JP2004021028A - Voice interaction device and voice interaction program - Google Patents

Abstract

An object of the present invention is to provide a user with discomfort when a user makes an invalid utterance in response to a question or confirmation by a guidance from a dialogue device, or when the device misrecognizes and needs the same content again. Generate guidance sentences that you do not hold.
SOLUTION: From an answer sentence, a judging means for judging whether or not a target answer of a guidance sentence has been obtained, and, when it is judged by the judging means that a target answer could not be obtained, the object is set as the target Another different guidance sentence for obtaining an answer is output as voice, and a re-requesting means for requesting the speaker again for the answer sentence is provided.
[Selection diagram] Fig. 1

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention provides a voice recognition device that outputs a guidance sentence in order to obtain a target answer from a speaker, analyzes a response sentence obtained from the speaker in response to the guidance sentence, and determines a target answer. About. Therefore, the present invention can be applied to, for example, an on-vehicle car navigation system and the like, and a voice dialogue for performing a dialogue for filling information corresponding to a so-called “slot” such as a destination facility name or address. It can be applied to devices and the like.
[0002]
[Prior art]
An example of a speech recognition device that outputs a guidance sentence in order to obtain a desired answer from a speaker and analyzes the answer sentence obtained from the speaker in response to the guidance sentence to determine a desired answer is an example. As an example, there is known a voice interactive device described in Japanese Patent Laid-Open Publication No. Hei 10-20884: Voice Interactive Device. In order to obtain a smoother answer from the speaker, the guidance sentence can be made easier to understand. In a speech dialogue device disclosed in Japanese Patent Application Laid-Open No. 10-20884, the time required for the user to input a speech, the ratio of a correct recognition result, In this method, a skill level of a user is estimated from a flow of a dialog, and a voice guidance using a plurality of different expressions is automatically selected according to the skill level.
[0003]
[Problems to be solved by the invention]
In the above-described voice dialogue apparatus disclosed in Japanese Patent Application Laid-Open No. H10-20884, a plurality of guidances are prepared in order to obtain a desired answer, but this is done by changing the guidance sentence in consideration of only the user's skill level. ing. That is, repeated guidance such as in the case of erroneous recognition is not considered. That is, in the conventional voice interaction device, the same guidance sentence is repeated when erroneous recognition is not performed or recognition is not performed. Therefore, the utterance may be invalid due to erroneous recognition of the voice interaction device even though the user has given a correct answer. In such a case, the voice interaction device needs to provide the same guidance again.
[0004]
In the conventional device, since the same guidance sentence is repeated, the user has an unpleasant sensation to the device, such as “a device that is inflexible”. Further, when the guidance sentence is difficult for the user to hear, it is difficult to understand even if the same guidance sentence is repeated, and there is also a problem that the user cannot obtain a desired answer of the voice interaction device after all. An object of the present invention is to reduce such discomfort of the user for the voice interaction device. A further object of the present invention is to realize a smooth voice dialogue between a user and a voice dialogue device.
[0005]
The present invention has been made in order to solve the above-described problems, and an object of the present invention is to realize a smooth voice dialogue with a voice dialogue device and obtain a desired answer without causing discomfort to a user. It is.
[0006]
It should be understood that one invention described above is not intended to achieve all the objects described above at the same time, and individual inventions are intended to achieve the respective objects.
[0007]
[Means for Solving the Problems]
In order to solve the above-mentioned problem, in the voice interactive device according to claim 1, a guidance sentence is output as voice to obtain a target answer from the speaker, and an answer sentence obtained from the speaker in response to the guidance sentence is provided. In the voice dialogue device that analyzes the answer and determines the target answer, the judgment means for judging whether or not the target answer of the guidance sentence is obtained from the answer sentence, and the target answer is not obtained by the judgment means If it is determined that the answer is different, another different guidance sentence for obtaining the intended answer is output as voice, and the speaker again has an answer sentence re-requesting means for requesting the answer sentence. I do. In other words, when the target answer cannot be obtained, the voice interaction apparatus outputs a different guidance sentence in voice to request the target answer, and requests the speaker's answer again.
[0008]
Further, according to the invention of claim 2, in the guidance sentence which is output by voice from the speaker in order to obtain the target answer, the target answer is obtained in at least one slot provided for each different category. Therefore, a sentence that prompts the speaker for an answer sentence, and a plurality of different guidance sentences are output as speech for at least one slot for which an answer is requested.
