieval function is performed according to the system of the cooking knowledge database 2, and in this case, information retrieval may be performed according to the recognition result of the current state recognition means. The system of the cooking knowledge database 2 is a cooking object associated with each cooking procedure for a certain cooking, its quantity, its processing method, degree of processing, and termination conditions. The cooking procedure is divided into work units suitable for human beings to grasp a series of work at a time. Therefore, the information search by the information search means is also performed by being divided into such work units. next,
The guidance function is a function as a guidance unit that announces necessary information retrieved by the information retrieval unit by voice using the speaker 13. The voice announcement by the speaker 13 is performed by the voice synthesizer 12 using the template to synthesize the voice based on the data stored in the voice language database 11. Existing speech synthesis technology is used for such processing. As the existing speech synthesis technology, there is, for example, a technology described in “Current State and Future of Research on Speech Synthesis” on page 39 to 45 of The Acoustical Society of Japan, Vol.

Here, work support processing by the dish navigation system of the present embodiment will be described. The work support process generally provides guidance on a cooking object and a processing method for each dish, and at this time, refers to a state in which the user is currently located and interactively interacts with the user in the following manner. That is, the process is determined and executed. Here, for example, in the case of cooking curry, a voice announcement such as "Please cut onion" is given, and "How many pieces of onion are cut" or "How to cut onion" While responding to related utterances from users, such as
At this time, for example, the current state of the user such as silence lasting for 10 seconds or more is detected. In this case, it is determined that the process of cutting the onion into two fine pieces has been completed, and the next procedure is announced by voice. Work support is provided. Therefore,
The actual processing will be described next with reference to the flowcharts of FIGS.

FIG. 4 is a flowchart showing an outline of the work support processing. As the work support processing, first, a search for work contents is performed (step S1). This process
This is done by the user declaring the name of the dish, or searching for the next process after the end of a specific procedure for a certain dish. The declaration of the dish name by the user is performed by recognizing the utterance of the user captured by the microphone 10 by the voice recognition technology.

When the work content is searched, it is determined whether there is any work content (step S2). This is performed by determining whether or not there is a corresponding dish item in the cooking knowledge database 2 and whether or not all procedures for a certain dish have not been completed yet. For example,
If "curry" is declared, since the item of curry is in the cooking knowledge database 2, it is determined in step S2 that there is work content. On the other hand, if there is no item in the cooking knowledge database 2 called "tom curly", it is called "tom curly". When the dish name is declared, it is determined in step S2 that there is no work content. Similarly, assuming that curry cooking procedures stored in the cooking knowledge database 2 are 1 to 30, procedure 3
After the completion of 0, it is determined in step S2 that there is no work content. If there is no work content, an end notification is sent to another module (step S3), and the process ends.

If it is determined in step S2 that there is work content, a trigger condition is set (step S4). For example, when the work content is curry procedure 3, the end condition in the data structure in the cooking knowledge database 2 is “report = amber” or “weight = １ ／”, so this is set as a trigger condition. In addition, "Report = finished"
For example, a user utterance indicating the end of a task such as “report = next” and synonyms thereof are set as trigger conditions.
The setting is performed using a work area of the RAM 16. Here, "weight = 1/2" is recognized by the current status recognition means, "report = amber", "report = finished",
User utterances such as "report = next" are recognized by the utterance recognition means. In the recognition of the user utterance by the utterance recognition means, the voice recognition unit 5 recognizes each item (target, color, end) of the recognition sentence template such as “<target><color>” and “<end word>”. This is a speech recognition process performed by applying a specific word retrieved from the speech language database 11 to the word or the like.

