JP2002297374A - Voice retrieving device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a voice retrieving device for enhancing precision in recognizing a voice. SOLUTION: When a conversation start button 14 is depressed, a choice setting part 16 sets a plurality of choices based on data which is stored in a candidate set DB 26 and reports it to a voice recognition processing part 12 and a guide sentence generating part 18. The guide sentence generating part 18 generates a guide sentence for presenting a plurality of choices and outputs them from a speaker 22 through a voice synthesizing part 20. The voice recognition processing part 12 performs a voice recognition processing in an input voice from a user and specifies which one of the choices reported from the choice setting part 16 is selected. The processing is repeated when more choices exist in relation to the specified choice. When the final choice is specified, the contents are reported to an operation instruction output part 30 and, then, a prescribed operation instruction is outputted.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a speech retrieval apparatus for retrieving various information using speech recognition.

[0002]

2. Description of the Related Art Conventionally, there has been known a voice search device which performs voice recognition processing on voice uttered by a user and searches for various information based on a recognition result. Such a voice search device is used in combination with a car navigation device or the like. For example, considering a case where a navigation device executes a function of searching for various facilities in order to be a destination of a route search, the voice search device performs (1) a function to be executed in response to a user's voice. (2) Search for facilities belonging to the facility type specified by the user (for example, a meal place or a gas station),
Information search is performed in such a manner that (3) a facility specified by a franchise name or the like designated by the user is further searched, and (4) one facility finally specified by the user is extracted.

[0003]

By the way, in the conventional voice search device, it is necessary for each user to remember what words can be input to the voice search device by voice. For example, in the above-described example, accurate voice input cannot be performed unless it is known in advance how various functions and facility types are set as recognition targets. However, many users cannot fully understand all the words to be recognized, so they have to try inputting an appropriate word for the time being, and cannot perform voice input using an accurate word. As a result, there is a problem that the accuracy of speech recognition may be reduced.

[0004] The present invention has been made in view of the above points, and an object of the present invention is to provide a voice search device capable of improving the accuracy of voice recognition.

[0005]

In order to solve the above-mentioned problems, a voice search apparatus according to the present invention has a search key associated with each of a plurality of search target items, and the content of a user input voice. When extracting the corresponding item from the search target items by comparing the search key with the search key, the maximum number of character strings that can be used as the search key is set. The pronunciation is output as voice by the recognition target character string output means. Then, predetermined voice recognition processing is performed by the voice recognition processing means on the voice of the user collected by the microphone, and the character string corresponding to this voice is subjected to voice output by the recognition target character string output means. A search target item corresponding to the search key specified by the character string selected from the character strings and selected by the voice recognition processing means is extracted by the item extracting means.

[0006] By outputting voice readings of a plurality of character strings that can be search keys, a character string to be recognized is presented to the user in advance, and the voice input by the user in response to the presentation is output. Since the corresponding character string is selected from the plurality of character strings that have been subjected to voice output and the search key is specified, the accuracy of voice recognition can be improved.

It is preferable that the apparatus further comprises a switch operated before the user speaks, and when the switch is operated, the voice recognition processing by the voice recognition processing means is started. Since the voice recognition processing only needs to be started when the switch is operated, the timing for starting the voice recognition processing becomes clear, and the processing can be simplified.

It is preferable that the apparatus further comprises a selected character string confirming means for outputting a voice of a character string selected by the voice recognition processing means. By outputting the reading of the selected character string by voice, the user can easily confirm the recognition result for the voice input by the user.

Further, when a negative opinion is given by the user with respect to the selection result of the character string by the voice recognition processing means, the reading of the plurality of character strings used to obtain the selection result is repeated. It is desirable to further include a re-selection instruction unit that issues an instruction to output a voice to the recognition target character string output unit. Thus, when a character string different from the one desired by the user is obtained as a selection result, the user can input a search key again by showing a negative opinion.

When the total number of character strings to be recognized exceeds the predetermined maximum number, the recognition target character string output means divides the number of the character strings within a range not exceeding the maximum number into a plurality of times. It is preferable that the reading of the character string is output as voice, and the selection of the character string is determined by the voice recognition processing means for each voice output. Even when there are a large number of character strings to be recognized, voice output is performed in a predetermined number of pieces, so that the user can select a character string by focusing only on the predetermined number of character strings. It is possible to reliably select a desired character string.

[0011] Further, there is further provided a voice output instructing means for instructing the character string output means for recognition to output the second and subsequent voices when a voice output of another selection candidate is instructed by the user. Is desirable. Thus, the user can easily obtain another selection candidate.

[0012] Further, when the user instructs to output the voice again, the re-voice which instructs the recognition target character string output means to again output the reading of the plurality of character strings output immediately before is output again. It is desirable to further include output instruction means. Thus, if the user misses the contents of the audio output, the user can output the audio again to check the content.

[0013] Further, a character string selecting means for selecting a character string without using the result of the voice recognition processing by the voice recognition processing means when an instruction to let the character string selecting operation be given by the user is further provided. When a character string is selected by the character string selecting means, the character string selected by the character string selecting means is used instead of the character string selected by the voice recognition processing means. It is desirable to perform an operation of extracting a search target item by using.
By giving an instruction to “leave”, the user can entrust the selection of a character string to the voice search device, so that the operation can be simplified in a case where any character string may be selected. Is possible.

Also, a plurality of search keys are associated with each of the search target items, and if one search target item cannot be narrowed down by the item extracting means corresponding to one search key, one search key is assigned. It is desirable to repeat the processing by the recognition target character string output unit, the speech recognition processing unit, and the item extraction unit using other search keys until the search target items can be narrowed down. As a result, one search target item can be reliably narrowed down.

Also, different priorities are associated with each of the plurality of search keys, and table information in which a plurality of character strings corresponding to the plurality of search keys are associated with each of the plurality of search target items. Based on the table information stored in the table storage means, based on the table information stored in the table storage means, the recognition target character string output means outputs the reading of the corresponding character string by voice in order from the search key with the highest priority. Is desirable. Since the content can be added or changed for each search key for which the priority is set, the data can be easily updated.

When a character string corresponding to one search key is selected, a search key to be selected next and a plurality of levels of tree structure information indicating a character string corresponding to the search key are stored in a tree. Stored in the structure storage means, and based on the tree structure information stored in the tree structure storage means,
The recognition target character string output means may extract a plurality of character strings corresponding to a search key to be subjected to voice output next, and voice-read the reading of these character strings. By simply tracing the tree structure in order from the upper hierarchy, a character string to be output next as a sound can be extracted, so that the processing can be simplified.

Further, the apparatus further comprises selection history storage means for storing past selection history information by the voice recognition processing means, and a character string having a high selection frequency is selected based on the selection history information stored in the selection history storage means. Is desirably determined by the recognition target character string output means, and the reading of this character string is preferentially output as voice. By preferentially performing voice output for a character string that is frequently selected, a character string that is frequently selected can be selected with a small number of voice inputs, and operability can be improved. .

When each of the plurality of character strings is composed of one of the Japanese 50 sounds, the item extracting means sets the first word to the one sound selected by the speech recognition processing means. It is desirable to extract matching search keys. Even when a large number of character strings to be selected exist, the character strings of the candidates can be easily narrowed down.

It is desirable that the voice recognition processing means selects a character string by comparing all characters constituting the character string with the entire voice recognition processing result. Since it is sufficient to select a character string in consideration of only a character string that completely matches the speech recognition result, comparison processing is facilitated and processing can be simplified.

Further, the voice recognition processing means may select a character string by comparing characters constituting a part of the character string with the entire voice recognition processing result.
By performing comparisons that take into account the characters that make up a part of a character string, in the case where a part of the character string has a characteristic, the user can input only the part that has this characteristic and is easy to remember Thus, operability can be improved.

Further, the voice recognition processing means, when the voice of the user is collected by the microphone before the voice output by the character string output means for recognition ends, starts the character string selecting operation from that point. It is desirable. In the case of, for example, selecting a recognition target character string guided in the first direction in voice output, this desired character string can be voice input without waiting for all voice outputs, thereby further improving operability. Improvement can be achieved.

It is desirable that the maximum number of character strings that can be used as a search key is set in a range of 7 ± 2. According to the theory of short-term memory in cognitive psychology,
If a chunk of information having some kind of unity is defined as a "chunk", the amount of information that a human can hold at one time is about 7 ± 2 chunks. For example, when storing a telephone number, basically, one numeral constituting the telephone number corresponds to one chunk. Further, when the numeral string “2983” is stored in a word-matching manner like “butcher”, the information of “butcher” corresponds to one chunk. Therefore, by setting the maximum number of character strings that can be a search key in the range of 7 ± 2 based on the concept of “chunk”, the user can reliably remember the character strings that can be the search key. be able to. The details of the above-described “chunk” are described in, for example, page 75 of the document “Cognitive Psychology 2 Memory, Yotaro Takano, 1995, University of Tokyo Press”.

