JP2008145697A - Voice information selection device and voice information selection method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a voice information selection device and a voice information selection method, capable of selecting a desired presentation candidate with little operation, by reflecting user's intention in a presentation order of the presentation candidate. <P>SOLUTION: When a score update section 5b receives update request of the presentation candidate before completing voice output for showing a name of the presentation candidate in a voice output section 7, the score of the presentation candidate recorded in a presentation candidate list is updated, and re-selection of the presentation candidate is instructed to a presentation candidate selection section 5a. Thereby, user's intention is reflected in the presentation order of the presentation candidate, and the desired presentation candidate is selected with little operation. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention selects a desired presentation candidate from, for example, a plurality of presentation candidates that are speech recognition results of car navigation, a plurality of presentation candidates that are search results of facility names and surrounding facilities, a plurality of presentation candidates in CTI, and the like The present invention relates to a voice information selection device and a voice information selection method.

When selecting a desired presentation candidate from a plurality of presentation candidates, the most commonly used method is to display a list of presentation candidates on the screen, and the user can use the mouse or touch panel to display the list. There is a method for selecting a desired presentation candidate from among the above.
This selection method is effective when there is a display that displays a list of presentation candidates, and the selection process of the presentation candidates can be performed while gazing at the screen of the display.
However, if you are in a situation where it is dangerous to look at the display screen or perform detailed device operations, such as when driving a car, or if you are using a device that does not have a display, such as a general phone, It is necessary to present and select candidates regardless of the display.

Conventionally, in such a situation, a voice information selection device that presents candidates by voice has been used.
In the conventional voice information selection device, when presenting candidates by voice, the plurality of presentation candidates are ranked with some priority (for example, if the plurality of presentation candidates are search results of surrounding facilities, the vehicle position is The voices representing the names of the presentation candidates are output in order from the shortest distance, in the order of the highest speech recognition likelihood (if the plurality of presentation candidates are voice recognition results).
The user confirms a plurality of presentation candidates while changing the voice output from the voice information selection device using the “next candidate” and “previous candidate” operation buttons. ”Is used to determine a desired presentation candidate (see, for example, Patent Document 1).

That is, the voice information selection device outputs voices representing the names of the presentation candidates in order from the presentation candidate with the highest ranking, but when the user listens to the name of the presentation candidate halfway, the presentation candidate is desired. Even when it can be understood that the candidate is not a presentation candidate, if all of the voice representing the name of the presentation candidate is output to the end, a lot of time is spent.
Therefore, from the viewpoint of shortening the selection time of the presentation candidate, the voice information selection device accepts the operation of “next candidate” or “previous candidate” by the user and is outputting the voice representing the name of the presentation candidate. , It is possible to switch to another presentation candidate.
However, when switching the candidate being presented to another presentation candidate, the conventional audio information selection device draws the user's intention that the candidate being presented is not the desired presentation candidate, and ranks the other presentation candidates. Since the ranking of other presentation candidates is not fixed, the number of operations for selecting the presentation candidates by the user cannot be reduced.

In addition to the voice information selection device described above, a voice information selection device has been developed that accepts user evaluation of candidates being presented and changes the ranking of other presentation candidates (see, for example, Patent Document 2).
In this audio information selection device, when the audio output related to the candidate being presented ends, the user inputs evaluations of “good”, “different”, and “very different”.
When “different” is input from the user, voice output about the next presentation candidate with a high score is started, and when “very different” is input from the user, the presentation candidates exceeding the threshold are deleted. The presentation candidate having the maximum score below the threshold is selected and the score is recalculated. And the audio | voice regarding the presentation candidate with the largest score is output.
When “good” is input from the user, the presentation candidate that is output immediately before is determined as the selection result. If there are no more selection candidates, speech recognition processing is performed again.

  As described above, in the above-described audio information selection device, the ranking of the presentation candidates is changed based on the evaluation of “good”, “different”, or “completely different”, but the distance (similarity) of the data structure of the system As a result of speech recognition, when a completely different one is presented as a candidate and the user inputs a “very different” evaluation, a data structure having a large difference is selected. Candidate changes different from the user's criteria are made, and as a result, there is a high possibility that the reduction in the number of operations will not work effectively.

For example, as a voice recognition result when the user utters “Yokohama Asahi Ward Office” (Yokohama Sahiku Yakusho), “Asahi Ward Yahagicho” (Yokohama Sahiku Yasashicho) ”(Yokohama Midokuryakusho) shall be obtained.
“Yokohama City Midori Ward Office” has the same genre as “Yokohama City Asahi Ward Office”, and the content of the utterance is only a small part, so the user is likely to evaluate as “different”.
On the other hand, “Yazawa-machi, Asahi-ku, Yokohama-shi” is geographically close and has many similar sounds, but the genre is different from “Asahi-ku government office in Yokohama-shi”.
For this reason, there is a high possibility that a user will be divided into an evaluation of “different” and an evaluation of “very different”.
As described above, since the determination criterion is completely different between the distance as the voice recognition result and the distance according to the category, there is a high possibility that the change in the presentation order cannot reflect the intention of the user.

Japanese Patent Laid-Open No. 2001-13978 (pages 1 to 9, FIG. 1) Japanese Patent Laid-Open No. 2000-305590 (pages 1 to 7, FIG. 1)

  Since the conventional audio information selection apparatus is configured as described above, the user can switch to another presentation candidate even while outputting the voice representing the name of the presentation candidate. However, when switching the presenting candidate to another presenting candidate, the user's intention that the presenting candidate is not the desired presenting candidate is taken into consideration, and the ranking of the other presenting candidates is not changed. Since the ranking of the presentation candidates is fixed, there is a problem that the number of operations for selecting the presentation candidates by the user cannot be reduced.

  The present invention has been made in order to solve the above-described problems, and an audio information selection device capable of selecting a desired presentation candidate with few operations by reflecting the user's intention in the presentation candidate presentation order, and An object is to obtain a voice information selection method.

  The audio information selection device according to the present invention includes a presentation candidate selection unit that selects a presentation candidate with the highest score from the presentation candidate list held in the presentation candidate list holding unit, and a presentation selected by the presentation candidate selection unit A voice output unit that generates a voice representing the name of the candidate and outputs a voice representing the name of the presentation candidate, and before the score update unit completes outputting the voice representing the name of the presentation candidate in the voice output unit When the request for changing the presentation candidate is received, the score of the presentation candidate recorded in the presentation candidate list is updated, and the re-selection of the presentation candidate is instructed to the presentation candidate selection means.

  According to this invention, the presentation candidate selection means for selecting the presentation candidate with the highest score from the presentation candidate list held in the presentation candidate list holding means, and the name of the presentation candidate selected by the presentation candidate selection means Voice output means for generating a voice representing the name of the presentation candidate and outputting a voice representing the name of the presentation candidate, and the score updating means before the completion of the output of the voice representing the name of the presentation candidate in the voice output means When the change request is received, the presentation candidate score recorded in the presentation candidate list is updated, and the presentation candidate selection unit is instructed to reselect the presentation candidate. There is an effect that a desired presentation candidate can be selected with a small number of operations, reflecting the ranking.

Embodiment 1 FIG.
FIG. 1 is a block diagram showing a speech information selection apparatus according to Embodiment 1 of the present invention. In the figure, a presentation candidate list input terminal 1 has a plurality of presentation candidate scores, facility names (texts mixed with kana and kanji), phonemes. This is a terminal for inputting a presentation candidate list in which reading information (hereinafter referred to as “intermediate language”) including symbols and prosodic symbols and geographical position information are recorded.
In the first embodiment of the present invention, it is assumed that the presentation candidates recorded in the presentation candidate list input from the presentation candidate list input terminal 1 are search results searched from the database under a predetermined search condition.
The presentation candidate list holding unit 2 is a memory that holds the presentation candidate list input from the presentation candidate list input terminal 1. The presentation candidate list holding unit 2 constitutes a presentation candidate list holding unit.

The user request input terminal 3 is, for example, a user request signal corresponding to a “selection” function button or a touch panel in addition to a user request signal corresponding to a “next candidate” or “previous candidate” function button or a touch panel (presentation candidate change request). This is a terminal for inputting a signal (request for selection of a presentation candidate).
The user request signal input unit 4 performs a process of notifying the presentation candidate control unit 5 of the user request signal input from the user request input terminal 3.