[0009]
Further, the invention according to claim 3 has a storage unit for storing a plurality of guidance sentences for each at least one slot in order to output different guidance sentences, and the answer sentence re-requesting unit includes at least one answer requesting answer. It is characterized in that a plurality of guidance sentences corresponding to the slots are sequentially selected from the storage means and output as voice.
[0010]
Further, according to the invention of claim 4, a set of a plurality of guidance sentences set for each of at least one slot for which an answer is required is derived from a set of guidance sentences of different expressions in order to derive a target answer of the slot. It is characterized by comprising.
[0011]
Further, according to a fifth aspect of the present invention, there is provided a voice dialogue program, wherein a computer of the voice dialogue apparatus outputs a guidance sentence by voice to obtain a desired answer from the speaker, The answer sentence is analyzed, and from the answer sentence, a judgment procedure for judging whether or not the intended answer for the guidance sentence has been obtained.If the judgment procedure indicates that the intended answer was not obtained, the The present invention is characterized in that it has another procedure for generating another different guidance sentence for obtaining a target answer, outputting the guidance sentence as voice, and requesting the speaker again for the answer sentence.
[0012]
Further, in the voice dialogue program according to the invention of claim 6, in the computer of the voice dialogue device, the answer sentence re-requesting procedure seeks an answer in order to obtain a target answer in at least one slot provided for each different category. The method further comprises a step of outputting a plurality of different guidance sentences by voice for at least one slot.
[0013]
Actions and effects of the present invention
This section mainly describes the functions and effects of the invention described in each claim. To facilitate understanding of the present invention, the present invention is exemplarily embodied and described, but does not limit the configuration of the claims. The part concretely described as an example is also an explanation of the embodiment of the invention.
[0014]
First, according to the first aspect of the present invention, a guidance sentence is output by voice in order to obtain a desired answer from a speaker, and an answer sentence obtained from the speaker is analyzed in response to the guidance sentence. Since the determination means is provided for determining whether or not an answer has been obtained, it can be determined whether or not an answer has been obtained for the guidance sentence. If it is determined that the intended answer could not be obtained, another different guidance sentence for obtaining the intended answer is output as a voice, and an answer requesting the speaker to request the answer again. Since a sentence re-requesting means is provided, when an answer is not obtained, a guidance sentence different from the previous sentence is output by the answer sentence re-requesting means. As a result, user discomfort is reduced. Furthermore, since the expressions are different, there is an effect that it is easy to understand what to answer. For example, the spoken dialogue device inquires about “store name”, and the user responds thereto. Here, it is assumed that a valid answer was not obtained because the user made an incorrect answer.
Dialogue device: "Please tell me the name of the store."
User: "Restaurant."
Recognition result: "Restaurant."
This is the type of business, and is not the store name that is the target answer of the guidance, so it is necessary to ask a question again, so that a guidance sentence different from the immediately preceding guidance sentence is output.
Dialogue device: "What kind of store is it?"
User: "It's a dandelion."
Recognition result: "It is a dandelion."
As described above, a different guidance sentence is output as a voice, so that the user is less likely to feel uncomfortable even when asked again from the device, and it is easier to understand what must be answered. In this way, changing the expression when the expected answer is not obtained is a very natural thing in a person-to-person conversation. Therefore, even in the case of a dialogue with the device, the dialogue can be smoothly advanced without feeling uncomfortable.
[0015]
Further, according to the invention of claim 2, a guidance sentence which is a sentence urging a speaker to answer in order to obtain an intended answer in at least one slot provided for each different category, which is an object of the voice interaction apparatus, is: Since a plurality of different guidance sentences are output by voice for at least one target slot for which an answer is requested, a plurality of different guidance sentences corresponding to the target slot can be output by voice.
[0016]
Further, the invention according to claim 3 enables a plurality of guidance sentences to be stored in the storage means for each at least one slot, and a plurality of guidance sentences corresponding to at least one slot for which an answer is requested by the answer sentence re-requesting means. Since it is possible to sequentially select and output voice from the storage means, even when the same answer as the previous guidance sentence is requested, a guidance sentence different from the immediately preceding guidance sentence can be output as voice. In this way, by sequentially selecting from the storage means and outputting the voice, for example, the expression of the first guidance sentence is general in the order of selecting the expression of the guidance sentence. The following guidance text uses expressions that emphasize a little more intelligibility than simplicity. It is also possible to output a guidance sentence with a ranking such as: In other words, the expression in the following guidance sentence emphasizes the comprehensibility. In other words, the simplicity and simplicity required of a guidance sentence are contradictory, and emphasis on one will cost one of them. However, since guidance sentences of different expressions can be sequentially selected and output as voice, it becomes possible to prepare guidance sentences of different specific gravity every time the guidance is repeated. As a result, even an unskilled user can use the apparatus without having to repeat the same guidance sentence and to have an uncomfortable or inflexible apparatus.