Further, the voice search device may be configured by distributing functions to a server and a terminal device connected via a network. Specifically, a function of storing information on search target items and search keys corresponding to the respective search target items is arranged on the server, and the terminal device includes a recognition target character string output unit,
The functions corresponding to the microphone, voice recognition processing means, and item extraction means are arranged, and prior to various processes,
It is preferable that the terminal device acquires necessary information from the server to configure the voice search device. Since the terminal device obtains information necessary for various processes from the server, the terminal device can obtain new information with updated contents from the server by communication and reflect the new information on the various processes.

It is preferable that the information sent from the server to the terminal device is difference information including a change content of the information sent up to the previous time. When the content is changed, only the difference information including the changed content needs to be acquired, and the communication cost can be reduced.

When a function is distributed to a server and a terminal device connected via a network to constitute a voice search device, information on a search target item and a search key corresponding to each are stored, and A server is provided with a function of performing a process of extracting a character string to be output as a sound by the recognition target character string output unit and a process of extracting a search target item by the item extraction unit. , A microphone, and a function corresponding to the voice recognition processing means, and the terminal device may acquire information necessary for these processes from the server. By arranging many functions on the server side, the processing load on the terminal device can be reduced and the configuration can be simplified, so that the cost of the terminal device can be reduced.

[0026]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, a voice search device according to an embodiment of the present invention will be described with reference to the drawings. [First Embodiment] FIG. 1 is a diagram showing the configuration of a vehicle-mounted system including a voice search device 1 according to a first embodiment. The in-vehicle system shown in FIG. 1 includes a voice search device 1 that determines and outputs various operation instructions in an interactive manner in response to a voice uttered by a user, and a vehicle search system that detects a vehicle position and detects a position around the vehicle position. A navigation device 2 for displaying a map, searching for a route to a destination selected by a user, guiding a route, and the like, and an audio device for playing music recorded on a recording medium such as a compact disk or a mini disk. 3 is included.

Next, a detailed configuration of the voice search device 1 will be described. The voice search device 1 shown in FIG. 1 includes a microphone 10, a voice recognition processing unit 12, a dialogue start button 14, a re-request button 15, an option setting unit 16, a guidance sentence generation unit 18,
Voice synthesis unit 20, speaker 22, selection item determination unit 24,
Candidate set database (DB) 26, DB update unit 2
8. It is configured to include the operation instruction output unit 30.

The microphone 10 collects the voice uttered by the user and converts it into a voice signal. Voice recognition processing unit 12
Analyzes a voice signal output from the microphone 10, performs a predetermined voice recognition process, and specifies a character string corresponding to a voice uttered by the user. The voice recognition processing unit 12 according to the present embodiment performs a predetermined recognition process on a character string corresponding to a predetermined number of options set by the option setting unit 16 as a recognition target.

The dialogue start button 14 is a push button switch that is pressed when the user starts a dialogue with the voice search device 1. The re-request button 15 is a push-button switch that is pressed when the user wants to hear the voice output from the voice search device 1 again.

The option setting section 16 stores a candidate set DB 26
Is set on the basis of the data stored in the option. In addition, this predetermined number is 7 ± in one presentation opportunity.
It is desirable to set within two ranges, and in this embodiment, five options are set. Details of the processing performed by the option setting unit 16 will be described later.

The guidance text generator 18 generates the content of the guidance voice to be output to the user, that is, the guidance text, based on the predetermined number of options set by the option setting unit 16. The voice synthesis unit 20 generates a voice signal for outputting a voice corresponding to the guidance text generated by the guidance text generation unit 18, and outputs the voice signal to the speaker 22. Speaker 22
The selection item determination unit 24 that outputs a guidance voice based on the input voice signal is based on the character string of the recognition result output from the voice recognition processing unit 12, and selects any item from a predetermined number of options. Is determined by the user.

The candidate set DB 26 stores the option setting unit 16
Stores the data required to set multiple options. FIG. 2 is a diagram showing the structure of data stored in the candidate set DB 26. As shown in FIG. 2, the candidate set DB 26 stores a predetermined candidate set (tree structure information) having a hierarchical structure. Each candidate set includes a predetermined number of options. The top-level candidate set includes, as options, a plurality of functions that can be executed by the navigation device 2 or the like. In addition, the candidate sets of the second and subsequent hierarchies include:
A plurality of options associated with any of the plurality of options included in the candidate set of the upper hierarchy are included.

FIG. 3 is a diagram showing the correspondence between the upper layer candidate set and the lower layer candidate set in the data structure shown in FIG. For example, the top-level candidate set 100 includes options such as “meal place search”, “service station search”, “facility search”, “parking lot search”, “audio operation”, and “other”. I have. These options are arranged based on a predetermined priority, and when generating a guidance voice for guiding these options, the options are guided in descending order of priority. For example, in the candidate set 100 shown in FIG. 3, “meal place search” has the highest priority, and in the guidance voice generated based on this candidate set 100, “meal place search”, “service station search” Each option is guided in the order of “,...,“ Other ”. A specific example in which each option is guided will be described later. The same applies to other candidate sets.

There is another candidate set 100a at the same level associated with the option "other", and this candidate set 100a includes "traffic information", "map display",
.., And the option “Other” is included. As for “others” of the options included in the candidate set 100a, if there are still other options, a new candidate set is provided and associated with the “others”.

For options other than "other" included in the candidate set 100 and the like, a candidate set including a plurality of options is provided in a lower hierarchy in association with this option. For example, as a candidate set of a lower hierarchy associated with “meal place search” included in the candidate set 100, a candidate set 102 exists. In this candidate set 102, “ A plurality of franchise names, such as restaurant a ", are included as options. Similarly, a candidate set 104 exists as a candidate set of a lower hierarchy associated with “service station search” included in the candidate set 100. In the candidate set 104, in order to select a service station, A plurality of franchise names such as “A Oil” are included as options.

As described above, in the present embodiment, the process of selecting one option and moving to the candidate set of the lower layer associated with the option in order from the candidate set of the highest layer is repeated. Then, the content of the operation instruction is determined by selecting one from a plurality of options included in the candidate set of the lowest hierarchy. In this case, a plurality of options included in the candidate set of the upper hierarchy correspond to “search key”, and a plurality of options included in the candidate set of the lowest hierarchy correspond to “search target item”. Note that FIG. 2 shows a candidate set having a four-layer hierarchical structure, but this is merely an example, and the number of layers may increase or decrease depending on the content of the operation instruction.

The DB update unit 28 obtains the vehicle position detection result from the navigation device 2 and, based on the result, stores data (such as a meal place, a gas station, and a parking lot) stored in the candidate set DB 26 with respect to the location of the facility ( Update the contents of (position data). For example, when the refueling station search is performed and the franchise name of the facility to be searched is selected, the DB update unit 28 searches the store corresponding to the franchise name for the vehicle position at that time. Is extracted, the position data of the extracted store is calculated, and the contents of the candidate set DB 26 are updated.

The operation instruction output unit 30 repeats the process of selecting one of a plurality of options, and outputs a predetermined operation instruction to the navigation device 2 or the audio device in accordance with the content of the finally selected item. Output to 3 The option setting unit 16, the guidance sentence generation unit 18, the speech synthesis unit 20, and the speaker 22 described above are used as a recognition target character string output unit and a selected character string confirmation unit, the speech recognition processing unit 12 is used as a speech recognition processing unit, 16, the selection item determination unit 24 is used as an item extraction unit, the dialogue start button 14 is used as a switch, the re-request button 15 is used as a re-voice output instruction unit, the candidate set DB 26 is used as a tree structure storage unit, and the selection item determination unit 24 is used as a voice. It corresponds to the output instruction means.

The voice search device 1 of the present embodiment has such a configuration, and its operation will be described next. FIG. 4 is a flowchart illustrating an operation procedure of the voice search device 1 according to the first embodiment. An operation procedure when outputting an operation instruction to the navigation device 2 in response to a voice uttered by the user is shown.

The option setting section 16 determines whether or not the user has pressed the dialog start button 14 (step 100). If the dialogue start button 14 has not been pressed, a negative determination is made, and the processing in step 100 is repeated. When the dialogue start button is pressed by the user, an affirmative determination is made, and the option setting unit 16 sets the top-level candidate set as the top candidate set using the data stored in the candidate set DB 26. (Step 101).

Next, the option setting unit 16 notifies the speech recognition processing unit 12 and the guidance sentence generation unit 18 of the character strings corresponding to the plurality of options included in the candidate set (step 10).
2). The guidance text generation unit 18 generates a predetermined guidance text for guiding a plurality of options included in the candidate set, and outputs the guidance text to the speech synthesis unit 20. A voice signal corresponding to the guidance sentence is generated by the voice synthesis unit 20 and output to the speaker 22, and a guidance voice for presenting an option is output from the speaker 22 (step 103).

The option setting unit 16 determines whether the user has pressed the re-request button 15 (step 104). When the re-request button 15 is pressed,
An affirmative determination is made in step 104, the process returns to step 103, and the subsequent processing is repeated. More specifically, the option setting unit 16 indicates that a request has been made to output the guidance message again.
Is notified to the guidance sentence generation unit 18 from In response to this notification, the guidance text generated during the previous processing is transmitted to the guidance text generation unit 18.
Is output to the speech synthesizer 20 again. As a result, the guidance voice is output again.