Next, the work content searched in step S1 is searched by the information search means, and the guidance is output by the guide means (step S5). First, the dialogue management unit 1
Is performed by searching the cooking knowledge database 2 to find guidance information, writing this in the utterance storage unit 3, and interrupting the speech synthesis unit 12. Thereby, the speech synthesis unit 12 creates an utterance sentence from the guidance information written in the utterance storage unit 3 using the utterance sentence template, synthesizes the utterance sentence, and outputs it from the speaker 13. For example, when the work content is curry procedure 3, the data structure has a data structure of “onion” to be cooked, quantity “2”, processing “fried”, and processing degree “medium heat”. Based on the data structure and the utterance template “<target><process>”,“<target> = onion, <process> = cut”
Is synthesized. As a result, a voice announcement “Please cut the onion” is given from the speaker 13. Here, from the data structure of the cooking object "onion", the quantity "2", the processing "fried", and the processing degree "medium heat", only the words of the cooking object "onion" and the processing "fried" are selected. While these are essential and definitive information that is not affected by knowledge or taste, the quantity “2” and the processing level “medium heat” are reference information that is influenced by knowledge or taste, and it is not necessary to provide guidance. Is considered.

After the guidance output of the work content (step S5),
The conversation management unit 1 outputs a predicted question to the utterance storage unit 3 (step S6). The prediction question is a question that is predicted after the voice announcement “Please cut the onions”
For example, "<target> is <quantity><question>" or "<degree>
It is output in the form of a recognition sentence template such as <processing><question>. The speech recognition unit 5 performs speech recognition processing by assigning specific words from the speech language database 11 to each item in the recognition sentence template. When the speech recognition unit 5 recognizes the question, for example, when the speech recognition unit 5 recognizes the question "How many onions" applies to the recognition template "<object><quantity><question>", the dialog management unit is interrupted. In step S7, the dialog management unit 1 searches the cooking knowledge database 2 and outputs the answer to the utterance storage unit 3, and the speech synthesis unit 12 transmits the answer to "<target> is <quantity>
Voice synthesis is performed using a template "<unit>", and a guidance "two onions" is output from the speaker 13 (step S5). That is, reference information is announced by voice with respect to the essential deterministic information first announced by voice, and the system and the user interact with each other. That is, in steps S5 to S7, until the trigger condition set in step S4 is detected (step S8), the system utterance (guidance)
Dialogues of user utterance (related utterance) and system utterance (answer) are repeated.

Thereafter, when the trigger condition set in step S4 is detected (step S8), after the trigger condition is released (step S9), the work contents of step S1 are searched again.

FIG. 5 is a flowchart showing an outline of the condition setting process. This is a process executed every few ms as a subroutine of the work support process shown in FIG. First, it is determined whether there is an end notification (step S3 in FIG. 4) (step S11). If there is no end notification, it is determined whether there is a trigger condition (step S5 in FIG. 4) (step S12). If the trigger condition does not exist, the process returns to the determination process of step S11. If the trigger condition exists, the current condition recognizing means or the user utterance recognizing device executes the current condition recognizing function or the user utterance recognizing function, respectively (step S1).
3). That is, the temperature sensor 6, gas sensor 7, weight sensor 8, timer 9, and microphone 10 collect
Based on the signal output from the / D converter 18, the current situation where the user is located and the user's speech are recognized. In this case, among the information collected from the microphone 10, sounds that are clues to the user's recognition of the current state include sounds to be cooked such as a sound of roasting, a sound of boiling, a sound of steam in a pressure cooker, a food processor, a hand mixer, and the like. , The sound of the user such as the sound of the knife hitting the cutting board, and the end sound of the microwave oven.

After the current state of the user is recognized in step S13, the setting condition is determined (step S14). In the setting condition determination, it is determined whether the current state of the recognized user satisfies the trigger condition (step S7 in FIG. 4). For example, if the work content is curry procedure 3, “report =” in the data structure in the cooking knowledge database 2
Since “Amber” or “Weight = １ ／” is set as the trigger condition, whether the user's utterance “It has become amber” or the user's utterance indicating the end of the work such as “Next” is recognized. Alternatively, if the situation recognition unit 4 recognizes from the output value of the weight sensor 8 that the weight of the fried onion has been reduced to half, it is determined that the current state of the user satisfies the trigger condition. . When the trigger condition is satisfied, the dialog management unit 1 is interrupted. Thereby, the trigger condition is canceled in the work support processing of FIG. 4 (step S9), and the work content is searched again (step S1). In the example of the operation content of the curry procedure 3 determined that the current state of the user satisfies the trigger condition, in step S1, the operation content of the curry procedure 4 is searched by searching for the operation content. Here, the function of the information search means is executed.