[0017]
Furthermore, in order to derive a target answer of at least one slot for which an answer is required, the invention of claim 4 is set for each at least one slot for which each answer is required so as to be composed of a set of guidance sentences having different expressions. Since a plurality of guidance sentences are configured, guidance of different expressions can be output as voice in order to obtain an answer for a target slot.
[0018]
Further, when the program according to the fifth aspect of the present invention is installed and used in the above-described voice interactive device, the above-mentioned voice interactive device can output a guidance sentence in order to obtain a desired answer from a speaker. Further, it is possible to analyze the answer sentence obtained from the speaker in response to the guidance sentence, and to judge from the answer sentence whether or not the intended answer of the guidance sentence has been obtained. Further, when it is determined that the intended answer was not obtained, another different guidance sentence for obtaining the intended answer is generated, the guidance sentence is output as a voice, and the speaker is requested again for the answer sentence. Will be able to do it.
[0019]
Further, when the program according to the invention of claim 6 is installed and used in the above-described voice interactive device, at least one slot for which an answer is required in order to obtain a target answer in at least one slot provided for each different category. A plurality of different guidance sentences can be output by voice.
[0020]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, the present invention will be described based on specific examples. However, the present invention is not limited to the embodiments described below.
[0021]
Here, a description will be given of an example of a dialogue in which a destination setting in a car navigation system is a target task.
[0022]
In this task, the spoken dialogue apparatus asks the user about three destination slots required for setting a destination, “shop name”, “address”, and “industry type”, and confirms an answer from the user. Here, it is assumed that the immediately preceding recognition result is confirmed (implicit confirmation) simultaneously with the question by putting the immediately preceding recognition result into the next question guidance.
[0023]
In the guidance generation unit, a guidance sentence template (FIG. 3) having a plurality of different expressions is stored in the guidance storage corresponding to the guidance slot including the combination of the slot to be asked (question slot) and the slot to be confirmed (confirmation slot). It is stored in the unit 171 and is selected from a group of templates corresponding to the combination of the corresponding question slot and the confirmation slot, and is used for generating guidance.
[0024]
FIG. 1 is a configuration diagram illustrating a logical configuration of a voice interaction device 100 according to an embodiment of the present invention.
[0025]
The voice interaction device 100 mainly includes a voice input unit 110, a voice recognition unit 120, a meaning understanding unit 130, a dialog control unit 140, a guidance generation unit 150, a voice output unit 160, a database 170, and the like. ing. Of course, the voice interactive device 100 has, as a physical hardware configuration, a man-machine interface unit such as a microphone included in the voice input unit 110 and a speaker included in the voice output unit 160, similarly to a known voice interactive device. It is embodied by a computer system provided.
[0026]
The voice recognition unit 120 recognizes the voice of the speaker as a character string. That is, the voice information input from the microphone (voice input unit 110) is converted into a character string by voice recognition processing using a voice recognition dictionary (a recognition language dictionary, a recognition acoustic dictionary, or the like).
[0027]
The meaning understanding unit 130 mainly includes a word extraction unit 131, a word determination unit 132, and the like. From these, necessary keywords (words serving as slot values) are extracted from the above character strings, it is determined whether or not the word is the target word of the guidance. Stored as slot value. More specifically, the word extracting unit 131 extracts a word having a slot value from a character string output as a result of speech recognition using a word dictionary or the like stored in the database 170. The word determining unit 132 determines whether the extracted word is a target word according to the guidance. To determine, for example, the category of the extracted word is identified using a word dictionary or a slot value candidate word list or the like, and it is determined whether the category belongs to the category assumed by the guidance. Judge whether or not. As the category, for example, any attribute such as an address, a place name, a facility type, a shop name, a business type, a facility name, a landmark name, or a user setting name can be defined. As a result of the determination, when it is determined that the target word is a target word, the target word is held as a slot value in a target slot. If the answer is "yes" or "no", the slot status of the corresponding slot is updated so as to evolve or retreat. If there is no negative word for the implicit confirmation, the confirmation is made and the slot state is updated. The slot state is, for example, a state in which no slot value is held is a slot (query slot) to be queried in order to obtain a slot value. The question slot state is a question slot state. The word “slot value” is held by the dialogue, but is not confirmed and the state of the slot to be confirmed (confirmation slot) is “confirmed” state. The state in which the slot value has been confirmed is called the “determined” state.