If the re-request button 15 is not depressed, a negative determination is made in step 104, and the voice recognition processing unit 12 allows the user to input a voice based on the presence or absence of a voice signal output from the microphone 10. It is determined whether or not the operation has been performed (step 105). If no voice input is made, a negative determination is made in step 105, and in this case, the process returns to step 104, and the subsequent processing is repeated.

If voice input is performed, step 1
In step 05, a positive determination is made, and the voice recognition processing unit 12 performs a predetermined voice recognition process on only the character strings corresponding to the plurality of options notified from the option setting unit 16 as a target for voice recognition, and selects the character string by the user. One selected option is specified (step 106). In the present embodiment, in addition to the plurality of options notified from the option setting unit 16, “other” is also assumed to be one of the options for speech recognition.

The selection item judging section 24 includes the voice recognition processing section 1
It is determined whether or not “other” has been selected from the options based on the speech recognition result output from Step 2 (Step 107). If "Other" is not selected,
A negative determination is made in step 107, and the selection item determination unit 2
4 sends an instruction to the option setting section 16 to determine whether or not there is a next candidate set (a lower-level candidate set) corresponding to the option selected by the user (step 108).

If the next candidate set exists, an affirmative determination is made in step 108 and the selection item determination unit 24
Is set in the option setting unit 16 so as to set the next candidate set.
To instruct. The option setting unit 16 that has received the instruction sets the next candidate set (step 109). Thereafter, the process returns to step 102, and the subsequent processing is performed.

If "other" is selected from the options, an affirmative determination is made in step 107 described above, and the selection item determination section 24 sets the option setting section 16 so as to set the next option. Notify. Based on the data stored in the candidate set DB 26, the option setting unit 16 having received the notification determines whether or not the next option exists (step 110).
The next option is set (step 111). Thereafter, the process returns to step 102 described above, and the next option is notified to the voice recognition processing unit 12 and the guidance sentence generation unit 18, and the subsequent processing is performed.

If the next option does not exist,
A negative determination is made in step 110, and in this case, the option setting unit 16 sends an instruction to the guidance sentence generation unit 18 to generate a guidance sentence indicating that there is no next option. Upon receiving the instruction, the guidance sentence generating unit 18 generates a predetermined guidance sentence and outputs it to the voice synthesizing unit 20, and outputs a guidance voice notifying that there is no next option from the speaker 22 (step 112). Then, the above-mentioned step 103
Returning to, the option guided during the previous processing is presented to the user again, and the subsequent processing is repeated.

Also, in the above-described processing for determining whether or not the next candidate set exists in step 108,
If the next candidate set does not exist, a negative determination is made, and the selection item determination unit 24 notifies the operation instruction output unit 30 of the content of the item finally selected by the user.
The operation instruction output unit 30 that has received the notification outputs an operation instruction corresponding to the content of the item selected by the user to the navigation device 2 or the like (step 113).

Next, the dialogue between the voice search device 1 and the user according to the processing shown in FIG. 4 will be described in detail. In the following description, an example of dialogue between the user and the voice search device 1 will be described as “U” and “S”. Also, along with the dialogue example, the candidate set DB
Reference will be made to the drawings showing the contents of the data read from 26 as appropriate.

(Interaction Example 1) Interaction Example 1 shows an example of an interaction when searching for the nearest gas station. FIG. 5 is a diagram showing the contents of data used in the first interactive example. U: The dialogue start button 14 is pressed. S: "Please select from meal place search, gas station search, facility search, parking lot search, audio operation, etc." ... (1) U: "gas station search" ... (2) S: " It ’s a service station search, so the franchise name is A
Please choose from petroleum, petroleum B, petroleum C, petroleum D, petroleum E, etc. "... (3) U:" petroleum B "... (4) S:" it is petroleum B, right 2 km ahead , 2.5km left, 3km left, 5km right, etc. "... (5) U:" 2km ahead right "... (6) S:" 2km ahead right, then B The destination is set at the petroleum Iwaki store. ”(7) As shown in FIG. 5, when the dialogue start button 14 is pressed by the user, the top-level candidate set is read out first, and the user A plurality of selectable functions are guided as in the above-mentioned voice (1).

In response to this voice, the above-mentioned voice (2)
When the user selects "service station search" as shown in (1), a candidate set of a lower hierarchy corresponding to the service station selection is read out, and a plurality of franchise names that can be selected by the user are described in the above-mentioned voice ( You will be guided as in 3).

Here, when the user selects the franchise name "B petroleum" as shown in the above-mentioned voice (4), the candidate set of the lower hierarchy corresponding to this B petroleum is read out, and the own vehicle is read. The position (relative distance) of each facility based on the position is guided as in the above-mentioned sound (5).

Next, as shown in the above-mentioned sound (6),
When the position “2 km ahead right” is selected by the user, “B Petroleum Iwaki Store” which is one gas station corresponding to the selected position is specified, and as shown in the above-mentioned voice (7), This gas station is set as the destination of the route search, and a series of processing ends.

(Interaction Example 2) Interaction Example 2 is a case in which the nearest gas station is searched in the same manner as in Interaction Example 1 described above, and shows an interaction example in which the re-request button 15 is pressed. The contents of the data used in the interactive example 2 are the same as those in FIG. 5 described above. U: The dialogue start button 14 is pressed. S: "Please choose from meal place search, gas station search, facility search, parking lot search, audio operation, etc." U: "gas station search" S: "gas station search, franchise name And A
Please select from petroleum, petroleum B, petroleum C, petroleum D, petroleum E, and others. "U: Press the re-request button 15. S: "Input is a gas station search, then select the franchise name from A, B, C, D, E, etc." ... (8) U: "B Oil: S: "In the case of B oil, please choose from 2km ahead right, 2.5km left, 3km left, 5km right, etc." U: "2km ahead right" S: "2km ahead On the right, let's set the destination at B Oil Iwaki Store. ”As shown in the voice (8) in the dialogue example above, when the user presses the re-request button 15, the candidate that was guided immediately before The contents of the set are presented again.

In this case, it is desirable to change the contents of the guide text at the first time and the second time. In the above example, the first guidance sentence is "Search for gas stations, then ..."
The second guidance sentence is "The input is a gas station search, followed by ...", and the contents of both are changed. In addition, since it is conceivable that the guidance voice is difficult to hear, if the request is made again, the utterance speed of the second voice may be made slower than the first voice.

(Interaction Example 3) Interaction Example 3 is a case in which the nearest gas station is searched for in the same manner as in the above-mentioned Dialog Example 1, and shows an interaction example in which "Other" is selected from the options. ing. FIG. 6 is a diagram showing the contents of data used in Dialogue Example 3. U: The dialogue start button 14 is pressed. S: "Please choose from meal place search, gas station search, facility search, parking lot search, audio operation, etc." U: "gas station search" S: "gas station search, franchise name And A
Please select from petroleum, petroleum B, petroleum C, petroleum D, petroleum E, or other. "U:" Other "... (9) S:" So, Petroleum F, Petroleum G, Petroleum H, Petroleum I, Please select from J Petroleum and others .... (10) U: "G Petroleum" S: "It's G Petroleum, 2km ahead right, 2.5km left, 3km ahead left, 5km ahead right, etc. U: "2 km ahead on the right" S: "2 km ahead on the right, then set the destination at G Petroleum Iwaki Store" As shown in the voice (9) in the above dialogue example When the user selects “other” from the options, a candidate set including the next option in the same layer is read out in response to the “other”, and the above-described voice (1)
As shown in (0), a franchise name that can be selected by the user is additionally provided.

(Interaction Example 4) In Interaction Example 4, similarly to Interaction Example 3 described above, when "Other" is selected from the options, the next candidate set in the same hierarchy does not exist. The example of the dialog in the case is shown. FIG. 7 is a diagram showing the contents of data used in Dialogue Example 4. U: The dialogue start button 14 is pressed. S: "Please choose from meal place search, gas station search, facility search, parking lot search, audio operation, etc." U: "gas station search" S: "gas station search, franchise name And A
Please select from petroleum, petroleum B, petroleum C, petroleum D, petroleum E, or other. "U:" other "S:" in, petroleum F, G, H, I, J, etc. U: “Other” ... (11) S: “I ’m sorry, there are no other candidates”… (12) As shown in the speech (11) in the above dialogue example If “other” is selected from the options by the user and there is no next candidate set in the same layer, as shown in the above-mentioned voice (12), the user can select Will be notified that no longer exists.