FIG. 5 is a flowchart showing an example of a method of fitting a word to each item in a template and a method of determining a priority order among a plurality of words. Here, in the case where the curry procedure 2 in the cooking knowledge database 2 in FIG. 3 is guided to “Please heat the salad oil” (FIG. 4, step S5), “<target> is <quantity><question 〉 ”. At this time, the word j of “salad oil” and the synonymous word “salad oil” included in the preceding guidance utterance are applied as candidates to the item i of <object> in the recognition sentence template (step S21). 23). In this case, since “salad oil” is a word included in the system utterance immediately before “Please heat salad oil” (step S24), its priority is raised by a constant a (step S25). If the previous user's utterance includes "salad oil" which is a synonym for "salad oil" (step S26), the priority is raised by a constant b (step S27). This is because, in addition to taking into account a pull-in phenomenon in a dialogue that the same word tends to be used as the word used by the other party, and taking into account that there is a word whose expression method is determined by a person.

Then, the word j is incremented one by one until the word j in a certain item i of a certain template is exhausted, and the processing of steps S23 to S27 is repeated (steps S28 and S29).
Item i is incremented until is exhausted, and S23-2
7 is repeated (steps S30 and S31).

On the other hand, as an example of a method of determining the priority of the template, it is assumed that a system utterance "Heat is high heat" is followed by a user utterance "How much is the heat?" In this case, after such a question, it tends to happen that one wants to forget or check other information that has already been issued. So, in such a case,
That is, the priority of the template regarding the predicted question regarding the quantity such as “<object><quantity><question>” is raised.

Further, even after a lapse of a predetermined time after the guidance announcement, there is a possibility that there is no question and the end of the work cannot be detected. This is the voice recognition unit 5 and the situation recognition unit 4
In such a case, there is a high possibility that the work guided by the search for the material to be cooked or the cooking utensil or other work is interrupted. Therefore, in this case, since it is expected that the content already guided will be returned again, the predicted question related to “<target><processing>” included in the immediately preceding system utterance, in this case, the first work content The priority of the template is raised.

A second embodiment of the present invention will be described with reference to FIGS. This embodiment is an example of application to a car navigation system.

FIG. 7 is a schematic diagram showing an outline of the car navigation system. This system is based on a dialog management unit 31 that centrally controls each unit and plays a central role in exchanging information between the units. The dialog management unit 31 includes a map database 32 as a database,
An utterance storage unit 33 for temporarily storing various information and an input / output unit for various information are connected to each other. The map database 32 has a hierarchical structure based on node information as a branch point and link information as a road connecting the branch points. The utterance storage unit 32 is a registration area necessary for voice recognition processing and voice synthesis processing between the dialog management unit 31 and other units. And the dialogue management unit 31
Are as follows.

As input units connected to the dialog management unit 31, there are provided a position measuring unit 34, a geographic search unit 35, a display / input control unit 36, and a voice recognition unit 37. The position measuring section 34 is a GPS receiver 3 as a situation collecting means.
8, processing various information collected from the direction sensor 39 and the vehicle speed sensor 40 and output in the form of electric signals;
It is transmitted to the dialogue management unit 31. That is, the information processing in the position measuring unit 34 includes position information based on triangulation based on radio waves from artificial satellites obtained from the GPS receiver 38, traveling direction information obtained from the direction sensor 39 such as an optical fiber gyro and a geomagnetic sensor, and vehicle speed. This is a process for integrating vehicle speed information and the like based on the vehicle speed pulse obtained from the sensor 40. The geographic search unit 35 includes the map database 32,
The current position is obtained by using a so-called map matching technique in which a signal from the position measuring unit 34 is compared with information in the map database 32. In addition, the geographic search unit 35
The route storage unit 41 is also provided, and a function of causing the route storage unit 41 to store and hold a route from a certain point to another certain point is also provided. Since such various kinds of processing are basic technologies of a car navigation system that is currently widely used,
Detailed description is omitted. Then, the display / input control unit 36
Is provided with a display input / output device 42, and controls input and output of necessary information to and from the dialog management unit 31. Further, the voice recognition unit 37 stores the voice information captured by the microphone 43 functioning as the user utterance collection means and output in the form of an electric signal as a voice language database 4 as a database.
The analysis is performed on the basis of the information stored in the storage unit 4 and the result is transmitted to the dialog management unit 31. In this case, the microphone 43 functioning as the user utterance collecting means is arranged near the driving user and picks up the utterance of the driving user.