[0028]
The dialog control unit 140 determines a question item or a confirmation item to be asked next, controls the flow of the dialog, and the progress of the dialog changes the slot state with respect to the retained slot value. Interact until they reach the “determined” state.
[0029]
The response sentence re-requesting unit 141 updates the template number so as to control the same guidance sentence as the previous guidance sentence.
[0030]
The guidance generation unit 150 generates a response sentence (a confirmation response sentence, a question response sentence, etc.) for the speaker (user), and further converts the response sentence (word string) into an acoustic digital signal (voice information). Combine. However, this conversion / synthesis processing may be performed by the audio output unit 160 as exemplified below. To generate a guidance sentence, specifically, a combination of a confirmation slot and a question slot is obtained by referring to the guidance sentence slot in FIG. Further, by obtaining the template number, a template number for generating a guidance sentence different from the immediately preceding guidance sentence is obtained. After that, a guidance sentence is generated by referring to the target slot based on the “store name”, “address”, and “business type” held in the confirmation slot and the question slot.
[0031]
The database 150 mainly includes a speech recognition dictionary, a word dictionary, a slot value candidate word list, a speech synthesis dictionary, and a guidance sentence template (FIG. 3) stored in the guidance storage unit 171.
(A) Dictionary for speech recognition
It consists of a language dictionary for recognition, a sound dictionary for recognition, and the like.
(B) Word dictionary
It has categories, related categories, other attributes, pronunciation information, and the like.
(C) Slot value candidate word list
Candidate word list. A table consisting of pairs of words and their categories.
(D) Speech synthesis dictionary
It has pronunciation rules for speech synthesis regarding utterance inflection, word connection, spacing, and the like.
(E) Guidance sentence template (Fig. 3)
It consists of a plurality of guidance sentence templates with template numbers corresponding to pairs of confirmation slots and question slots.
[0032]
FIG. 4 is a flowchart illustrating a procedure of a process executed by the above-described voice interaction apparatus 100. In this procedure, first, an initial process is executed in step 400 to initialize a target slot, a conversation state, a guidance sentence slot, and a template number. In this process, an initial process such as loading a program and data expected to be frequently used from the database 170 onto a memory having a relatively high access speed is executed. For example, when the voice interactive device 100 has a display device (not shown), other initial processing such as displaying an initial menu screen may be performed.
[0033]
In step 402, the guidance generation unit 150 generates a guidance sentence. In generating the guidance sentence, a combination of a “confirmation slot”, a “question slot”, and a “template number” are obtained by referring to the guidance sentence slot shown in FIG. The guidance sentence template is selected from the guidance sentence template based on the combination of the “confirmation slot” and the “question slot” and the “template number”.
[0034]
In an initial state, first, all target slots do not hold slot values, and thus there are no slots to be checked. First, since the shop name is asked, "shop name" is held in the "question slot", and 1 is set as the template number as an initial value. Accordingly, in the guidance generation unit, the guidance sentence template with the template number “1” whose “confirmation slot” is “empty” and the “question slot” is “store name” is selected. Here, the following guidance sentence is prepared. Example: Dialogue device: "Please say the name of the store."
[0035]
Next, in step 404, in the voice output unit 160, the guidance sentence generated in step 402 is converted and synthesized into an acoustic digital signal (voice information), and is output to the speaker as voice.
[0036]
In step 406, the voice input unit 110 captures the voice of the speaker from the microphone (voice input unit 110) as voice information.
[0037]
In step 408, the speech recognition unit 120 converts the speech information into a character string using a speech recognition dictionary (such as a recognition language dictionary or a recognition acoustic dictionary).
[0038]
In step 410, the word extracting unit 131 of the meaning understanding unit 130 extracts a necessary word (word serving as a slot value) from the above-mentioned character string, and identifies the category of the word. Next, in step 412, the word judgment unit 132 judges whether the word is a target word. If it is the target word, the process proceeds to step 414, and if the extracted word is a word as the slot value of the target slot, the extracted word is held in the target slot as the slot value, and the slot state is updated. If the target word is a confirmation word such as "Yes" or "No", the slot value is deleted and the slot state is updated as necessary. At the same time, the template number is initialized and set to "0" (determination means).