(Interaction Example 5) Interaction Example 5 shows an example of an interaction when searching for a desired facility. FIG. 8 shows dialogue example 5
FIG. 4 is a diagram showing the contents of data used in the method. U: The dialogue start button 14 is pressed. S: "Please select from meal place search, gas station search, facility search, parking lot search, audio operation, etc." U: "facility search" ... (13) S: "facility search Please select the region of the facility from Hokkaido, Tohoku region, Kanto region, Chubu region, Kinki region and others. "... (14) U:" Tohoku region "... (15) S:" Tohoku region Then, please select the prefecture of the facility from Fukushima prefecture, Akita prefecture, Iwate prefecture, Miyagi prefecture, Aomori prefecture, and others .... (16) U: "Fukushima prefecture" ... (17) S: "Fukushima It is a prefecture, so please select the first letter of the facility name from A line, ka line, sa line, ta line, na line, etc. "... (18) U:" A line "... (19) S: "Ah, then, R pie, R pie Giken, R pie information system, R pi establishment, R pi distribution, Please choose from the other. "... (20) U:" R pie "... (21) S:" R pie, then set the destination on R pie "... (22) Like this A plurality of selectable functions are guided by voice, and the user selects a target function according to the voice. As shown by the above-mentioned sound (13), when "facility search" is selected by the user, a candidate set of a lower hierarchy corresponding to this facility search is read, and the above-mentioned sound (1) is selected.
As shown in 4), the region where the facility is located is guided as an option.

Here, as shown by the above-mentioned voice (15), the user locates the facility in the "Tohoku district" as the district where the facility is located.
Is selected, the candidate set of the lower hierarchy corresponding to the "Tohoku region" is read, and the name of the prefecture where the facility is located is guided as in the above-mentioned voice (16).

When the user selects "Fukushima prefecture" as shown in the above-mentioned voice (17), the corresponding lower-level candidate set is read out, and as shown in the above-mentioned voice (18), The first character of the name of the facility (a line, line, etc.) is provided as an option. As shown in the above-mentioned sound (19), when the user selects "A line", the corresponding lower-level candidate set is read out and the above-mentioned sound (2) is selected.
As shown in (0), the name of a facility located in "Fukushima Prefecture" and whose first character of the facility name belongs to "A line" is guided as an option.

Here, as shown in the above-mentioned sound (21), when one facility name “R pie” is selected by the user, one facility “R pie” is specified. As shown in the sound (22), this facility is set as the destination of the route search, and a series of processing ends.

As described above, in the first embodiment, data including a candidate set having a predetermined hierarchical structure is stored in the candidate set DB 26, and based on this candidate set, the next audio output target A plurality of character strings corresponding to certain options are extracted, and the reading of these character strings is output as voice. Then, a character string to be recognized is presented to the user in advance, and a character string corresponding to a voice input by the user in response to the presentation is selected from a plurality of character strings targeted for voice output. Since the option is selected and the option selected by the user is specified, the accuracy of voice recognition can be improved. In particular, since it is possible to extract the character string to be output next by simply tracing the candidate set having the hierarchical structure in order from the upper layer, the processing can be simplified. [Second Embodiment] In the above-described first embodiment, candidate sets having a hierarchical structure are prepared and stored in advance in the candidate set DB 26. However, in the first embodiment, a database having a general table format structure is used. The same processing as described above can be performed.

FIG. 9 shows a voice search device 1 according to the second embodiment.
FIG. 1 is a diagram showing a configuration of an on-vehicle system configured including A. The voice search device 1A of the second embodiment shown in FIG.
Is different from the voice search device 1 according to the first embodiment in that the candidate set DB 26 is replaced with a candidate set DB 26a having a different data content, and with the change in the data content, The operation procedure for narrowing down the contents of the operation instruction according to the voice uttered by the user is different. Hereinafter, the description will be given focusing on differences from the first embodiment.

The candidate set DB 26a stores the option setting unit 1
6 stores data required for setting a plurality of options. FIG. 10 shows a candidate set DB according to the second embodiment.
It is a figure showing the structure of the data stored in 26a. Figure 1
0, as shown in FIG.
The data stored in a is in the form of a table, unlike the case of the first embodiment described above.

The data (table information) stored in the candidate set DB 26a is composed of three elements: "priority", "candidate set title", and "option". “Priority” has the same meaning as the hierarchy in the first embodiment described above. That is, when deciding any operation instruction, one option is selected from a plurality of options in order from the “function” of priority 1. A specific example of a process for presenting and selecting options will be described later. Note that each option associated with the priority 1 to 6 shown in FIG. 10 corresponds to the “search key”, and each option of priority 7 which is the finally specified option is set as the “search target item”. Yes, it is.

In the candidate set DB 26a, one row in the horizontal direction forms one data group (hereinafter, this is referred to as a "record"). For example, the record on the first line shown in FIG. 10 is a data group related to the facility name “B Petroleum Iwaki Store”, and is related to “Fuel Station Search” as a function, and the franchise name is “B Petroleum”. This indicates that the region where the facility is located belongs to "Tohoku region", the prefecture where the facility is located belongs to "Fukushima prefecture", and the first letter of the facility name belongs to "ha line". The content of the position is updated by the above-described DB updating unit 28. The candidate set DB 26a includes a plurality of such records. The candidate set DB 26a corresponds to a table storage.

The voice search device 1A of the present embodiment has such a configuration, and its operation will be described next.
FIG. 11 is a flowchart showing a partial operation procedure of the voice search device 1A of the second embodiment. Note that the music search device 1
The basic operation procedure of A is the same as that of the music search apparatus 1 of the first embodiment shown in FIG.
1 is different from the processing contents after step 107. FIG. 11 mainly shows the difference between the processing contents.

The option setting unit 16 determines whether or not the user has pressed the dialog start button 14 (step 100). If the dialogue start button 14 is not pressed by the user, a negative determination is made, and the process of step 100 is repeated. If the dialogue start button 14 is pressed, an affirmative determination is made in step 100, and the option setting unit 16 extracts data belonging to the column of “priority 1” from the candidate set DB 26a and uses the extracted data. A first candidate set is set (step 101A). Specifically, as shown in FIG. 10, in the present embodiment, the data of the column of “priority 1” includes the contents of various functions,
A candidate set including the contents of these functions as options is set. Thereafter, the processing shown in Steps 102 to 107 shown in FIG. 4 is performed in the same manner as in the first embodiment.

The selection item judging section 24 includes the voice recognition processing section 1
It is determined whether or not “other” has been selected from the options based on the speech recognition result output from Step 2 (Step 107). If "Other" is not selected,
A negative determination is made in step 107, and the selection item determination unit 24 notifies the option setting unit 16 of that. The option setting unit 16 that has received the notification narrows down options that are candidates to be presented as the next candidate set, in accordance with the options selected by the user (step 120).
For example, if “search for gas station” is selected by the user, the selection item determination unit 24 narrows down records corresponding to “search for gas station”.

Next, the option setting section 16 determines whether or not there is a candidate set including two or more options in order from the candidate set having the highest priority (step 121). 2 choices
If there is a candidate set including one or more, an affirmative determination is made in step 121, and then the option setting unit 16
A predetermined number of options are extracted corresponding to the candidate set (candidate set including two or more options) specified in step 121, and the next candidate set is set (step 12).
2).

FIG. 12 is a flowchart showing a detailed procedure of the process shown in step 122. First, the option setting unit 16
Selects a candidate set having two or more alternatives of high priority and different types based on the data stored in the candidate set DB 26a (step 130).

Next, the option setting section 16 determines whether or not the number of options included in the selected candidate set is equal to or less than a predetermined number (five in this embodiment) (step 13).
1). If the type of option is not equal to or less than the predetermined number, a negative determination is made in step 131, and then the option setting unit 16
Extracts a predetermined number of options (step 132).

If the number of options is equal to or less than the predetermined number, an affirmative determination is made in step 131, and then the selection system setting unit 16 extracts all the available options (step 133). Next, the option setting unit 16 proceeds to step 13
2 or the options extracted in the processing shown in step 133 are set as the next candidate set (step 1).
34), the processing in step 122 shown in FIG. 11 ends. Thereafter, the process returns to step 102, and the subsequent processing is repeated.

In step 107 described above, if “other” is selected from the options, an affirmative determination is made, and the selection item determination unit 24 notifies the option setting unit 16 to set the next option. I do. Based on the data stored in the candidate set DB 26a, the option setting unit 16 that has received the notification determines whether there is another option other than the option already presented in the previous process (step 123). If there is, a positive determination is made and the next option is set (step 124). Thereafter, the process returns to step 102 described above, and the next option is notified to the voice recognition processing unit 12 and the guidance sentence generation unit 18, and the subsequent processing is performed.

If the next option does not exist,
In step 123, a negative determination is made. In this case,
The option setting unit 16 sends an instruction to the guidance message generation unit 18 to generate a guidance message for guiding that there is no next option. Upon receiving the instruction, the guidance sentence generation unit 18 generates a predetermined guidance sentence and outputs it to the speech synthesis unit 20, and outputs a guidance voice notifying that there is no next option from the speaker 22 (step 125). Thereafter, the process returns to the above-described step 103, where the option guided during the previous processing is presented to the user again, and the subsequent processing is repeated.