As an output section connected to the dialog management section 31, a speech synthesis section 45 is provided. The voice synthesizing unit 45 generates necessary voice information based on the information stored in the voice language database 44 under the control of the dialog management unit 31, and uses the voice information as the speaker 4 as a guide unit.
6 to output.

FIG. 7 is a diagram schematically showing the outline of the car navigation system according to the present embodiment.
Actually, each unit is configured in a form like a block diagram shown in FIG. 2 exemplifying a block diagram of the dish navigation system (not shown). That is, a microcomputer including a CPU, a ROM, and a RAM is provided. The microcomputer includes a GPS receiver 38, a direction sensor 39, and a vehicle speed sensor 4.
0, the display input / output device 42, the microphone 43, and the speaker 46 are connected, and a large-capacity storage medium for storing and holding the map database 32 and the audio language database 44 is connected, thereby forming a system. A predetermined area in the RAM is used for the utterance storage unit 33 and the route storage unit 41, and the position measurement unit 34, the geographic search unit 35, and the display /
The functions of the input control unit 36 are executed by microcomputer control. Further, the speech recognition method in the speech recognition unit 37 and the speech synthesis method in the speech synthesis unit 45 are as follows.
Since this is the same as the speech recognition technology and the speech synthesis technology in the first embodiment, the description is omitted.

Here, various means executed by the microcomputer function include a current state recognition means, a user utterance recognition means, an information search means, and a guidance means. The current status recognition means is a GPS receiver 3 as a situation collection means.
8. Based on the signals collected and output by the direction sensor 39 and the vehicle speed sensor 40, the current state recognition function of recognizing the current situation where the user is placed is performed. Next, based on the information collected by the microphone 43, the user utterance recognition means performs a function of recognizing a related utterance made by the user with respect to the guide information in association with a previous system utterance or a user utterance. Here, “related utterance” means
It means a question, a return of a question, a request for related information, or the like made by the user to the guidance information. Further, “system utterance” means a voice announcement by the car navigation system according to the present embodiment, and “user utterance” means a user utterance. Next, the information search means performs an information search function of searching the map database 32 and the speech language database 44 for information necessary to support the work performed by the user according to the recognition result of the current state recognition means. The guidance means performs a guidance function of announcing the necessary information retrieved by the information retrieval means through the speaker 46. The voice announcement by the speaker 46 is performed by the voice synthesizer 45 using the template to synthesize the voice based on the data stored in the voice language database 44. Such processing is performed in the first embodiment. The existing speech synthesis technology as described in the embodiment is used.

FIG. 8 is a schematic view exemplifying the contents stored and stored in the path storage unit 41. The route storage unit 41 stores, for each intersection ID that specifies an intersection, a point number, a direction, a distance,
Intersection names and landmarks are set. Such setting is performed by an input operation of a predetermined item from the display input / output device 42, but the input operation itself is no different from the conventional car navigation system. That is, a point number, a direction, a distance, an intersection name, and a landmark are automatically set by simply inputting a departure point and a destination point. Other information is automatically set when a direction is input for each intersection ID. Any existing input operation method, such as a method, is acceptable.

The work support processing by the car navigation system according to the present embodiment will be described. This work support processing is generally performed with a general voice guide that determines a route to a destination before traveling and stores and retains the route in a route storage unit 41, and announces the route during actual traveling. This is a function of the car navigation system. Therefore, the description is omitted. On the other hand, among such work support functions, a function unique to the present embodiment is that the user refers to the current state where the user is placed and finds an audio announcement suitable for the state in an interactive manner. Therefore,
Hereinafter, functions unique to this embodiment will be described with reference to the flowchart of FIG.