The presentation candidate control unit 5 includes a presentation candidate selection unit 5a and a score update unit 5b. The presentation candidate selection unit 5a presents the presentation with the highest score from the presentation candidate lists held in the presentation candidate list holding unit 2. A process for selecting candidates is performed. In addition, the presentation candidate selection part 5a comprises the presentation candidate selection means.
When the score update unit 5b of the presentation candidate control unit 5 receives a user request signal indicating a request for changing a presentation candidate from the user request signal input unit 4 before the output of the voice in the voice output unit 7 to be described later is completed, The process updates the presentation candidate score recorded in the candidate list and instructs the presentation candidate selection unit 5a to reselect the presentation candidate. The score update unit 5b constitutes score update means.

The speech synthesizer 6 generates a synthesized sound representing the facility name of the presentation candidate selected by the presentation candidate selection unit 5a of the presentation candidate control unit 5, and the voice output start time of each accent phrase in the synthesized sound, the synthesized sound To generate time information indicating the voice output end time and the input time of the change request for the presentation candidate.
The voice output unit 7 outputs the synthesized sound generated by the voice synthesizing unit 6 to the speaker 8 and, depending on the progress of the output of the synthesized sound, the voice output start time of the accent phrase and the voice output end time of the synthesized sound The process which notifies the time information which shows this to the presentation candidate control part 5 is implemented. The voice synthesizing unit 6, the voice output unit 7, and the speaker 8 constitute voice output means.

FIG. 2 is a flowchart showing the audio information selection method according to Embodiment 1 of the present invention, and FIG. 3 is a flowchart showing details of step ST13 in FIG.
FIG. 4 shows the relationship between the synthesized speech generated by the speech synthesizer 6 and the time information indicating the speech output start time of each accent phrase, the speech output end time of the synthesized speech, and the input time of the presentation candidate change request. It is explanatory drawing which is.
In the figure, 21 represents a synthesized sound, and the synthesized sound 21 shows an example composed of three accent phrases “XX donut”, “Higashikanagawa”, and “Station front shop”.
Reference numeral 22 denotes a score reduction range when the first accent phrase “XX donut” is a score reduction target.
23 is the voice output start time of the accent phrase “XX donut”, 24 is the voice output start time of the accent phrase “Higashikanagawa”, 25 is the voice output start time of the accent phrase “Ekimae Shop”, and 26 is the voice of the synthesized sound 21 The output end time, 27 indicates the input time of the presentation candidate change request.
Note that the input time 27 indicates that a user request signal (request for changing a presentation candidate) corresponding to “next candidate” is input when the syllable “o” of “donato” is being output by voice.

FIG. 5 is an explanatory diagram showing a presentation candidate list in which a plurality of presentation candidates are recorded as a search result when a peripheral facility is searched by the car navigation system.
In the figure, the presentation candidate list 31 is a list in which the presentation candidate list input terminal 1 is input from the car navigation system, and the presentation candidate list 41 is a score of the presentation candidate list 31 updated by the score update unit 5b of the presentation candidate control unit 5. It is a list.
The presentation candidate list 51 is a list in which the score of the presentation candidate list 41 is updated by the score update unit 5 b of the presentation candidate control unit 5, and the presentation candidate list 61 is presented by the score update unit 5 b of the presentation candidate control unit 5. This is a list with updated scores.
The presentation candidates 32 to 34 are recorded in the presentation candidate list 31, the presentation candidates 42 to 48 are recorded in the presentation candidate list 41, and the presentation candidates 52 to 58 are recorded in the presentation candidate list 51. The presentation candidates 62 and 63 are the presentation candidates recorded in the presentation candidate list 61.

In the reading information of the presentation candidates, the underlined portion indicates a portion including a word representing a place name. The determination of whether or not the name is a place name can be made by, for example, the following method.
Because the location is known from the location information of the facility and the address of the facility can be taken out, the word included in the address, the word included in the adjacent address, or the word including the name of the nearest station that can be calculated from the location is the place name Judge.
FIG. 6 is an explanatory diagram for explaining a score reduction calculation method when the score reduction word is a place name. In the figure, 71 indicates the current vehicle position, and 72-74 indicate the positions of the presentation candidates 52-54. ing.

Next, the operation will be described.
In the first embodiment, it is assumed that the car navigation system is used, and for example, the car navigation system searches for surrounding facilities at the position of the vehicle.
For example, when the user utters “near fast food” toward the voice input unit of a car navigation system (not shown), the voice recognition unit of the car navigation system (not shown) correctly recognizes “near fast food”. Assume that a data search unit of a car navigation system (not shown) searches a facility of the corresponding genre “fast food” from the facility name database.
At this time, the data search unit of the car navigation system generates the presentation candidate list 31 by sorting the presentation candidates including the facility name, the reading information, and the position information in the order of score as the retrieval result (presentation candidate). Is input from the presentation candidate list input terminal 1.

When the presentation candidate list 31 is input from the presentation candidate list input terminal 1, the presentation candidate selection unit 5a of the presentation candidate control unit 5 stores the presentation candidate list 31 in the presentation candidate list holding unit 2 (step ST1).
Further, the presentation candidate selection unit 5a of the presentation candidate control unit 5 initially sets “1” to the information presentation position variable I indicating the information presentation position (step ST2).

The presentation candidate selection unit 5 a of the presentation candidate control unit 5 selects the I-th presentation candidate from the presentation candidates recorded in the presentation candidate list 31.
At this time, since I = 1, the first candidate for presentation, the facility name “XX Donut Higashi-Kanagawa Ekimae Shop” is selected, and the reading information of the candidate 32 is read out by the speech synthesizer 6. Output to.

When the speech synthesis unit 6 receives the reading information of the presentation candidate 32 from the presentation candidate selection unit 5a of the presentation candidate control unit 5, the speech synthesis unit 6 generates a synthesized sound 21 representing the facility name of the presentation candidate 32 from the intermediate language that is the reading information. Then, the synthesized sound 21 is output to the sound output unit 7 (step ST3).
Here, although the method of converting the intermediate language which is the reading information of the presentation candidate 32 into the synthesized sound 21 is assumed, the synthesized sound 21 may be generated by analyzing text mixed with kana and kanji.
When the speech synthesis unit 6 generates the synthesized sound 21 representing the facility name of the presentation candidate 32, time information indicating the speech output start times 23, 24, and 25 of each accent phrase in the synthesized sound 21, and the synthesized sound 21 is generated, and the time information is output to the sound output unit 7.

Upon receiving the synthesized sound 21 and time information from the speech synthesizing unit 6, the audio output unit 7 outputs the synthesized sound 21 to the speaker 8 in order from the head of the synthesized sound 21, thereby starting reproduction of the synthesized sound 21 (step ST4). ).
In addition, when the sound output unit 7 starts to reproduce the synthesized sound 21, time information indicating the sound output start time of the accent phrase and the sound output end time of the synthesized sound is displayed according to the progress of the reproduction of the synthesized sound 21. The presentation candidate control unit 5 is notified.
Specifically, time information indicating the audio output start time 23 is notified to the presentation candidate control unit 5 at the timing of starting playback of the accent phrase “XX donut”, and playback of the accent phrase “Higashi Kanagawa” is started. The time information indicating the audio output start time 24 is notified to the presentation candidate control unit 5 at the timing, and the time information indicating the audio output start time 25 is displayed at the timing when the accent phrase “shop in front of the station” is started to be reproduced. 5 is notified.
In addition, time information indicating the audio output end time 26 of the synthesized sound 21 is notified to the presentation candidate control unit 5 at the timing when the reproduction of the accent phrase “shop in front of the station” is completed.