In step 121 described above, when there is no longer any candidate set including two or more options, a negative determination is made, and the selection item determination unit 24 determines whether the option finally selected by the user is Operation instruction output unit 3 for contents
Notify 0. The operation instruction output unit 30 that has received the notification outputs an operation instruction corresponding to the content of the option selected by the user to the navigation device 2 or the like (step 12).
6).

Next, the dialogue between the voice search device 1A and the user will be specifically described according to the processing shown in FIG. 11 described above, and together with this dialogue example, the dialogue is stored in the candidate set DB 26a. The manner in which a necessary record is extracted from the collected data will be described with reference to the drawings as appropriate.

(Interaction Example 6) Interaction Example 6 shows an example of dialog when searching for the nearest gas station. FIG. 13 is a diagram illustrating the contents of a record extracted from the candidate set DB 26a in the interactive example 6. U: The dialogue start button 14 is pressed. S: "Please select from meal place search, gas station search, facility search, parking lot search, audio operation, etc." ... (23) U: "gas station search" ... (24) S: " It ’s a service station search, so the franchise name is A
Please choose from petroleum, petroleum B, petroleum C, petroleum D, petroleum E, etc. "... (25) U:" petroleum B "... (26) S:"it's B petroleum, right 2km ahead , 5km left,
Other, please choose from ... "(27) U:" 2 km ahead right side "... (28) S:" 2 km ahead right side, then set the destination at B Yukino Iwaki store "... (29) 13) As shown in FIG.
Is pressed, a plurality of options are first extracted corresponding to the "function" which is a candidate set title of priority 1, and a plurality of functions that can be selected by the user are guided as in the above-mentioned voice (23). Is done.

In response to the voice (23), when the user selects "fuel station search" as shown in the voice (24), only records corresponding to the "fuel station selection" are narrowed down. Then, a plurality of options are extracted in accordance with the “franchise name” which is the next highest priority candidate set title of priority 2, and a plurality of franchise names that can be selected by the user are described in the above-mentioned voice (25). You will be guided as follows.

When the user selects the franchise name "B petroleum" as shown in the above-mentioned voice (26),
Only records corresponding to this "B Petroleum" are narrowed down,
Next, a plurality of alternatives are extracted corresponding to the priority “position” of 6 which is a candidate set title having a higher priority and including two or more types of alternatives, and the position of the own vehicle is set as a reference. The positions (relative distances) of the respective facilities are guided as in the above-mentioned sound (27).

As shown in the above-mentioned sound (28), when the user selects the position "2 km ahead right", one of the refueling stations corresponding to the selected position, "B Petroleum Iwaki Store", is opened. The specified gas station is set as the destination of the route search as indicated by the voice (29) described above, and a series of processing ends.

Note that when the re-request button 15 is pressed, the dialog example 2 in the first embodiment described above is performed.
Is performed, and the contents of the data used in this case are the same as those shown in FIG. (Interaction Example 7) Interaction Example 7 is a case in which the nearest gas station is searched in the same manner as in the above-described Interaction Example 6, and shows an interaction example in which “Other” is selected from the options. FIG. 14 is a diagram illustrating the contents of the record extracted from the candidate set DB 26a in the interactive example 7. U: The dialogue start button 14 is pressed. S: "Please choose from meal place search, gas station search, facility search, parking lot search, audio operation, etc." U: "gas station search" S: "gas station search, franchise name And A
Please select from petroleum, petroleum B, petroleum C, petroleum D, petroleum E, or other. "U:" Other "... (30) S:" Now, select from petroleum F, petroleum G, or other Please ... "(31) U:" G Petroleum "... (32) S:"It's G Petroleum, let's set the destination at G Petroleum Iwaki Store "... (33) Voice in Dialogue Example 7 above As shown in (30),
When the user selects “other” from the options, the petroleum A, petroleum B, petroleum C, petroleum D,
Records excluding E oil are narrowed down, and a franchise name that can be selected by the user is additionally provided as shown in the above-mentioned voice (31).

When the user selects the franchise name "G petroleum" as shown in the above-mentioned voice (32),
Only records corresponding to this "G petroleum" are narrowed down. In this case, at the time of narrowing down based on the franchise name, one record is narrowed down,
Since the “G oil Iwaki store” which is one of the gas stations to be guided is specified, as shown in the above-mentioned voice (33),
This gas station is set as the destination of the route search, and a series of processing ends.

(Dialogue Example 8) In Dialogue Example 8, similarly to Dialogue Example 7 described above, when “Others” is selected from the options and there is no next presentable option. An example of the dialogue is shown. FIG. 15 is a diagram illustrating the contents of the record extracted from the candidate set DB 26a in the interactive example 8. U: The dialogue start button 14 is pressed. S: "Please choose from meal place search, gas station search, facility search, parking lot search, audio operation, etc." U: "gas station search" S: "gas station search, franchise name And A
Please select from petroleum, petroleum B, petroleum C, petroleum D, petroleum E, etc. "U:" other "S:" Please select from petroleum F, petroleum G, other, etc. " U: "Other" ... (34) S: "Sorry, there are no other candidates" ... (35) As shown in the voice (34) in the above dialogue example, the user selects "Other" from among the options. Is selected, and when there is no option that can be presented next, as shown in the above-mentioned voice (35), the user is notified that there are no more options that can be selected by the user. You.

As described above, in the second embodiment, predetermined table information having a predetermined table format, in which a plurality of options are associated with different priorities, is stored in the candidate set DB 26a. Based on the information,
The reading of the corresponding character string is output as voice in order from the option with the highest priority. A character string to be recognized is presented to the user in advance, and a character string corresponding to the voice input by the user in response to the presentation is selected from a plurality of character strings to be output. Since the option selected by the user is specified, the accuracy of voice recognition can be improved. Particularly, since data is stored in a table format, there is an advantage that records can be easily added or changed.

[Modifications] The present invention is not limited to only the above-described embodiments, and various modifications can be made within the scope of the present invention. For example, in the above-described embodiment, the user sequentially selects any one of a plurality of options to be presented to select the final option, and the operation instruction corresponding to the content is determined by navigation. Although the processing has been performed on the device 2 and the like, the option may be automatically selected when the user desires.

FIG. 16 is a diagram showing the configuration of a voice search device according to a modification in which options are automatically selected. Figure 1
6 is different from the voice search device 1 in the first embodiment described above in comparison with the selection frequency learning unit 3.
2 in that a learning result storage unit 34 is added. Hereinafter, the configuration and operation will be described, mainly focusing on differences from the first embodiment.

The selection frequency learning section 32 learns the frequency of selection by the user for a plurality of options presented to the user. The learning result storage unit 34 includes the selection frequency learning unit 3
2 is stored. In the present embodiment, when performing a voice input for selecting an option, the user inputs “leave it”, so that the option is automatically selected. When this "leave" is input, the selection item determination unit 24 automatically selects an option with a high selection frequency in the past using the learning result stored in the learning result storage unit 34, and outputs an operation instruction. Content is determined. In this case, the selection item determination unit 24 corresponds to a character string selection unit.

The voice search device 1B has such a configuration. Next, an operation in the case where an option is automatically selected according to the selection frequency of past options will be described. Figure 1
FIG. 7 is a flowchart showing a partial operation procedure of the voice search device 1B in a case where an option is automatically selected according to the selection frequency of a past option. Note that the basic operation procedure of the music search device 1B is the same as that of the music search device 1 of the first embodiment shown in FIG. 4 described above, and the processing contents after step 107 are different. FIG. 17 mainly shows the difference between the processing contents.

[0091] The selection item judging section 24 includes the voice recognition processing section 1
It is determined whether or not “other” has been selected from the options based on the speech recognition result output from Step 2 (Step 107). The process when "other" is selected is the same as that of the music search device 1 of the first embodiment shown in FIG. 4, and the description is omitted.

If "other" is not selected,
A negative determination is made in step 107, and then the selection item determination unit 24 determines whether “leave” is selected from the options based on the voice recognition result output from the voice recognition processing unit 12. (Step 140).

If "leave" is selected from the options, an affirmative determination is made in step 140, and the selection item determination unit 24 reads the learning result stored in the learning result storage unit 34, and An option is automatically selected based on the selection frequency (step 141). For example, in the present embodiment, all options up to the final option are automatically selected.

When the final option is automatically selected according to the past selection frequency, or when a negative determination is made in step 108, the selection item determining unit 24 operates the final option. The instruction output unit 30 is notified. Upon receiving the notification, the operation instruction output unit 30 sends an operation instruction corresponding to the content of the item selected by the user to the navigation device 2.
And the like (step 113).