FIG. 9 is a flowchart showing the flow of the process in which the dialogue management section 31 creates guidance contents and predicted questions. First, the point number i to be processed is set to 0 (step S41), then the point number i is incremented by one (step S42), and it is confirmed whether all the points to be guided have ended (step S43). .
If the guidance has been completed for all points, the process is terminated. Otherwise, the information of the point number i is obtained from the information stored in the route storage unit 41 (step S4).
4). Then, when the current position detected by the position measurement unit 34 approaches the intersection of the point number i of the map information searched from the map database 32 by the map search unit 35, the dialog management unit 31 transmits the information on the guide point to the route. The data is loaded from the storage unit 41, and first, only the direction and the distance are output to the utterance storage unit 33 as utterance contents (step S45). This is because the information of the direction and the distance is the urgent necessary information required as the voice navigation at the intersection.

When the contents of the utterance are stored and held in the utterance storage unit 33, a subroutine different from the flowchart shown in FIG. 9 is executed, the voice information is voice-synthesized by the voice synthesis unit 45, and the guidance is announced from the speaker 46. You. At this time, the speech synthesis unit 45 performs speech synthesis using an utterance sentence template “<distance><unit> forward <direction> direction”. For example, in the utterance storage unit 33, the storage information of the route storage unit 41 as illustrated in FIG.
= Right, <distance> = 300 ", the synthesized voice is" 300 meters ahead to the right ". This guidance content is urgent necessary information.

Next, the dialogue management section 31 stores the route storage section 41
Is output to the utterance storage unit 33 (step S46). At this time, the voice recognition unit 37 prepares all recognition sentence templates related to the predicted question, which is the user's related utterance with respect to the system utterance at the guide point i, and prepares for the user's question taken in through the microphone 43. As the recognition sentence template, for example, "<what>, <name><intersection>
<Question>, <landmark><existence><question>, etc. are prepared. Then, in this case, a is added to the priority of the predicted question regarding the item that was not guided (step S47), and the priority is set high.

Thereafter, the intersection of the point number i of the map information retrieved from the map database 32 by the map retrieval unit 35 and the current position point detected by the position measurement unit 34 are compared, and the current position point is determined by the point number i. The user utterance recognized by the voice recognition unit 37 is recognized (step S49) on the premise that it is determined that the vehicle does not pass through the intersection (step S48). That is, the voice recognition unit 37 recognizes the user's utterance by applying the voice generated by the user to the recognition sentence template. For example, the question "What is the intersection?"
By applying user utterances to recognition sentence templates such as "what", "name", "intersection", and "question",
Recognition of the user utterance is performed. In this case, speech recognition
This is performed by using the existing speech recognition technology.
Since the priority of the prediction question regarding the item not guided in 47 is set high, the accuracy of the voice recognition is high.

In the following step S50, the voice recognition unit 37
Evaluates the question item and outputs the result to the utterance storage unit 33. Thereafter, a subroutine different from the flowchart shown in FIG. 9 is executed, the voice information is voice-synthesized by the voice synthesizer 45, and the guidance is announced from the speaker 46. At this time, when assuming a user's question "What is the name of the intersection?", The speech synthesis unit 45 performs speech synthesis using the utterance sentence template "<Intersection name>Intersection." Therefore, in the example of FIG. 8, a voice synthesis “is an intersection in front of the arena” is performed, and this is announced from the speaker 46. This guidance content is additional information.

Thereafter, the voice recognition section 37 lowers the priority by adding b to the priority of the recognition sentence template relating to the already guided item, in the example of FIG. 8, the distance and direction (step S51). , And further wait for a question from the user. If it is determined that the current position has passed the guide point i (step S48), the point number i is incremented by one, and the same processing is repeated (steps S42 to S51). This is determined in step S43. Continue until all points have been completed.