The score update unit 5b of the presentation candidate control unit 5 waits for notifications from the audio output unit 7 (time information indicating the audio output start times 23 to 25 and time information indicating the audio output end time 26), and inputs a user request signal. Processing for waiting for notification from the unit 4 (user request signal indicating a request for changing a presentation candidate, user request signal indicating a request for selecting a presentation candidate) is performed (step ST5).
When receiving the notification from the voice output unit 7 or the user request signal input unit 4, the score update unit 5 b of the presentation candidate control unit 5 determines whether the notification has been received from either the voice output unit 7 or the user request signal input unit 4. (Step ST6).
In the example of FIG. 4, first, the time information indicating the sound output start time 23 is notified from the sound output unit 7 at the timing of starting the reproduction of the accent phrase “XX donut”, and then the syllable “ Since the user request signal corresponding to the “next candidate” is sent from the user request signal input unit 4 at the timing when “O” is output as a voice, at this point, the voice output unit 7 Time information indicating the audio output start time 23 is notified.

When the score update unit 5b of the presentation candidate control unit 5 is notified of the time information indicating the audio output start time 23 at the timing of starting the reproduction of the accent phrase “XX donut” from the audio output unit 7 (step ST5, ST6) The reproduction position information indicating the reproduction position of the synthesized sound 21 is updated so that it can be understood that it is “currently in the middle of reproducing the accent phrase“ OO donut ”” (step ST7).
When the score update unit 5b of the presentation candidate control unit 5 updates the reproduction position information indicating the reproduction position of the synthetic sound 21, it determines whether or not the reproduction of the synthetic sound 21 is completed (step ST8).
At this time, since the notification of the time information indicating the audio output end time 26 has not been received from the audio output unit 7, it is determined that the reproduction of the synthesized sound 21 has not been completed, and the process returns to step ST5.

Next, the score update unit 5b of the presentation candidate control unit 5 outputs the user request signal (“candidate”) corresponding to the “next candidate” from the user request signal input unit 4 at the timing when the syllable “o” of “Donut” is output by voice. When a notification of a presentation candidate change request is received (steps ST5 and ST6), it is determined whether or not the user request signal is a selection request for the presentation candidate 32 (step ST10).
If the user request signal is a request for selecting the presentation candidate 32 (a request to be determined as the presentation candidate 32), the series of processing ends. At this point, the user request signal notified from the user request signal input unit 4 is Since it is a user request signal (request for changing a presentation candidate) corresponding to “next candidate”, the process proceeds to step ST11.

In step ST11, the score update unit 5b of the presentation candidate control unit 5 has the user request signal notified from the user request signal input unit 4 as a user request signal (presentation candidate change request) corresponding to “next candidate”. It is determined whether or not.
At this time, as described above, the user request signal notified from the user request signal input unit 4 is the user request signal corresponding to the “next candidate” (request for changing the presentation candidate), and therefore the I + 1th and subsequent presentations The score of the candidate (including the I + 1th presentation candidate) is re-evaluated (step ST13).
At this time, since I = 1, the scores of the second to eleventh presentation candidates are reevaluated.

That is, the score update unit 5b of the presentation candidate control unit 5 outputs the user request signal (presentation) corresponding to the “next candidate” from the user request signal input unit 4 at the timing when the syllable “o” of “Donut” is output by voice. Since the notification of “candidate change request” has been received, the accent phrase “XX donut” is determined as the score-lowering word (step ST21 in FIG. 3).
After receiving notification of time information indicating the audio output end time 26 from the audio output unit 7, the user request signal input unit 4 notifies the user request signal (presentation candidate change request) corresponding to “next candidate”. In such a case, it is not possible to determine a score-decreasing word.

When the score update unit 5b of the presentation candidate control unit 5 determines a score-lowering word (step ST22), it determines whether the score-lowering word is a place name (step ST23).
At this point in time, the accent phrase “XX donut” has been determined as the score-lowering word, so it is determined that the score-lowering word is not a place name.
When the score update unit 5b of the presentation candidate control unit 5 determines that the score-lowering word is not a place name, among the second to eleventh presentation candidates recorded in the presentation candidate list 31, the score-lowering word “XX” Recalculation is performed to lower the scores of the presentation candidates 32, 33, and 34 including “donut” (step ST25).

Here, as the recalculation of the score, for example, a method is adopted in which 1/5 of the original score is used as the recalculated score.
As a result, the score of the presentation candidate 32 is 0.18, the score of the presentation candidate 33 is 0.144, and the score of the presentation candidate 34 is 0.076.
Although the presentation order of the already presented candidate data 32 does not change, the presentation candidate 33 is changed to the presentation candidate 47 and the presentation candidate 34 is changed to the presentation order of the presentation candidate 48. As for the other presentation candidates, the presentation ranks up by the amount of the presentation candidates 33 and 34 missing.
When the score update unit 5b of the presentation candidate control unit 5 changes the presentation rank of the presentation candidates recorded in the presentation candidate list 31 as described above, the presentation candidate list 41 reflecting the changed presentation rank. Is reset in the presentation candidate list holding unit 2 (step ST26).

The presentation candidate selection unit 5a of the presentation candidate control unit 5 resets the presentation candidate list 41 reflecting the changed presentation rank in the presentation candidate list holding unit 2 when the score update unit 5b resets the presentation candidate list to be presented next. Is selected, the current information presentation position variable I is incremented (step ST14).
At this time, since I = 1, I = 2 by I ← I + 1.

The presentation candidate selection unit 5 a of the presentation candidate control unit 5 selects the I-th presentation candidate from the presentation candidates recorded in the presentation candidate list 41.
At this time, since I = 2, the presentation candidate 42 whose facility name as the second presentation candidate is “◯ ΔNardo Higashi Kanagawa Ekimae” is selected, and the reading information of the presentation candidate 42 is converted into the speech synthesis unit 6. Output to.

When the speech synthesis unit 6 receives the reading information of the presentation candidate 42 from the presentation candidate selection unit 5 a of the presentation candidate control unit 5, the speech synthesis unit 6 presents it from the intermediate language that is the reading information of the presentation candidate 42 as in the case of the presentation candidate 32. A synthesized sound representing the facility name of the candidate 42 is generated, and the synthesized sound is output to the voice output unit 7 (step ST3).
Moreover, when the speech synthesis unit 6 generates a synthesized sound representing the facility name of the presentation candidate 42, as in the case of the presentation candidate 32, the time information indicating the voice output start time of each accent phrase in the synthesized sound, Time information indicating the voice output end time of the synthesized sound is generated, and the time information is output to the voice output unit 7.

When the voice output unit 7 receives the synthesized sound and time information from the voice synthesis unit 6, as in the case of the presentation candidate 32, the voice output unit 7 outputs the synthesized sound in order from the head of the synthesized sound, thereby reproducing the synthesized sound. Is started (step ST4).
In addition, when the sound output unit 7 starts the reproduction of the synthesized sound, the time information indicating the sound output start time of the accent phrase and the sound output end time of the synthesized sound is displayed according to the progress of the reproduction of the synthesized sound. Notify the control unit 5.
In this example, for convenience of explanation, the user request signal corresponding to the “next candidate” from the user request signal input unit 4 at the timing when the syllable “sa” of “Marsan Cañardo” is output as a voice ) Is notified to the presentation candidate control unit 5.

As in the case of the presentation candidate 32, the score update unit 5b of the presentation candidate control unit 5 waits for a notification (time information indicating the voice output start time, time information indicating the voice output end time) from the voice output unit 7. Then, a process of waiting for a notification from the user request signal input unit 4 (a user request signal indicating a request for changing a presentation candidate, a user request signal indicating a request for selecting a presentation candidate) is performed (step ST5).
When receiving the notification from the voice output unit 7 or the user request signal input unit 4, the score update unit 5 b of the presentation candidate control unit 5 determines whether the notification has been received from either the voice output unit 7 or the user request signal input unit 4. (Step ST6).
In this example, first, the time information indicating the sound output start time is notified from the sound output unit 7 at the timing of starting the reproduction of the accent phrase “○ Δ Nardo”, and then the syllable “sal“ ”Is output from the user request signal input unit 4 at a timing at which the user request signal (request for changing the presentation candidate) is notified from the user request signal input unit 4. Time information indicating the output start time is notified.