Next, the dialog between the voice search device 1B and the user according to the processing shown in FIG. 17 will be described in detail. (Dialogue Example 9) Dialogue Example 9 is a case in which the nearest gas station is searched in the same manner as Dialogue Example 1 described above, and shows a dialogue example in a case where “leave” is selected as an option.
Note that the contents of the data used in the dialogue example 9 are the same as those in FIG. 5 described above. U: The dialogue start button 14 is pressed. S: "Please choose from meal place search, gas station search, facility search, parking lot search, audio operation, etc." U: "gas station search" S: "gas station search, franchise name And A
Please choose from petroleum, petroleum B, petroleum C, petroleum D, petroleum E, etc. Or say "leave it" .... (36) U: "leave it" ... (37) S: "I can leave it to you, then set your destination at B Oil Iwaki Store." As shown in the voice (36), a new option "leave it" is added to the user. On the other hand, as shown in the voice (37), when “leave” is selected by the user, the option with the highest selection frequency is automatically selected according to the past selection frequency. In the example described above, the options after the franchise name are automatically selected according to the past selection frequency. Specifically, “B Oil” is automatically selected as the franchise name, and “2 km ahead right” is automatically selected as the location, so that the option “B Oil Iwaki Store” is finally selected. Selected.

In the example described above, all the options are automatically selected up to the final option. However, only the option at that time may be automatically selected. For example, if “leave” is selected when selecting a franchise name, only the franchise name is automatically selected, and the process moves to the lower-level candidate set “position”, A plurality of included options may be presented. Further, in the above-described example, a case has been described in which processing for automatically selecting an option is added to the first embodiment, but this function may be added to the second embodiment in the same manner. it can.

Further, in the above-described modified example, when “leave” is selected as an option, the option is selected according to the past selection frequency. However, the option is randomly selected regardless of the past selection frequency. You may make it. In this case, the process of learning the past selection frequency becomes unnecessary, and the configuration can be simplified.

When performing a process of learning the past selection frequency of an option, this learning result is stored in the candidate set DB 26.
(Or 26a). For example, in each of the embodiments described above,
When guiding a plurality of options related to the franchise name, “G Petroleum” was presented at the second guidance performed when “Other” was selected from the options presented at the first time. However, if a learning result that “G oil” is selected with high frequency is obtained, this “G oil” may be presented at the first guidance. Alternatively, even among a plurality of options included in one guidance, the guidance order may be changed according to the past selection frequency. For example, 1 in the initial state
In the second guidance, if the guidance was in the order of petroleum A, petroleum B, petroleum C,..., The learning result shows that the selection frequency was high in the order of petroleum G, petroleum B, petroleum A,. If it has been obtained, the first guidance may be changed to G petroleum, B petroleum, A petroleum, and so on. In this case, the learning result storage unit 34 corresponds to a selection history storage unit.

In each of the above-described embodiments, the re-request button 15 is pressed to hear the guidance voice again. However, the re-request operation may be performed by voice input. In this case, for example, a voice such as "again" may be input, and the immediately preceding guidance content may be output again in response to these voices.

In each of the above-described embodiments, the operation in the case where the once selected option is canceled and a new option is selected has not been described, but such processing may be performed. FIG. 18 is a flowchart partially showing an operation procedure of the voice search device in a case where a once-selected option is canceled and a new option is selected. For example,
The description will be made assuming that this processing is performed in the voice search device 1 described in the first embodiment. The basic operation procedure in this case is the same as the flowchart shown in FIG. 4 described above, and a new process is added after step 107. FIG. 18 mainly shows newly added processing contents. In this modification, the selection item determination unit 24 corresponds to a reselection instruction unit.

The selection item judging section 24 includes the voice recognition processing section 1
It is determined whether or not “other” has been selected from the options based on the speech recognition result output from Step 2 (Step 107). The process when “other” is selected by the user is the same as that in FIG. 4 described above, and a description thereof will be omitted.

If "other" is not selected,
A negative determination is made in step 107, and then the selection item determination unit 24 determines whether or not a voice of “correction” has been input based on the voice recognition result output from the voice recognition processing unit 12 (step S107). 150). Specifically, by giving a negative opinion by the voice input of “correction”, a process of canceling the once selected option is performed. "Back" instead of "Modify"
A negative opinion may be shown by performing a voice input such as “No” or “No”.

When the voice of "correction" is input, an affirmative determination is made in step 150, and the selection item determination unit 24 instructs the option setting unit 16 to select a value higher than the level of interest at that time. It instructs to set the candidate set of one higher hierarchy again. In response to this instruction, the option setting unit 16 sets the candidate set of the upper hierarchy again (step 151). Thereafter, step 1 shown in FIG.
02, the character string corresponding to the option included in the candidate set in the upper hierarchy is notified as a recognition target,
This option is output as voice, and the subsequent processing is repeated.

If the voice of "correction" has not been input, a negative determination is made in step 150. In this case, the process proceeds to step 108, and the subsequent processing is performed. Next, a dialog performed between the voice search device 1 and the user according to the above-described processing illustrated in FIG. 18 will be specifically described.

(Dialogue Example 10) Dialogue Example 10 is a case in which the nearest gas station is searched in the same manner as in Dialogue Example 1 and the like described above, and shows a dialogue example in which a voice “correct” is input. I have. U: The dialogue start button 14 is pressed. S: "Please choose from meal place search, gas station search, facility search, parking lot search, audio operation, etc." U: "gas station search" S: "gas station search, franchise First name, A
Please select from petroleum, petroleum B, petroleum C, petroleum D, petroleum E, etc. "U:" Petroleum A "... (38) S:"It's E petroleum, 2 km ahead on the right, 2.5 km Please choose from the left side, 3 km ahead, 5 km left, etc. "... (39) U:" Modify "... (40) S:" Then the franchise name is A, B, C, Please choose from D oil, E oil, etc. "... (41) U:" A oil "S:" It is A oil, 2 km ahead on the left, 5 km left,
Others, please choose from among them. "U:" 2 km ahead on the left "S:" 2 km ahead on the left, then set the destination at A Oil Iwaki Store "In the above dialogue example 10, voice (38) Although the user has selected "A petroleum" from the options as shown in Fig. 7A, an erroneous recognition has occurred as shown in the voice (39), and "E petroleum" has been selected. . In this case, as shown in the voice (40), the user performs "correction" and voice input, and as shown in the voice (41), based on the franchise name which is a candidate set of the upper hierarchy, The franchise name that can be selected is presented again.

(Interaction Example 11) Interaction Example 11 shows another example of the dialog when the voice of "correction" is input. U: The dialogue start button 14 is pressed. S: "Please choose from meal place search, gas station search, facility search, parking lot search, audio operation, etc." U: "gas station search" S: "gas station search, franchise First name, A
Please select from petroleum, petroleum B, petroleum C, petroleum D, petroleum E, etc. "U:" Facility search "... (42) S:"It's C petroleum, 2 km ahead on the right, 2.5 km Please choose from the left side, 3 km ahead right side, 5 km left side, etc. "... (43) U:" Modify "... (44) S:" Then the franchise name is A, B, C, Please select from D Petroleum, E Petroleum, etc. "... (45) U:" Modify "... (46) S:" Mealing place search, gas station search, facility search, parking lot search, audio operation, Other, please select from among. "(47) U:" Facility search "Hereinafter, the dialogue after selecting the facility search is performed in the same manner as in the above-described dialogue example 5, and the description will be given here. Omitted.

In the above-described dialogue example 11, the user once selected “service station search” but wanted to select “facility search”. "Facility search", which is a different option from the selected option, is input by voice. Even in this case, as shown in the voice (43), the voice search device 1 selects the one closest to the input voice from the options at that time and continues the process.

As shown in voice (44), when the user inputs "correction" and voice, as shown in voice (45), based on the franchise name which is a candidate set of the upper hierarchy, The franchise name that can be selected is presented again. As shown in the voice (46), the user further inputs "correction" and voice, and as shown in the voice (47), based on the "function", which is a higher-level candidate set. Then, the options that can be selected by the user are guided again.

By the way, in the above-mentioned dialogue example 11, even when a voice input not conforming to the contents of the plurality of options presented is performed, one of the plurality of options at that time is selected. However, if a voice input that does not conform to the content of the option is made and it is difficult to specify the option, it may be notified that an effective recognition result has not been obtained. In this case, the validity of the recognition result is determined in the processing of step 106 shown in FIG. 4 described above, and if the matching rate of the character strings is extremely low (for example, 10% or less), the option can be specified. What is necessary is just to notify that it did not exist. In the following, the dialogue when a valid recognition result is not obtained will be specifically described.

(Dialogue Example 12) U: The dialogue start button 14 is pressed. S: "Please choose from meal place search, gas station search, facility search, parking lot search, audio operation, etc." U: "gas station search" S: "gas station search, franchise First name, A
Please select from petroleum, petroleum B, petroleum C, petroleum D, petroleum E, or other. "U:" H petroleum "... (48) S:" Sorry. The input word could not be recognized. The franchise names are A Petroleum, B Petroleum,
Please select from C oil, D oil, E oil, and others. ”(49) In the above dialogue example 12, as shown in the voice (48),
Since “H Petroleum”, which is not presented as an option, is selected by the user, the voice search device 1 cannot obtain a valid recognition result. Therefore, the voice search device 1
As shown in the voice (49), the user is notified that the option could not be specified, and guidance for prompting the user to input the option again is provided.