Here, when the guide point is close to the intersection that is the guide point, the user may not be able to identify the guided intersection. Thus, in preparation for such a case, the voice recognition unit 37 sets “<current location> ｛to |
Then, a recognition sentence template such as <direction change><question> or "[intersection] <current location><question>" is provided. [] In such a recognition sentence template indicates the possibility of omission. For example, when receiving the question "is here?", The voice recognition unit 37 determines that the <intersection> is omitted by the recognition sentence template "[intersection] is <current location><question>", and records the utterance. The intersection ID 0023 is searched from the utterance history of the unit 33, and <intersection> is set to the intersection ID.
The name is replaced before the arena, which is the intersection name of 0023 (the example in FIG. 8). Also, the instructional pronoun “here” is replaced by the speech recognition unit 37 with the item “current location”. To resolve this semantically, the dialog management unit 31 sends the location information of the current location via the geographic search unit 35 to the nearest location. And the intersection ID, intersection ID, etc. of the user are acquired, and “here” is replaced with <current location>. As a result, “intersection ID 23 = intersection ID
23 "and" intersection ID23 = intersection ID22 ". In the former case, the evaluation is “true”, which is sent to the speech synthesizer 37 and a guidance output such as “yes” is made. On the other hand, in the latter case, the evaluation is “false”, which is sent to the speech synthesizer 37 and a guidance output such as “not correct” is made.

[0055]

According to the first aspect of the present invention, the present state where the user is placed is collected, and support information corresponding to the current state is provided.
When the user makes a related utterance such as a question, a question, or a related information request, the user recognizes and responds to the utterance in association with the previous system utterance or the user utterance. It can provide the required information and enhance the work support effect.

According to the second and third aspects of the present invention, in the work support system according to the first aspect, the user utterance recognizing means includes a word contained in the immediately preceding system utterance and a synonymous word thereof as a user utterance recognition candidate. Therefore, the prediction accuracy of the user utterance after the system utterance can be increased, and the recognition rate of the voice recognition can be improved. In particular, according to a third aspect of the present invention, in the work support system according to the second aspect, the user utterance recognizing means sets the priority of a word included in the immediately preceding system utterance higher than that of the synonymous word. If a phonetic word contains a user's utterance that appeared before, the priority of the synonymous word is raised, so that the prediction accuracy of the user's utterance after the system utterance is further improved, and the recognition rate of speech recognition is improved. It can be further improved.

The inventions according to claims 4 to 6 are as described in claim 1
In the work support system described above, the user utterance recognition unit infers whether or not the user's attention target is in the current guidance information, and if it is determined that the user's attention target has shifted from the current guidance information, the previous guidance It prepares for the re-request for information, so that the movement of the user's attention can be inferred from the natural flow of the dialogue, and thus the user is able to use complex mechanisms, special sensors, and no extra dialogue for them. Can follow the natural behavior pattern of the system, and the recognition rate of voice recognition can be improved.

The inventions according to claims 7 to 9 are as follows.
In the work support system described above, the guide means sets a priority order for the guide information to be guided by the guide section and sequentially guides the guide information to the guide section according to the information request utterance of the user. The user himself / herself can make a selection, and thus, can provide appropriate information according to the user's request. In particular, the invention according to claim 8 is the work support system according to claim 7, wherein the guide means guides the guide unit with the essential deterministic information, and provides the guide information with the reference information according to the user's information request utterance. Since the guidance is performed, it is possible to provide a guide that does not miss the machine required for the cooking operation. According to a ninth aspect of the present invention, in the work support system according to the seventh aspect, the guide unit guides the urgent necessary information to the guide unit, and guides the additional information to the guide unit in response to the user's information request utterance. Therefore, guidance that does not impair the safety of the vehicle during driving can be provided.

According to the tenth and eleventh aspects of the present invention, in the work support system according to the first aspect, the user utterance recognizing means resolves the anaphora or omission included in the user utterance based on the recognition result of the current state recognizing means. Therefore, it is possible to perform speech recognition on anaphors and omissions that cannot be solved only from the linguistic context, and it is possible to eliminate the ambiguity of the user utterance and improve the reliability of the system.

[Brief description of the drawings]

FIG. 1 is a schematic diagram showing an outline of a cooking navigation system as a first embodiment of the present invention.