When the score update unit 5b of the presentation candidate control unit 5 is notified of time information indicating the audio output start time at the timing of starting the reproduction of the accent phrase “◯ ΔNardo” from the audio output unit 7 (steps ST5 and ST6). ), The playback position information indicating the playback position of the synthesized sound is updated so that it is understood that it is “currently in the middle of playing back the accent phrase“ ◯ Δ Nardo ”” (step ST7).
When the score update unit 5b of the presentation candidate control unit 5 updates the reproduction position information indicating the reproduction position of the synthesized sound, the score update unit 5b determines whether or not the reproduction of the synthesized sound is completed (step ST8).
At this time, since the notification of the time information indicating the sound output end time has not been received from the sound output unit 7, it is determined that the reproduction of the synthesized sound is not completed, and the process returns to step ST5.

Next, the score update unit 5b of the presentation candidate control unit 5 outputs a user request corresponding to the “next candidate” from the user request signal input unit 4 at the timing when the syllable “sa” of “Marsan Cakunald” is output as voice. Upon receiving the notification of the signal (presentation candidate change request) (steps ST5 and ST6), it is determined whether or not the user request signal is a request for selecting the presentation candidate 42 (step ST10).
If the user request signal is a request for selecting the presentation candidate 42 (a request to be determined as the presentation candidate 42), the series of processing ends. At this point, the user request signal notified from the user request signal input unit 4 is Since it is a user request signal (request for changing a presentation candidate) corresponding to “next candidate”, the process proceeds to step ST11.

In step ST11, the score update unit 5b of the presentation candidate control unit 5 has the user request signal notified from the user request signal input unit 4 as a user request signal (presentation candidate change request) corresponding to “next candidate”. It is determined whether or not.
At this time, as described above, the user request signal notified from the user request signal input unit 4 is the user request signal corresponding to the “next candidate” (request for changing the presentation candidate), and therefore the I + 1th and subsequent presentations The score of the candidate (including the I + 1th presentation candidate) is re-evaluated (step ST13).
At this time, since I = 2, the scores of the third to eleventh presentation candidates are reevaluated.

That is, the score update unit 5b of the presentation candidate control unit 5 outputs the user request signal corresponding to the “next candidate” from the user request signal input unit 4 at the timing of outputting the syllable “sa” of “Marsan Cakunald” by voice. Since the notification of (change request for presentation candidate) has been received, the accent phrase “◯ ΔNardo” is determined as the score-lowering word (step ST21 in FIG. 3).
When the score update unit 5b of the presentation candidate control unit 5 determines a score-lowering word (step ST22), it determines whether the score-lowering word is a place name (step ST23).
At this time, since the accent phrase “◯ ΔNardo” is determined to be a score-reduced word, it is determined that the score-reduced word is not a place name.
When the score update unit 5b of the presentation candidate control unit 5 determines that the score-lowering word is not a place name, among the third to eleventh presentation candidates recorded in the presentation candidate list 31, the score-lowering word “◯ △ Recalculation is performed to lower the scores of the presentation candidates 43 to 46 including “Nardo” (step ST25). The score recalculation method adopts the same method as described above.

As a result, the presentation order of the presented presentation candidates 42 is not changed, but the presentation candidates 43 to 46 are changed to the presentation orders of the presentation candidates 55 to 58. As for the other presentation candidates, the presentation ranks up because the presentation candidates 43 to 46 are missing.
When the score updating unit 5b of the presentation candidate control unit 5 changes the presentation rank of the presentation candidates recorded in the presentation candidate list 41 as described above, the presentation candidate list 51 reflecting the changed presentation rank. Is reset in the presentation candidate list holding unit 2 (step ST26).

The presentation candidate selection unit 5a of the presentation candidate control unit 5 resets the presentation candidate list 51 reflecting the changed presentation rank in the presentation candidate list holding unit 2 when the score update unit 5b resets the presentation candidate list to be presented next. Is selected, the current information presentation position variable I is incremented (step ST14).
At this time, since I = 2, I = 3 by I ← I + 1.

The presentation candidate selection unit 5 a of the presentation candidate control unit 5 selects the I-th presentation candidate from the presentation candidates recorded in the presentation candidate list 51.
At this time point, since I = 3, the presentation candidate 52 whose facility name is “Δ Burger Yokohama Santimachi” as the third presentation candidate is selected, and the reading information of the presentation candidate 52 is output to the speech synthesizer 6. To do.
In this example, for convenience of explanation, a user request signal (request for changing a presentation candidate) corresponding to the “next candidate” is output from the user request signal input unit 4 at the timing when the syllable “ma” of “tanmachichiten” is output by voice. It is assumed that the presentation candidate control unit 5 is notified.

In the speech synthesis unit 6, the speech output unit 7, and the score update unit 5 b of the presentation candidate control unit 5, the same processing as that of the presentation candidates 32 and 42 is performed for the presentation candidate 52. In this case, at the timing when the voice output unit 7 outputs the syllable “ma” of “tanmachiten” as a voice, the score update unit 5 b of the presentation candidate control unit 5 receives a user corresponding to the “next candidate” from the user request signal input unit 4. Since the notification of the request signal (request for changing the presentation candidate) is received, the accent phrase “Sanmachi store” is determined as the score-lowering word in step ST21 of FIG.
When the score update unit 5b of the presentation candidate control unit 5 determines the accent phrase “Sanmachi store” as a score-lowering word, the score-lowering word includes the place name “Sanmachi” (step ST23). Considering the position information of the presentation candidates recorded in the list 51, the scores of the fourth to eleventh presentation candidates are recalculated (step ST24).

Hereinafter, the recalculation of the score of the presentation candidate in the score update unit 5b of the presentation candidate control unit 5 will be specifically described.
The vehicle position of the vehicle on which the navigation system is currently mounted is the vehicle position 71 in FIG. 6, and the position of the presentation candidate 52 currently being presented is the position 72 in FIG. 6.
The score update unit 5b of the presentation candidate control unit 5 uses the own vehicle position 71 and the position 72 of the currently presented presentation candidate 52 as the reference position, and the fourth to eleventh presentation candidates are close to the own vehicle position 71. It is determined whether there is a presentation candidate 52 currently being presented.

The score update unit 5b of the presentation candidate control unit 5 determines the scores of the fourth to eleventh presentation candidates according to the determination result.
For example, since the position 73 of the fourth presentation candidate 53 is closer to the vehicle position 71 than the present presentation candidate 52, the score of the fourth presentation candidate 53 is maintained.
On the other hand, since the position 74 of the fifth presentation candidate 54 is closer to the present presentation candidate 52 than the own vehicle position 71, the score of the fifth presentation candidate 54 is lowered.
Here, for convenience of explanation, since the positions of the sixth to eleventh presentation candidates are closer to the vehicle position 71 than the currently presented presentation candidate 52, the scores of the sixth to eleventh presentation candidates are maintained. To do.

As a result, the score of the fifth presentation candidate 54 is lowered, and the ranking of the presentation candidate 54 is lowered to the position of the candidate data 63 in the presentation candidate list 61.
When the score updating unit 5b of the presentation candidate control unit 5 changes the presentation rank of the presentation candidates recorded in the presentation candidate list 51 as described above, the presentation candidate list 61 reflecting the changed presentation rank. Is reset in the presentation candidate list holding unit 2 (step ST26).

The presentation candidate selection unit 5a of the presentation candidate control unit 5 sets the presentation candidate list 61 reflecting the changed presentation order in the presentation candidate list holding unit 2 when the score update unit 5b resets the presentation candidate list holding unit 2, and then presents the presentation candidate to be presented next. Is selected, the current information presentation position variable I is incremented (step ST14).
At this time, since I = 3, I = 4 by I ← I + 1.

The presentation candidate selection unit 5 a of the presentation candidate control unit 5 selects the I-th presentation candidate from the presentation candidates recorded in the presentation candidate list 61.
At this time, since I = 4, the presentation candidate 62 whose facility name is “◯ □ Fried Chicken Rokkakubashi” as the fourth presentation candidate is selected, and the speech synthesis unit 6 converts the reading information of the presentation candidate 62 into the speech synthesis unit 6. Output to.
In this example, the accent phrase “Marshikaku”, the accent phrase “Fried Chicken”, and the accent phrase “Rocka Bashten” are played in order, but before the playback of the accent phrase “Rocka Bashten” is completed, After the user request signal corresponding to the “next candidate” is not input from the request signal input unit 4 and the reproduction of the accent phrase “Rocka Bashten” is completed, the user request signal input unit 4 outputs “ A user request signal (presentation candidate selection request) corresponding to the “select” function button is notified to the presentation candidate control unit 5.