[0039]
In step 416, the dialog control unit 140 determines whether or not all the target slots are in the “determined” state. If it is determined that the state is determined, the end guidance is output in step 420, and the process of the voice interaction device ends. On the other hand, if it is determined that the target slot has not been determined yet, in step 418, the guidance statement slot is set and the template number is updated based on the slot state. Here, if it is determined in step 412 that the word is not the target word, the slot state is not updated. Therefore, in step 418, the same combination of “question slot” and “confirmation slot” as the previous guidance sentence is obtained. A group of guidance sentence templates requesting the same answer is selected (answer sentence re-requesting means). However, since the template number is updated in step 418, even if a template of the same group is selected, the guidance sentence becomes a selection of a guidance sentence template having a different expression. Steps 418 to 418 are repeated to set a word serving as a slot value to all target slots, and set the slot state to “determined”.
[0040]
Hereinafter, the slot value, slot state, and guidance slot state of each target slot will be described in detail using specific examples.
[0041]
(1) The spoken dialogue apparatus asks a question about the “store name”, and the user responds to the question. Here, it is assumed that the speech recognition unit has correctly recognized.
Dialogue device: "Please say the name of the store."
User: "It's a dandelion."
Recognition result: "It is a dandelion."
As a result, “dandelion” is held in the “store name slot”, and the status of the slot and the “store name slot” is updated to “confirm”. In the guidance sentence slot, “store name” is set in “confirmation slot”, “address” is set in “question slot”, and “0” is set as the template number.
[0042]
(2) A “dandelion” is stored in the “store name slot”, and a guidance for asking an “address” is generated while confirming the result. In response, the user answers “address”. Here, it is assumed that erroneous recognition is performed and a valid recognition result is not obtained.
Dialogue device: "Please tell me the address of the dandelion."
User: "I'm Nagoya City."
Recognition result: "How many?"
Since a valid recognition result for “address slot” was not obtained, the slot value and the slot state are not updated, but the template number is updated. Therefore, a guidance sentence template is selected next, and a guidance sentence is generated.
[0043]
(3) Here, a guidance sentence having a different wording from the immediately preceding guidance sentence is generated. In response, the user answers “address”. Here, it is assumed that the speech recognition unit has correctly recognized.
Dialogue device: "Where is the dandelion?"
User: "I'm Nagoya City."
Recognition result: "Nagoya city."
“Nagoya City” is held in the “address slot”, and the slot state is updated to “confirmation”. The slot state of the “store name slot” is updated to “determined”, the template number is initialized, and “0” is set. For the guidance sentence slot, “address” is set to “confirmation slot”, “business type” is set to “question slot”, and “0” is set for the template number. Therefore, the guidance for asking the "business type" is generated while confirming the result of the "address slot" and the result.
[0044]
(4) The user answers "business type". Here, it is assumed that erroneous recognition is performed and a valid recognition result is not obtained.
Dialogue device: "What shop in Nagoya?"
User: "Restaurant."
Recognition result: "Yes, it is."
Since a valid recognition result for the “business type” slot was not obtained, the slot value and the slot state are not updated. In step 418, when the guidance sentence slot is set again, the guidance slot becomes the same as the question slot immediately before, but since the template number is updated, it does not become the same guidance sentence.
[0045]
(5) The guidance for asking the "business type" is selected again, but the guidance sentence template of "2" is selected. In response, the user answers "business type".
Dialogue device: "What kind of business is in Nagoya?"
User: "Restaurant."
Recognition result: "It is a restaurant"
“Restaurant” is held in the “industry” slot, and the slot status is updated to “confirm”. Further, "Nagoya City" is updated to a "fixed" state. For the template number, an initial value “0” is set. In step 418, since the template number is updated when setting the guidance sentence slot, the guidance sentence template with the template number "1" is selected here.
[0046]
(6) Finally, confirmation is made for the “industry slot”. In response, the user answers "yes" and the dialog ends.
Dialogue device: "Are you a restaurant?"
User: "Yes"
Recognition result: "Yes"
[0047]
The guidance sentence output in step 420 refers to the target slot, for example, and outputs the following guidance sentence.
Dialogue device: "Now, set a dandelion at a restaurant in Nagoya City."
Thus, by implementing the above-described invention, the above-described voice interactive device can be realized.
[0048]
As a result, even when the same content is repeatedly provided, it is possible to prompt the user to answer with a guidance sentence having a different expression without performing the same guidance, and without causing discomfort to the user. It is possible to expect a smoother conversation with the voice interaction device.
[0049]
In the above-described embodiment, an example has been described in which three slots of “store name”, “address”, and “business type” are determined. However, at least one other slot may be determined. Further, as other slots are determined, a suitable guidance sentence template may be prepared.