FIG. 2 is a block diagram illustrating electrical connection of each unit.

FIG. 3 is a schematic diagram showing a data structure of a cooking knowledge database.

FIG. 4 is a flowchart showing an outline of a work support process.

FIG. 5 is a flowchart showing an outline of a condition setting process.

FIG. 6 is a flowchart showing a method of fitting a word to each item in a template and a method of determining a priority order among a plurality of words.

FIG. 7 is a schematic diagram showing an outline of a car navigation system as a second embodiment of the present invention.

FIG. 8 is a schematic diagram illustrating a part of the route information stored and held in the route storage unit.

FIG. 9 is a flowchart illustrating a flow of a process of creating a guidance content and a predicted question.

[Explanation of symbols]

 6 to 10, 38 to 40, 43 Situation collecting means 10, 43 User utterance collecting means 2, 11, 32, 44 Database 13, 46 Guide unit

────────────────────────────────────────────────── ─── front page continued (51) Int.Cl. ⁶ identifications FI G10L 3/00 571 G10L 3/00 571H G01C 21/00 G01C 21/00 H // G06F 3/16 340 G06F 3/16 340L

Claims (11)

[Claims] 1. A work support system for guiding a user who is performing a predetermined task to information for supporting the task, wherein a current state where the user is located is collected from an information source other than the user's utterance. State collecting means for converting the current state into an electric signal and outputting the electric signal; current state recognizing means for recognizing the current state of the user based on the output signal of the state collecting means; A user utterance collecting unit that outputs and outputs a user utterance based on an output signal of the user utterance collecting unit, and recognizes an utterance related to guidance information performed by the user in association with a previous system utterance or a user utterance. An utterance recognizing means, a database for systematically storing and holding information supporting a user's work, Information search means for searching the database for information necessary to support the work performed by the user in accordance with the knowledge result; and guidance means for guiding the necessary information searched by the information search means to a guidance unit. A work support system comprising: 2. The work support system according to claim 1, wherein the user utterance recognizing means selects a word included in the immediately preceding system utterance and its synonymous word as a recognition candidate of the user utterance. 3. The user utterance recognizing means sets the priority of a word included in the immediately preceding system utterance higher than that of the synonymous word, and includes a user utterance that previously appeared in the synonymous word. 3. The work support system according to claim 2, wherein the priority order of the synonyms is raised. 4. The user utterance recognizing means infers whether or not the user's attention target is in the current guidance information. If it is determined that the user's attention target has shifted from the current guidance information, the previous guidance information 2. The work support system according to claim 1, wherein preparations are made for a re-request. 5. When the information search means searches for information that is the answer to the user's related utterance recognized by the user utterance recognition means, it is determined that the attention target of the user has shifted from the current guide information. 5. The work support system according to claim 4, wherein the priority of an item that is not a target of the user-related utterance is raised. 6. After a certain period of time has elapsed after the information has been guided from the guidance unit by the guidance means, if sufficient progress of the work cannot be recognized from the recognition results of the current state recognition means and the user utterance recognition means, the user's attention target is 5. The method according to claim 4, wherein it is determined that the information has shifted from the guidance information, and the priority order of an item that has not been a target of the immediately preceding user-related utterance is raised.
The work support system described. 7. The guide unit according to claim 1, wherein the guide unit sets a priority order for the guide information to be guided by the guide unit, and sequentially guides the guide information to the guide unit in response to an information request utterance of the user. Work support system. 8. The cooking device according to claim 1, wherein the guidance unit guides the guide unit with essential definite information, and guides the reference information to the guide unit in response to a user's information request utterance. The work support system according to claim 7. 9. The vehicle according to claim 1, wherein the guide unit guides the urgent necessary information to the guide unit and guides the guide unit to the additional information in response to an information request utterance of the user. The work support system according to claim 7, which performs the operation. 10. The work support system according to claim 1, wherein the user utterance recognizing means seeks a semantic solution of anaphor and omission included in the user utterance based on the recognition result of the current state recognizing means. 11. The work support system according to claim 10, wherein the recognition result of the current state recognition means is current position information, and the driving operation of the vehicle is to be supported.