  In the speech synthesis unit 6, the voice output unit 7, and the score update unit 5 b of the presentation candidate control unit 5, the same processing as that of the presentation candidates 32, 42, and 52 is performed for the presentation candidate 62. In the case of 62, the score update unit 5b of the presentation candidate control unit 5 receives the notification of the user request signal (request for changing the presentation candidate) corresponding to the “next candidate” from the user request signal input unit 4, and the voice output unit 7 receives the notification of the time information indicating the voice output end time, so in step ST8, it is determined that the reproduction of the synthesized sound representing the facility name of the presentation candidate 62 has been completed.

The score update unit 5b of the presentation candidate control unit 5 determines that the reproduction of the synthesized sound representing the facility name of the presentation candidate 62 has been completed, and then notifies the user request signal from the user request signal input unit 4 within a predetermined time t2. It is determined whether or not it has been done (step ST9).
When the user request signal is notified from the user request signal input unit 4 within the predetermined time t2, the score update unit 5b of the presentation candidate control unit 5 receives the user request corresponding to the “select” function button. It is determined whether or not it is a signal (presentation candidate selection request) (step ST10), and if the user request signal is a user request signal (presentation candidate selection request) corresponding to the “select” function button, presentation The candidate 62 is recognized as the finally selected presentation candidate, and the series of processing ends.
In the case of the first embodiment, the “□□ Fried Chicken Rokkakubashi store”, which is the sixth presentation candidate in the presentation candidate list 31, is changed to the fourth presentation candidate in the presentation candidate list 61, and is desired with fewer operations. Presentation candidates 62 can be selected.

If the user request signal is not notified from the user request signal input unit 4 within the predetermined time t2 (step ST9), the score update unit 5b of the presentation candidate control unit 5 presents next without changing the score of the presentation candidate. In order to select a presentation candidate, the current information presentation position variable I is incremented (step ST14).
At this time, since I = 4, I = 5 by I ← I + 1.
Here, the case where the user request signal is not notified from the user request signal input unit 4 within the predetermined time t2 (step ST9) has been described as shifting to the process of step ST14, but the process is suspended until the user performs an operation. You may do it.

  As apparent from the above, according to the first embodiment, the presentation candidate selection unit 5a that selects the presentation candidate with the highest score from the presentation candidate list held in the presentation candidate list holding unit 2, and the presentation A speech synthesizing unit 6 that generates speech representing the name of the presentation candidate selected by the candidate selection unit 5a and a speech output unit 7 that outputs speech representing the name of the presentation candidate are provided, and the score update unit 5b outputs the speech. When the request for changing the presentation candidate is received before the output of the voice representing the name of the presentation candidate in the unit 7 is completed, the score of the presentation candidate recorded in the presentation candidate list is updated and the reselection of the presentation candidate is presented. Since the configuration is such that the candidate selection unit 5a is instructed, the user's intention is reflected in the presentation candidate presentation order, and the desired presentation candidate can be selected with few operations.

In the first embodiment, the voice output start notification is performed in units of accent phrases. However, the present invention is not limited to this. For example, the voice output start notification is performed in syllable units, morpheme units, or phrase units. May be performed.
In the first embodiment, the user request signal corresponding to the three function buttons “next candidate”, “previous candidate”, and “select” is input, but the function button “correct” is supported. When a user request signal corresponding to the “correction” function button is input, the presentation candidate control unit 5 inputs the user's voice and performs voice recognition processing on the voice. Then, the presentation order of candidates including the word corresponding to the voice recognition result may be increased.
In addition, the presentation candidate control unit 5 may recalculate the presentation order by lowering the score of a presentation candidate that does not include a word that is acoustically close to the speech recognition result.

  In the first embodiment, the user request signal is input by the user operating the function button. However, the presentation candidate control unit 5 recognizes the user's utterance and the result of the voice recognition is displayed. It may be interpreted as an operation instruction.

  Furthermore, in the first embodiment, as a method for creating a synthesized sound, a method for creating a synthesized sound based on reading information (intermediate language) stored in a database in advance in the system has been described. It is also possible to create a synthesized speech when the presentation candidate list is created, and it is also possible to generate morpheme information or accent phrase information together with syllable measurement time information and use it for control.

  In the first embodiment, when recalculating the score of the presentation candidate list, the position information of the own vehicle, the position information of the current presentation candidate whose score reduction target is the place name, and the position information of other presentation candidates are used. In addition to that, the route to the destination that has already been set, the positional relationship of the candidate facilities, information on the presence or absence of parking lots, detailed genre information given to the facilities, etc. It may be recalculated.

Embodiment 2. FIG.
In the first embodiment, the case where the presentation candidate recorded in the presentation candidate list is a search result retrieved from the database under a predetermined search condition has been described. However, the presentation candidate recorded in the presentation candidate list is the user. It may be a voice recognition result for the voice.
In this case, the score update algorithm in the case where the presentation candidate recorded in the presentation candidate list is a search result searched from the database under a predetermined search condition, and the presentation candidate recorded in the presentation candidate list is the user's voice. It is assumed that the score update algorithm in the case of the speech recognition result for is different from

The configuration diagram of FIG. 1 and the flowchart of FIG. 2 are also used in the second embodiment.
FIG. 7 is a flowchart showing details of step ST13 in FIG. 2, and FIG. 8 is an explanatory diagram showing an example of a score correction calculation formula.
Score _New indicates the score after recalculation, and Score _Org indicates the speech recognition result score.
n represents the number of accent phrases in reading, and i represents the position of the accent phrase from the beginning. Further, a and b are appropriate coefficients from 0 to 1, α _i is an evaluation value when the i-th accent phrase is positively evaluated from the beginning of the facility name, and β _i is a minus value of the i-th accent phrase The evaluation value when the evaluation is received is shown. That is, as α _i increases, the numerator of the calculation formula increases, and as β _i increases, the numerator of the calculation formula decreases.

FIG. 9 is an explanatory diagram showing a presentation candidate list in which a plurality of presentation candidates are recorded as a voice recognition result of the car navigation system.
In the figure, a presentation candidate list 81 is a list in which the presentation candidate list input terminal 1 is inputted from the car navigation system, and a presentation candidate list 91 is a score of the presentation candidate list 81 updated by the score update unit 5b of the presentation candidate control unit 5. It is a list.
The presentation candidate list 101 is a list in which the score of the presentation candidate list 91 is updated by the score update unit 5 b of the presentation candidate control unit 5.
The presentation candidates 82 to 88 are recorded in the presentation candidate list 81, the presentation candidates 92 to 95 are presented in the presentation candidate list 91, and the presentation candidates 102 to 104 are recorded in the presentation candidate list 101. Is a presentation candidate.
FIG. 10 is an explanatory diagram showing the presentation candidate list 111 in which the presentation candidate list 101 in FIG. 9 is updated, and the presentation candidate 112 is a presentation candidate recorded in the presentation candidate list 111.

Next, the operation will be described.
In the second embodiment, it is assumed that a destination is set by voice recognition of a facility name in a car navigation system, or a facility search using voice recognition of a facility name in a telephone number guidance service.
Specifically, assuming an application in which an address or facility name is input by voice and the recognition result of the voice is selected, the user utters “Aikawacho Tanagawa, Kanagawa” (Kanagawa Ken Aikawa Chotanazawa) Will be described.

When the presentation candidate list 81 is input from the presentation candidate list input terminal 1, the presentation candidate selection unit 5 a of the presentation candidate control unit 5 displays the presentation candidate list 81 as a presentation candidate list holding unit as in the first embodiment. 2 (step ST1).
Further, the presentation candidate selection unit 5a of the presentation candidate control unit 5 initially sets “1” to the information presentation position variable I indicating the information presentation position (step ST2).

The presentation candidate selection unit 5 a of the presentation candidate control unit 5 selects the I-th presentation candidate from the presentation candidates recorded in the presentation candidate list 81.
At this point in time, since I = 1, the presentation candidate 82 whose facility name is “Tadotsucho Takami, Kagawa Prefecture” as the first presentation candidate is selected, and the reading information of the presentation candidate 82 is output to the speech synthesizer 6. To do.