Further, when performing the voice recognition processing, a partial match between the character string corresponding to the plurality of presented options and the character string corresponding to the voice input by the user is taken into consideration.
The recognition accuracy may be increased. For example, it is conceivable that, when selecting the “service station search” presented as an option, some users may utter only the “service station” part. In such a case, if only the entire character string “Search for gas station” is targeted for speech recognition, the match rate is low because it only partially matches with “Gas station” spoken by the user, Of course, since the matching rate is low with other options (such as “meal place search”), it may be difficult to accurately specify the option selected by the user. Therefore, for example, for a gas station search, the character strings to be recognized are “gas station search” and “gas station”, and for a meal place search, “meal place search”, “meal place”, “meal”, etc. Since it is possible to determine both a general match and a partial match between a character string corresponding to a plurality of options and a character string for a voice input by a user, recognition accuracy can be improved.

Even when such partial matching is considered, it is preferable to output the entire character string when replying to the recognition result. For example, even when the user inputs "fuel station", it is preferable to reply the entire character string, such as "fuel station search," as a response to the corresponding recognition result.

In each of the above-described embodiments, a plurality of options are presented and one of them is input by voice. However, a predetermined code is added to each option to present the desired option. May be input by voice. Specifically, the predetermined code is “1, 2,
.. And letters such as "A, B, C,...". For example, in the case of presenting a plurality of functions as options, "1: meal place search, 2: gas station search, 3: facility search, 4: parking lot search, 5: audio operation, 6: other, etc. Please select a corresponding number from "" and a user may input any number from 1 to 6 by voice. As described above, when a predetermined code is used, the user only has to utter a code associated with a desired option, and voice input can be further simplified. In addition, since the character string to be subjected to the voice recognition processing can be a simple character string such as a numeral, the recognition accuracy can be improved.

In each of the above-described embodiments, the user selects one option and performs voice input after all of the plurality of options are presented. If a voice input is made by the user at the time, the voice recognition processing may be started at that time. When listening to the output voice guidance, some users may start voice input immediately after a desired option is output. In such a case, the operability can be further improved by immediately starting the voice recognition process even after all the options are not presented.

Further, in each of the above-described embodiments, the voice search device used alone has been described. However, the functions are distributed to the server and the terminal device connected via the network to configure the voice search device. May be. FIG.
FIG. 3 is a diagram illustrating a configuration example of a voice search device in a case where functions are distributed to a server and a terminal device connected via a network. The voice search device shown in FIG. 19 includes a voice search terminal device 4 and a server 5 connected via a predetermined network 6.

The voice search terminal device 4 has basically the same configuration as the voice search device 1A in the second embodiment described above, except that a communication processing unit 36 is added. The configuration of the voice search terminal device 4 may be the same as the voice search device 1 in the first embodiment described above.

The communication processing unit 36 provided in the voice search terminal device 4 performs a communication process for acquiring information necessary for updating data stored in the candidate set DB 26a from the server 5 via the network 6. . DB update unit 2
8 updates the data stored in the candidate set DB 26a based on the information received by the communication processing unit 36. This updating process is performed prior to the predetermined process by the voice search terminal device 4.

The server 5 includes a server control unit 50, a candidate set DB 52, and a communication processing unit 54. The server control unit 50 controls the overall operation of the server 5. The candidate set DB 52 stores basically the same content data as the candidate set DB 26a provided in the voice search terminal device 4 described above. This candidate set DB 52
Is updated to new contents at any time.

When a predetermined request is made from the voice search terminal 4, the server control unit 50 stores the predetermined difference information including the change to the content already transmitted to the voice search terminal 4, in the candidate set DB 52. , And transmits the difference information to the voice search terminal device 4. The communication processing unit 54 performs communication processing necessary for the server 5 to transmit and receive data to and from the voice search terminal device 4.

As described above, since the contents of the candidate set DB 26a provided in the voice search terminal device 4 can be updated based on the predetermined difference information transmitted from the server 5, the voice search terminal device 4 The updated new information can be reflected in various processes. In particular, since the information sent from the server 5 to the voice search terminal device 4 is difference information including changes to the information sent up to the previous time, the amount of data to be sent and received is reduced, and the communication cost is reduced. be able to.

The above-described server 5 has a function of storing information on search target items and search keys corresponding thereto. The voice search terminal device 4 has a function corresponding to a recognition target character string output unit, a microphone, a voice recognition processing unit, and an item extraction unit, and is required from the server 5 before performing various processes using these. Information is acquired.

FIG. 20 is a diagram showing another example of the configuration of the voice search device in the case where functions are distributed to a server and a terminal device connected via a network. The voice search device shown in FIG. 20 includes a voice search terminal device 4A and a server 5A connected via a predetermined network 6.

In the voice search device shown in FIG. 20, the candidate set DB 26 provided in the voice search device 1 of the first embodiment described above (or the candidate set DB 26a provided in the voice search device 1A of the second embodiment) ), The configuration corresponding to the function realized by each of the option setting unit 16 and the selection item determination unit 24 is arranged in the server 5A. Specifically, the server 5A includes a server control unit 50, a candidate set DB 52, a communication processing unit 54, an option setting unit 5
6, a selection item determination unit 58 is provided.

The voice search terminal device 4A is different from the above-described voice search terminal device 4 in that the option setting section 16, the selection item determination section 24, the candidate set DB 26a, and the DB update section 28 are omitted. Has been added. When outputting an operation instruction to the navigation device 2 or the like in response to the voice uttered by the user, the control unit 38 in the voice search terminal device 4A transmits the minimum data necessary for presenting an option. It is obtained from the server 5A via the communication processing unit 36. The guidance sentence generation unit 18 generates and outputs a predetermined guidance sentence according to an instruction from the control unit 38. Server 5
A obtains a voice recognition result for the user's voice from the voice search terminal device 4A, sets the next candidate set, transmits data necessary for presenting options to the voice search terminal device 4A, For example, a process of extracting one option selected in the above is performed.

The server 5A stores the information on the search key corresponding to each of the search target items, extracts the character string to be output as a sound by the recognition target character string output means, and performs the item extraction. It has a function of performing a search target item extraction process by means. Also,
The voice search terminal device 4A has functions corresponding to a recognition target character string output unit, a microphone, and a voice recognition processing unit, and stores information necessary for these processes in the server 5A described above.
From.

FIG. 21 is a diagram showing another example of the configuration of the voice search device in the case where functions are distributed to a server and a terminal device connected via a network. The voice search device shown in FIG. 21 includes a voice search terminal device 4B and a server 5B connected via a predetermined network 6.

The voice search device shown in FIG. 21 is different from the voice search device shown in FIG. 20 in that the function of the guidance sentence generation unit 18 is further arranged on the server side. Specifically, the server 5B includes a server control unit 50, a candidate set DB 52, a communication processing unit 54, an option setting unit 5
6, a selection item determination unit 58 and a guidance sentence generation unit 60 are provided. The voice search terminal device 4B is a voice search terminal device 4B.
The difference is that the guidance sentence generation unit 18 is deleted from A. In the voice search device shown in FIG. 21, since the guidance sentence is generated by the server 5B, the control unit 38 in the voice search terminal device 4B receives the guidance sentence generated by the server 5B, and Output to The other operation contents are the same as those of the voice search device shown in FIG.

FIG. 22 is a diagram showing another example of the configuration of the voice search device when the functions are distributed to the server and the terminal device connected via a network. The voice search device shown in FIG. 22 includes a voice search terminal device 4C and a server 5C connected via a predetermined network 6. The voice search device shown in FIG. 22 is different from the voice search device shown in FIG.
2 in that the function of the voice synthesis unit 20 is arranged on the server side. Specifically, the server 5C includes a server control unit 50, a candidate set DB 52, a communication processing unit 54, an option setting unit 56, a selection item determination unit 58, a guidance sentence generation unit 60, a speech recognition processing unit 62, and a speech synthesis unit 64. It has. The voice search terminal device 4C is different from the voice search terminal device 4B in that the voice recognition processing unit 12 and the voice synthesis unit 20 are deleted.

In the voice search device shown in FIG. 22, the voice of the user collected by the microphone 10 is transmitted to the control unit 38.
Is converted into digital audio data by the server 5C.
Sent to. Then, based on the transmitted voice data, a predetermined voice recognition process is performed by the voice recognition processing unit 62 in the server 5C. In addition, a predetermined voice synthesis process is performed by the voice synthesizing unit 64 in accordance with the guidance text generated by the guidance text generation unit 60, and voice data corresponding to the guidance text is generated. The generated voice data is transmitted to the voice search terminal device 4C, converted into an analog signal by the control unit 38 in the voice search terminal device 4C, and
2 is output.

In the voice search device of the modified example shown in FIGS. 20 to 22, since many functions are arranged on the server side, the processing load on the voice search terminal device side is reduced, and the configuration can be simplified. Therefore, there is an advantage that the cost of the voice search terminal device can be reduced. Further, in each of the above-described embodiments and modifications, various embodiments have been described in the case where the voice search device of the present invention is applied to an in-vehicle system.
The application range of the present invention is not limited to a vehicle-mounted system, and can be applied to various other systems.