[0050]
Since the guidance sentence template in the above embodiment is one example, other guidance sentence templates may be prepared.
[0051]
Further, a combination of “question slot” and “confirmation slot” of the guidance sentence template may be prepared in addition to the above. A guidance sentence template for “question slot” and “confirmation slot” alone may be prepared.
[0052]
Furthermore, in the above embodiment, the implicit confirmation was performed by putting the immediately preceding recognition result into the next guidance sentence. However, depending on the contents of the intended answer, the question and confirmation may be performed for each slot. Dialogue control may be performed. In this case, the prepared guidance sentence template prepares a plurality of question guidances and confirmation guidances for each target slot. Further, any other necessary guidance may be provided.
[0053]
Furthermore, in the above-described embodiment, an example is shown in which control is performed for implicitly confirming one slot for which an answer was obtained immediately before and for determining an answer for the next one slot. The number of slots for which an answer is required is not limited to one. If necessary, each may be singular or plural. Further, the control may be a control for requesting confirmation only for at least one slot, or a control for requesting only an answer for at least one slot.
[0054]
Further, in the above embodiment, the guidance sentence also shows one example for each of the slot for asking for confirmation and the slot for asking for an answer. Further, any guidance sentence corresponding to the control of the question slot may be used. Therefore, guidance sentences corresponding to at least one slot for each of a question and a confirmation may be prepared as needed.
[0055]
Furthermore, a plurality of confirmation sentence slots and question slots may be prepared as needed.
[0056]
The above-described embodiment is an example of the present invention, and the present invention is not limited to the embodiment. Various modifications can be considered in light of the essence of the present invention.
[Brief description of the drawings]
FIG. 1 is a configuration diagram illustrating a logical configuration of a voice interaction device according to an embodiment of the present invention.
FIG. 2 is an explanatory diagram of a slot table used by a word determination unit 132, an answer sentence re-request unit 141, and a guidance generation unit 150.
FIG. 3 is an explanatory diagram of a guidance sentence template used by a guidance generation unit 150.
FIG. 4 is a flowchart illustrating a procedure of a process executed by the voice interaction device 100;
[Explanation of symbols]
100… voice interaction device
110 ... voice input unit
120… voice recognition unit
130… semantic understanding
131… word extraction unit
132… word judgment unit
140… Dialogue control unit
141… answer sentence re-request section
150 ... Guidance generation unit
160 ... audio output unit
170… Database
171: Guidance storage unit

Claims (6)

A voice recognition device that outputs a guidance sentence to obtain a desired answer from a speaker, analyzes a response sentence obtained from the speaker in response to the guidance sentence, and determines a desired answer.
From the answer sentence, a determining means for determining whether or not the answer intended by the guidance sentence has been obtained,
If it is determined by the determining means that the target answer was not obtained, another different guidance sentence for obtaining the target answer is output by voice, and the answer requesting the speaker again for the answer sentence A spoken dialogue apparatus comprising: a sentence re-requesting unit. The guidance sentence is a sentence that prompts a speaker for an answer sentence in order to obtain a target answer in at least one slot provided for each different category. The voice interaction device according to claim 1, wherein the guidance sentence is output by voice. Storage means for storing a plurality of guidance sentences for each of the at least one slot, wherein the answer sentence re-requesting means sequentially selects a plurality of guidance sentences corresponding to at least one slot for which an answer is sought from the storage means; The speech recognition device according to claim 2, wherein the speech recognition device outputs the speech. The set of a plurality of guidance sentences set for at least one slot for which the answer is requested is composed of a set of guidance sentences having different expressions in order to derive a target answer of the slot. The speech recognition device according to claim 2 or 3. Outputting a guidance sentence to obtain a desired answer from the speaker;
Analyzing the answer sentence obtained from the speaker in response to the guidance sentence, from the answer sentence, a determination procedure to determine whether the intended answer of the guidance sentence has been obtained,
When it is determined that the target answer is not obtained by the above-described determination procedure, another different guidance sentence for obtaining the target answer is generated, the guidance sentence is output as a voice, and the answer is again sent to the speaker. A spoken dialogue program that causes a computer to execute a requested answer sentence re-request procedure. The answer sentence re-requesting step further includes a step of outputting a plurality of different guidance sentences for at least one slot for which an answer is sought in order to obtain a target answer in at least one slot provided for each different category. The speech dialogue program according to claim 5, wherein:

Images