When the speech synthesis unit 6 receives the reading information of the presentation candidate 82 from the presentation candidate selection unit 5a of the presentation candidate control unit 5, as in the first embodiment, the facility of the presentation candidate 82 is read from the intermediate language as the reading information. A synthesized sound representing the name is generated, and the synthesized sound is output to the voice output unit 7 (step ST3).
Also, when the speech synthesis unit 6 generates a synthesized sound representing the facility name of the presentation candidate 82, a time indicating the voice output start time of each accent phrase “Kagawa Prefecture”, “Tadotsucho”, and “Takami” in the synthesized sound Information and time information indicating the sound output end time of the synthesized sound are generated, and the time information is output to the sound output unit 7.

When the voice output unit 7 receives the synthesized sound and the time information from the voice synthesizing unit 6, as in the first embodiment, the voice output unit 7 outputs the synthesized sound in order from the head of the synthesized sound to reproduce the synthesized sound. Is started (step ST4).
Further, when the reproduction of the synthesized sound is started, the sound output unit 7 starts the sound output start time of the accent phrase and the sound output of the synthesized sound according to the progress of the reproduction of the synthesized sound, as in the first embodiment. Time information indicating the end time is notified to the presentation candidate control unit 5.

The score update unit 5b of the presentation candidate control unit 5 waits for notifications from the audio output unit 7 (time information indicating the audio output start time, time information indicating the audio output end time) and from the user request signal input unit 4 Processing for waiting for notification (a user request signal indicating a request for changing a presentation candidate, a user request signal indicating a request for selecting a presentation candidate) is performed (step ST5).
When receiving the notification from the voice output unit 7 or the user request signal input unit 4, the score update unit 5 b of the presentation candidate control unit 5 determines whether the notification has been received from either the voice output unit 7 or the user request signal input unit 4. (Step ST6).
Here, for convenience of explanation, after the time information indicating the voice output start time is notified from the voice output unit 7 at the timing of starting the reproduction of the accent phrase “Kagawa Prefecture”, the syllable “ga” of “Kagaken” is voiced. It is assumed that the user request signal (presentation candidate change request) corresponding to the “next candidate” is notified from the user request signal input unit 4 at the output timing.

When the score update unit 5b of the presentation candidate control unit 5 is notified of time information indicating the audio output start time at the timing of starting the reproduction of the accent phrase “Kagawa Prefecture” from the audio output unit 7 (steps ST5 and ST6). Then, the reproduction position information indicating the reproduction position of the synthesized sound is updated so that it is understood that “the accent phrase“ Kagawa Prefecture ”is currently being reproduced” (step ST7).
When the score update unit 5b of the presentation candidate control unit 5 updates the reproduction position information indicating the reproduction position of the synthesized sound, the score update unit 5b determines whether or not the reproduction of the synthesized sound is completed (step ST8).
At this time, since the notification of the time information indicating the sound output end time has not been received from the sound output unit 7, it is determined that the reproduction of the synthesized sound is not completed, and the process returns to step ST5.

Next, the score update unit 5b of the presentation candidate control unit 5 outputs the user request signal (“candidate”) corresponding to the “next candidate” from the user request signal input unit 4 at the timing when the syllable “ga” of “Kagaken” is output by voice. Upon receiving the notification of a presentation candidate change request (steps ST5 and ST6), it is determined whether or not the user request signal is a presentation candidate selection request (step ST10).
If the user request signal is a request for selecting the presentation candidate 82 (a request to be determined as the presentation candidate 82), the series of processing ends. At this point, the user request signal notified from the user request signal input unit 4 is Since it is a user request signal (request for changing a presentation candidate) corresponding to “next candidate”, the process proceeds to step ST11.

In step ST11, the score update unit 5b of the presentation candidate control unit 5 has the user request signal notified from the user request signal input unit 4 as a user request signal (presentation candidate change request) corresponding to “next candidate”. It is determined whether or not.
At this time, as described above, the user request signal notified from the user request signal input unit 4 is the user request signal corresponding to the “next candidate” (request for changing the presentation candidate), and therefore the I + 1th and subsequent presentations The score of the candidate (including the I + 1th presentation candidate) is re-evaluated (step ST13).
At this time, since I = 1, the scores of the second to eighteenth presentation candidates are reevaluated.

That is, the score update unit 5b of the presentation candidate control unit 5 outputs the user request signal (presentation) corresponding to the “next candidate” from the user request signal input unit 4 at the timing when the syllable “ga” of “Kagaken” is output by voice. Since the notification of “candidate change request” has been received, the accent phrase “Kagawa Prefecture” is determined as the score-lowering word (step ST31 in FIG. 7).
When the score update unit 5b of the presentation candidate control unit 5 determines a score-lowering word (step ST32), the score of the presentation candidate recorded in the presentation candidate list 81 is corrected (step ST33).
Hereinafter, a correction example of the score will be specifically described. In the score correction calculation formula of FIG. 8, it is assumed that a = 0.9 and b = 0.5.

As a specific evaluation value, when the “scoring-down word” is determined, the value of α _i is incremented by 1 when there is a matching word at the corresponding position with respect to the word immediately before the down-scoring word. .
On the other hand, if there is a matching word at the corresponding position with respect to the “scoring-down word”, the value of β _i is incremented by 1.
For the “score-lowering word”, the maximum syllable length that partially matches even if it does not match as a word is obtained, and (maximum syllable length / total syllable length) is added to β _i . However, the initial values are 0 for both α _i and β _i .

In the case of the presentation candidate 81, since the first accent phrase “Kagawa Prefecture” is a score-lowering word, for the presentation candidate including the accent phrase “Kagawa Prefecture” at the head as the corresponding part, the value of β _i is incremented from 0 to 1 Then, the score is recalculated as 1.
On the other hand, Kanagawa Prefecture, in the accent phrase “Kanagawa Prefecture”, matches four syllable “Gawaken” as the longest continuous syllable, so 0.66 (= 4/6) is similarly added to β _i .
As described above, the scores of all the presentation candidates recorded in the presentation candidate list 81 are recalculated, and the presentation candidate list 81 is changed to the presentation candidate list 91.
Thereby, the presentation candidate beginning with the accent phrase “Kagawa Prefecture” is changed to the thirteenth and subsequent.
When the score updating unit 5b of the presentation candidate control unit 5 changes the presentation rank of the presentation candidates recorded in the presentation candidate list 81 as described above, the presentation candidate list 91 reflecting the changed presentation rank. Is reset in the presentation candidate list holding unit 2 (step ST34).

The presentation candidate selection unit 5a of the presentation candidate control unit 5 sets the presentation candidate list 91 reflecting the changed presentation order in the presentation candidate list holding unit 2 when the score update unit 5b resets the presentation candidate list holding unit 2, and then presents the presentation candidate to be presented next. Is selected, the current information presentation position variable I is incremented (step ST14).
At this time, since I = 1, I = 2 by I ← I + 1.

The presentation candidate selection unit 5 a of the presentation candidate control unit 5 selects the I-th presentation candidate from the presentation candidates recorded in the presentation candidate list 91.
At this time, since I = 2, the presentation candidate 92 whose facility name as the second presentation candidate is “Samukawa Kanchu Honpo, Kanagawa” is selected, and the reading information of the presentation candidate 92 is sent to the speech synthesizer 6. Output.
In this example, for convenience of explanation, the user request signal corresponding to the “next candidate” from the user request signal input unit 4 at the timing when the syllable “M” of “Samkawa Monaka Honpo” is output as a voice is changed. Request) is notified to the presentation candidate control unit 5.

  In the speech synthesis unit 6, the speech output unit 7, and the score update unit 5 b of the presentation candidate control unit 5, the same processing as in the case of the presentation candidate 82 is performed for the presentation candidate 92. The score update unit 5b of the presentation candidate control unit 5 corresponds to the “next candidate” from the user request signal input unit 4 at a timing when the voice output unit 7 outputs the syllable “mu” of “Samkawa Monaka Honpo” by voice. Since the notification of the user request signal (presentation candidate change request) is received, in step ST31 of FIG. 7, the accent phrase “Samukawa-Nakachu Honpo” is determined as the score-lowering word.