[0131]

As described above, according to the present invention, a character string to be recognized is presented to a user in advance by reading out a plurality of character strings that can be used as search keys. Since only the character string is targeted for speech recognition, the accuracy of speech recognition can be improved.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating a configuration of a vehicle-mounted system including a voice search device according to a first embodiment;

FIG. 2 is a diagram showing a structure of data stored in a candidate set DB.

FIG. 3 is a diagram showing a correspondence relationship between an upper-layer candidate set and a lower-layer candidate set in the data structure shown in FIG. 2;

FIG. 4 is a flowchart showing an operation procedure of the voice search device of the first embodiment.

FIG. 5 is a diagram showing the contents of data used in Dialogue Example 1;

FIG. 6 is a diagram showing the contents of data used in Dialogue Example 3;

FIG. 7 is a diagram showing the contents of data used in Dialogue Example 4;

FIG. 8 is a diagram showing the contents of data used in Dialogue Example 5;

FIG. 9 is a diagram illustrating a configuration of a vehicle-mounted system including a voice search device according to a second embodiment.

FIG. 10 is a diagram illustrating a structure of data stored in a candidate set DB according to the second embodiment.

FIG. 11 is a flowchart showing a partial operation procedure of the voice search device of the second embodiment.

FIG. 12 is a flowchart showing a detailed procedure of a process shown in step 122;

FIG. 13 is a diagram showing the contents of a record extracted from a candidate set DB in Dialog Example 6;

FIG. 14 is a diagram illustrating the contents of a record extracted from a candidate set DB in Dialog Example 7;

FIG. 15 is a diagram showing the contents of a record extracted from a candidate set DB in Dialogue Example 8;

FIG. 16 is a diagram showing a configuration of a voice search device in a modification in which an option is automatically selected.

FIG. 17 is a flowchart showing a partial operation procedure of the voice search device when an option is automatically selected according to a selection frequency.

FIG. 18 is a flowchart partially showing an operation procedure of the voice search device in a case where a once selected option is canceled and a new option is selected.

FIG. 19 is a diagram illustrating a configuration example of a voice search device when functions are distributed to a server and a terminal device connected via a network.

FIG. 20 is a diagram illustrating another configuration example of the voice search device when functions are distributed to a server and a terminal device connected via a network.

FIG. 21 is a diagram illustrating another configuration example of the voice search device when functions are distributed to a server and a terminal device connected via a network.

FIG. 22 is a diagram illustrating another configuration example of the voice search device when functions are distributed to a server and a terminal device connected via a network.

[Explanation of symbols]

 1, 1A, 1B Voice search device 2 Navigation device 3 Audio device 4, 4A, 4B, 4C Voice search terminal device 5, 5A, 5B, 5C server 6 Network 10 Microphone 12, 62 Voice recognition processing unit 14 Dialogue start button 15 Re Request button 16, 56 Option setting unit 18, 60 Guidance sentence generation unit 20, 64 Voice output unit 22 Speaker 24, 58 Selection item determination unit 26, 26a, 52 Candidate set DB (database) 28 DB update unit 30 Operation instruction output unit 32 selection frequency learning unit 34 learning result storage unit

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat ゛ (Reference) G06F 17/30 G06F 17/30 310B 340 340B G10L 15/06 G10L 3/00 521W 15/18 537J 15/00 551Q 15/28 (72) Inventor Nozomu Saito 1-18 Nishi Gotanda, Shinagawa-ku, Tokyo Alpine, Inc. F-term (reference) 5B075 NK02 NK43 PP07 PP13 PQ04 PQ46 PQ75 PR03 5D015 AA04 BB01 DD02 HH00 KK01

Claims (20)

[Claims] 1. A search key is associated with each of a plurality of search target items, and by comparing the content of voice input by a user with the search key, a corresponding search target item is selected. A recognition target character string output unit that sets a maximum number of character strings that can be the search key, and outputs a voice of a plurality of the character strings in a range not exceeding the number. A microphone that collects the voice of the voice, and performs voice recognition processing on the voice collected by the microphone, and outputs a character string corresponding to this voice as a target of voice output by the recognition target character string output unit. Voice recognition processing means for selecting from a character string; and the search target corresponding to the search key specified by the character string selected by the voice recognition processing means Voice search apparatus characterized by comprising an item extracting means for extracting an eye, a. 2. The apparatus according to claim 1, further comprising: a switch operated before the user speaks, wherein when the switch is operated, a voice recognition process by the voice recognition processing unit is started. Voice search device. 3. The voice search device according to claim 1, further comprising a selected character string confirmation unit that outputs a voice of the character string selected by the voice recognition processing unit. 4. The method according to claim 3, wherein when a negative opinion by the user is given to a result of the selection of the character string by the voice recognition processing means,
A voice search apparatus further comprising reselection instructing means for instructing the recognition target character string output means to output again a voice reading of the plurality of character strings used for obtaining the selection result. . 5. The character string output unit according to claim 1, wherein the recognition target character string output unit divides the character string into a plurality of times when the total number of the character strings to be recognized exceeds the maximum number. A voice search device, which outputs voice readings of a number of character strings not exceeding a maximum number, and wherein the voice recognition processing unit performs selection determination of the character strings for each voice output. 6. The voice output instruction according to claim 5, wherein when the user instructs voice output of another selection candidate, the recognition target character string output means is instructed to output a second or subsequent voice. A voice search device further comprising means. 7. The plurality of characters that have been output immediately before to the recognition target character string output means when the user instructs to output sound again, according to any one of claims 1 to 6. A speech retrieval apparatus further comprising a re-speech output instructing unit for instructing to read a row again by speech. 8. The apparatus according to claim 1, wherein when the user instructs to perform the character string selecting operation, the result of the voice recognition processing by the voice recognition processing unit is not used. Further comprising a character string selecting means for selecting the character string, wherein when the character string is selected by the character string selecting means, the item extracting means is selected by the voice recognition processing means. A voice search device, wherein an operation of extracting the search target item is performed using the character string selected by the character string selection means instead of a character string. 9. The method according to claim 1, wherein a plurality of the search keys are associated with each of the search target items, and the item extraction means corresponds to one of the search keys. When the one search target item cannot be narrowed down, the recognition target character string output unit using the other search key, the speech recognition processing unit, and the voice recognition unit until the one search target item can be narrowed down. A voice search device characterized by repeating processing by an item extracting means. 10. The plurality of characters corresponding to the plurality of search keys, wherein each of the plurality of search keys is associated with a different priority, and each of the plurality of search target items corresponds to the plurality of search keys. Further comprising a table storage unit for storing table information associated with a column, wherein the recognition target character string output unit, based on the table information, in order from the search key with a higher priority, the corresponding character string A voice search device characterized by outputting voice reading of a voice. 11. The method according to claim 9, wherein when the character string corresponding to one of the search keys is selected, the search key to be selected next and the character string corresponding to the search key are selected. A tree structure storage unit for storing tree structure information of a plurality of hierarchies, wherein the recognition target character string output unit is configured to store a plurality of tree structures corresponding to the search keys to be output next based on the tree structure information. Wherein the character strings are extracted and the readings of these character strings are output as voice. 12. The apparatus according to claim 1, further comprising a selection history storage unit configured to store past selection history information by the voice recognition processing unit, wherein the recognition target character string output unit includes the selection history information. A voice search device that determines the character string having a high selection frequency based on the character string, and preferentially outputs voice reading of the character string. 13. The character string according to claim 1, wherein each of the plurality of character strings is composed of one of 50 Japanese sounds. A voice search device, wherein the search key that matches the one sound selected by the voice recognition processing unit is extracted. 14. The speech recognition processing device according to claim 1, wherein the speech recognition processing unit compares all characters constituting the character string with the entire speech recognition processing result, A voice search device for making a selection. 15. The character string processing device according to claim 1, wherein the voice recognition processing unit compares a character constituting a part of the character string with a whole voice recognition processing result. A voice search device characterized by making a selection. 16. The voice recognition processing unit according to claim 14, wherein the voice of the user is collected by the microphone before the voice output by the recognition target character string output unit ends. , A character string selecting operation is started. 17. The voice search device according to claim 1, wherein the maximum number is set in a range of 7 ± 2. 18. The server according to claim 1, wherein a function is distributed to a server and a terminal device connected via a network, and the server corresponds to the search target item. The terminal device has a function of storing information related to a search key, and the terminal device has a function corresponding to the recognition target character string output unit, the microphone, the voice recognition processing unit, and the item extraction unit, and uses these. A voice search device for acquiring necessary information from the server prior to various processes. 19. The voice search device according to claim 18, wherein the information sent from the server to the terminal device is difference information including a change to the information sent up to the previous time. 20. The function according to claim 1, wherein a function is distributed to a server and a terminal device connected via a network, and the server corresponds to the search target item. A function of storing information related to a search key and performing a process of extracting the character string to be output as a sound by the recognition target character string output unit and a process of extracting the search target item by the item extraction unit, The terminal device has a function corresponding to the recognition target character string output unit, the microphone, and the voice recognition processing unit, and acquires information necessary for these processes from the server. .