The correction of the score here is that the accent phrase “Kanagawa Prefecture” precedes the score-decreasing word, so α _i of the corresponding part of the candidate that starts with “Kanagawa Prefecture” is _added by 1 and the accent phrase “Samukawa Saiban” 1 is added to β _i of the corresponding part of the candidate including “Nakahonpo”.
For the score of the corresponding part of the candidate for presentation including “Samukawa”, the score is corrected by calculating the ratio with “Samukawa most central Honpo”.
As described above, the scores of all the presentation candidates recorded in the presentation candidate list 91 are recalculated, and the presentation candidate list 91 is changed to the presentation candidate list 101. Thereby, the scores of the presentation candidates 92 to 95 including “Samukawa” are lowered.
When the score updating unit 5b of the presentation candidate control unit 5 changes the presentation rank of the presentation candidates recorded in the presentation candidate list 91 as described above, the presentation candidate list 101 reflecting the changed presentation rank. Is reset in the presentation candidate list holding unit 2 (step ST34).

The presentation candidate selection unit 5a of the presentation candidate control unit 5 resets the presentation candidate list 101 reflecting the changed presentation rank in the presentation candidate list holding unit 2 when the score update unit 5b resets the presentation candidate to be presented next. Is selected, the current information presentation position variable I is incremented (step ST14).
At this time, since I = 2, I = 3 by I ← I + 1.

The presentation candidate selection unit 5 a of the presentation candidate control unit 5 selects the I-th presentation candidate from the presentation candidates recorded in the presentation candidate list 101.
At this time point, since I = 3, the presentation candidate 102 whose facility name is “Tadayo Aikawa-machi, Kanagawa” as the third presentation candidate is selected, and the reading information of the presentation candidate 102 is output to the speech synthesizer 6. To do.
In this example, for convenience of explanation, a user request signal (request for changing a presentation candidate) corresponding to “next candidate” is output from the user request signal input unit 4 at the timing when the syllable “shi” of “Tashiro” is output as voice. It is assumed that the presentation candidate control unit 5 is notified.

  In the speech synthesis unit 6, the speech output unit 7, and the score update unit 5 b of the presentation candidate control unit 5, the same processing as that of the presentation candidates 82 and 92 is performed for the presentation candidate 102. In this case, the score update unit 5b of the presentation candidate control unit 5 corresponds to the “next candidate” from the user request signal input unit 4 at the timing when the voice output unit 7 outputs the syllable “shi” of “Tashiro”. Since the notification of the request signal (request for changing the presentation candidate) is received, the accent phrase “Tashiro” is determined as the score-lowering word in step ST31 of FIG.

The correction of the score here is because the accent phrases “Kanagawa” and “Aikawa-cho” are in front of the score-decreasing word, so α _i of the corresponding part of the presentation candidate including “Kanagawa” and “Aikawa-cho” is 1 For presentation candidates including the accent phrase “Tashiro”, β _i is added to the syllable length ratio.
As described above, the scores of all the presentation candidates recorded in the presentation candidate list 101 are recalculated, and the presentation candidate list 101 is changed to the presentation candidate list 111. Thereby, the ranking of the presentation candidate 103 is increased by 1, and the ranking of the presentation candidate 104 is decreased.
When the score update unit 5b of the presentation candidate control unit 5 changes the presentation rank of the presentation candidates recorded in the presentation candidate list 101 as described above, the presentation candidate list 111 that reflects the changed presentation rank is reflected. Is reset in the presentation candidate list holding unit 2 (step ST34).

Finally, after the presentation candidate 112 whose facility name is “Aikawacho Tanagawa, Kanagawa Prefecture”, which is the fourth presentation candidate recorded in the presentation candidate list 111, is voice-output, the user operates the “select” function button. When the user request signal (presentation candidate selection request) corresponding to the “select” function button is notified from the user request signal input unit 4 to the presentation candidate control unit 5, the presentation candidate 112 is finally selected. It recognizes that it is a presentation candidate and ends a series of processing.
In the case of the second embodiment, “Tanazawa, Kanagawa Pref.”, Which was the tenth presentation candidate in the presentation candidate list 81, is changed to the fourth presentation candidate in the presentation candidate list 111, and the desired presentation can be performed with few operations. Candidate 112 can be selected.

  In this Embodiment 2, even when there is a presentation candidate that matches before the word that the user has determined to be inappropriate, it has been shown that the presentation order is changed by recalculating the score. You may make it present preferentially from the thing with the same front.

Embodiment 3 FIG.
In the first and second embodiments, the timing at which the score update unit 5b of the presentation candidate control unit 5 receives the notification of the user request signal (presentation candidate change request) corresponding to the “next candidate” from the user request signal input unit 4 In the above description, what determines the score-lowering word has been described. However, there is a time lag (delay time) from when the user listens to the synthesized sound output from the voice output unit 7 until the user operates the function button corresponding to the “next candidate”. Will occur.
Therefore, when listening to an accent phrase and trying to determine that accent phrase as a score-lowering word and operating the function button corresponding to “next candidate”, the voice output of that accent phrase has already ended. In some cases, the accent phrase cannot be determined as a score-reduced word.
Therefore, in the third embodiment, the score update unit 5b of the presentation candidate control unit 5 receives a notification of a user request signal (presentation candidate change request) corresponding to the “next candidate” from the user request signal input unit 4. The score-lowering word is determined on the assumption that the user request signal (request for changing the presentation candidate) is received at a timing earlier than the timing by the delay time t1 (predetermined time).

Hereinafter, the processing content of the score update part 5b of the presentation candidate control part 5 is demonstrated concretely.
FIG. 11 shows the relationship between the synthesized speech generated by the speech synthesizer 6 and the time information indicating the speech output start time of each accent phrase, the speech output end time of the synthesized speech, and the input time of the change request for the presentation candidate. It is explanatory drawing which is.
In the figure, 21 represents a synthesized sound, and the synthesized sound 21 shows an example composed of three accent phrases “XX donut”, “Higashikanagawa”, and “Station front shop”.
23 is the voice output start time of the accent phrase “XX donut”, 24 is the voice output start time of the accent phrase “Higashikanagawa”, 25 is the voice output start time of the accent phrase “Ekimae Shop”, and 26 is the voice of the synthesized sound 21 Indicates the output end time.
131 indicates the delay time t1, 132 to 134 indicate the score reduction word switching time considering the delay time t1, and 135 indicates the delay consideration score reduction range of “XX donut”. Further, 136 indicates a notification time of a user request signal (presentation candidate change request) corresponding to “next candidate”.

The voice output unit 7 notifies the presentation candidate control unit 5 of time information indicating the voice output start time 23 of the accent phrase “XX donut” when starting the reproduction of the synthesized sound of “XX Donut Higashikanagawa Station”. To do.
When the reproduction of the accent phrase “XX donut” is completed, the audio output unit 7 notifies the presentation candidate control unit 5 of time information indicating the audio output start time 24 of the accent phrase “Higashi Kanagawa”.
Here, it is assumed that the user operates the function button corresponding to “next candidate” simultaneously with the completion of the reproduction of the accent phrase “XX donut”.
In this case, since the user listens to the synthesized sound of the accent phrase “XX donut” and then operates the function button corresponding to “next candidate”, a delay time occurs, and the user request signal (request to change the presentation candidate) is generated. The timing notified to the presentation candidate control unit 5 is the timing at which the synthesized sound of the accent phrase “Higashi Kanagawa” is output, as indicated by the notification time 136 in FIG. 11.

Therefore, when the score update unit 5b of the presentation candidate control unit 5 receives a notification of a user request signal (request for changing a presentation candidate) corresponding to the “next candidate” from the user request signal input unit 4, the timing at which the notification is received The score-lowering word is determined based on the notification of the user request signal (presentation candidate change request) at a timing earlier by the delay time t1.
Therefore, in the example of FIG. 11, the accent phrase “XX donut” is determined as the score-lowering word.
According to the third embodiment, there is an effect that it is possible to determine a score-lowering word according to the user's intention.

Embodiment 4 FIG.
In the third embodiment, the score update unit 5b of the presentation candidate control unit 5 receives the notification of the user request signal (presentation candidate change request) corresponding to the “next candidate” from the user request signal input unit 4. In the above description, the score-decreasing word is determined as being received by the user request signal (request for changing the presentation candidate) at the timing earlier by the delay time t1, but the score update unit 5b is finally executed by the presentation candidate selection unit 5a. The delay time t1 may be adjusted in consideration of the presentation candidate selected in (1).

FIG. 12 is a flowchart showing the adjustment process of the delay time t1.
FIG. 13 is an explanatory diagram showing a presentation candidate list in which a voice output end time and a “next candidate” presentation request time of a score-lowering word are recorded in addition to a plurality of presentation candidates.
121 is a presentation candidate list that has been updated by the processing of the second embodiment, and 122 to 124 are presentation candidates in which the voice output end time and the “next candidate” presentation request time of the score-lowering word are recorded.
Although FIG. 13 illustrates the updated presentation candidate list 121 by the processing of the second embodiment, the updated presentation candidate list by the processing of the first and third embodiments may be used.

Next, the operation will be described.
The score update unit 5b of the presentation candidate control unit 5 sets the presentation candidate finally selected by the presentation candidate selection unit 5a as a comparison source.
Here, it is assumed that the presentation candidate “Aikawacho Tanagawa, Kanagawa” is finally selected, and the presentation candidate “Aikawacho Tanagawa, Kanagawa” is set as the comparison target.

When the score update unit 5b of the presentation candidate control unit 5 sets the finally selected presentation candidate as a comparison target source, the comparison target source is compared with the other presentation candidates 122 to 124 from the front, and the other presentation candidates In 122 to 124, first, a word different from the comparison source is extracted (step ST41).
In FIG. 13, the underlined part of the facility name represents a different word.

When the score update unit 5b of the presentation candidate control unit 5 first extracts a word that is different from the comparison target, the underlined part in FIG. 13 indicates the time when the voice output ends and the “next candidate” presentation request time from the user. Each of them is compared, and if “next candidate” presentation request time> the voice end time, the difference is calculated (step ST42).
For example, in the case of the presentation candidate 122, the voice end time is 570 ms, the “next candidate” presentation request time is 720 ms, and the “next candidate” presentation request time> the voice end time is satisfied, so the difference 150 ms is obtained.
In the case of the presentation candidate 124, since the voice end time = 1630 ms, the “next candidate” presentation request time = 1920 ms, and the “next candidate” presentation request time> the voice end time is satisfied, the difference 290 ms is obtained.
On the other hand, in the case of the presentation candidate 123, since the voice end time = 1620 ms, the “next candidate” presentation request time = 11210 ms, and the “next candidate” presentation request time> the voice end time is not satisfied, the difference is not obtained.

The score update unit 5b of the presentation candidate control unit 5 obtains an average difference of 220 ms in the presentation candidates 122 and 124.
The score updating unit 5b resets the delay time t1 in consideration of the average difference of 220 ms from the current delay time t1 (step ST43).
For example, an arithmetic average value between the current delay time t1 and the average difference of 220 ms is set as the reset value of the delay time t1.
If the current delay time t1 is 350 ms, the arithmetic average value 260 ms is reset to the delay time t1.
According to the fourth embodiment, the delay time t1 can be adjusted according to the user's corresponding operation speed.

It is a block diagram which shows the audio | voice information selection apparatus by Embodiment 1 of this invention. It is a flowchart which shows the audio | voice information selection method by Embodiment 1 of this invention. It is a flowchart which shows the detail of step ST13 of FIG. It is explanatory drawing which shows the relationship between the synthetic | combination sound produced | generated by the audio | voice synthesis | combination part, and the time information which shows the audio | voice output start time of each accent phrase, the audio | voice output end time of a synthetic | combination sound, and the input time of the presentation candidate change request. is there. It is explanatory drawing which shows the presentation candidate list | wrist in which the some presentation candidate is recorded as a search result when a periphery facility is searched by the car navigation system. It is explanatory drawing explaining the score reduction calculation method in case a score reduction word is a place name. It is a flowchart which shows the detail of step ST13 of FIG. It is explanatory drawing which shows an example of the calculation formula of score correction. It is explanatory drawing which shows the presentation candidate list in which the several presentation candidate is recorded as a speech recognition result of a car navigation system. It is explanatory drawing which shows the presentation candidate list with which the presentation candidate list of FIG. 9 was updated. It is explanatory drawing which shows the relationship between the synthetic | combination sound produced | generated by the audio | voice synthesis | combination part, and the time information which shows the audio | voice output start time of each accent phrase, the audio | voice output end time of a synthetic | combination sound, and the input time of the presentation candidate change request. is there. It is a flowchart which shows the adjustment process of delay time t1. It is explanatory drawing which shows the presentation candidate list | wrist where the audio | voice output end time of the score reduction word and the "next candidate" presentation request | requirement time are recorded other than a some presentation candidate.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 Presentation candidate list input terminal 2 Presentation candidate list holding part (presentation candidate list holding means) 3 User request input terminal 4 User request signal input part 5 Presentation candidate control part 5a Presentation candidate selection part (presentation candidate selection means ), 5b, score update unit (score update unit), 6 speech synthesis unit (speech output unit), 7 speech output unit (speech output unit), 8 speaker (speech output unit).

Claims (7)

  A presentation candidate list holding means for holding a presentation candidate list in which scores of a plurality of presentation candidates are recorded, and a presentation candidate having the highest score from the presentation candidate list held in the presentation candidate list holding means is selected. A presentation candidate selection unit, a voice output unit that generates a voice representing the name of the presentation candidate selected by the presentation candidate selection unit, and outputs a voice representing the name of the presentation candidate; a presentation candidate in the voice output unit When the request for changing the presentation candidate is received before the output of the voice representing the name of the presentation is completed, the score of the presentation candidate recorded in the presentation candidate list is updated, and the presentation candidate selection unit re-selects the presentation candidate. A voice information selection device comprising score update means for instructing the user.   The score updating means includes a score update algorithm in the case where the presentation candidates recorded in the presentation candidate list are search results retrieved from the database under a predetermined search condition, and the presentation candidates recorded in the presentation candidate list. The speech information selection device according to claim 1, wherein a score update algorithm for a speech recognition result for a user's speech is different.   If the presentation candidate recorded in the presentation candidate list is a search result searched from the database under a predetermined search condition, the score update means is configured to display the above presentation if the presentation candidate that has received the change request includes a place name. 3. The audio information selection device according to claim 2, wherein a score of another presentation candidate is updated in consideration of a geographical positional relationship between the candidate and another presentation candidate.   The score update means, when the presentation candidate recorded in the presentation candidate list is a speech recognition result for the user's voice, when the change request is received among the words constituting the presentation candidate that has received the change request 3. The score of another presentation candidate that includes a word that matches a previous word is increased, and the score of another presentation candidate that includes a word that matches the word at the time of receiving the change request is lowered. Voice information selection device.   The score updating means handles a request for changing a candidate for presentation as if it has received a request for changing a candidate for presentation by a predetermined time earlier than the timing of actually receiving a request for changing a candidate for presentation. The voice information selection device according to any one of claims 1 to 4.   6. The speech information selection apparatus according to claim 5, wherein the score update means adjusts the predetermined time in consideration of the presentation candidate finally selected by the presentation candidate selection means.   A presentation candidate list holding step in which a presentation candidate list holding unit holds a presentation candidate list in which scores of a plurality of presentation candidates are recorded; and a presentation candidate list holding unit holding a presentation candidate list in the presentation candidate list holding unit. A presentation candidate selection step for selecting a presentation candidate having the highest score from the above, and a voice that represents the name of the presentation candidate selected by the presentation candidate selection means by the voice output means, and represents the name of the presentation candidate And when the score update means receives a request to change the presentation candidate before the output of the voice representing the name of the presentation candidate in the voice output means is completed, the presentation recorded in the presentation candidate list A voice information selection method comprising: a score update step for updating a candidate score and instructing the presentation candidate selection means to reselect a presentation candidate .

Images