JP2004271979A - Voice synthesizer - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To fast forwardly reproduce a synthesized voice with a high audibility in such a way that a user can easily understand contents of an entire sentence. <P>SOLUTION: Text information stored in a text storage device 100 is inputted to a morphological analysis device 102 and a parsing device 103 by a text acquisition device 101 and is converted to reading information through a read analysis device 104 and is inputted to a synthesized voice waveform generator 105. The synthesized voice waveform generator 105 generates a synthesized voice waveform by using reading information inputted from the read analysis device 104, morphological analysis information inputted from the morphological analysis device 102, and voice information stored in a voice information database 107 and outputs the synthesized voice waveform to a waveform buffer device 108. The synthesized voice waveform generator 105 thins the generated synthesized voice waveform per morpheme. The outputted synthesized voice waveform is reproduced as a voice from a speaker device 109. <P>COPYRIGHT: (C)2004,JPO&NCIPI

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a speech synthesizer for generating speech information corresponding to text information from the text information.
[0002]
[Prior art]
Conventional audio reproducing apparatuses and audio synthesizing apparatuses typically use the following method when trying to make a user output audio in a short time.
(1) To increase the speed of the output sound and reduce the reproduction time (Patent Document 1).
(2) The audibility is ensured while shortening the reproduction time by shortening the steady part where the voice component is little changed (Patent Document 2).
(3) The audio to be reproduced is divided at certain time intervals, and only a part of each time interval is reproduced, thereby ensuring the audibility while shortening the reproduction time (Patent Document 3).
(4) Reproduction of only the beginning of a chapter or paragraph greatly reduces the reproduction time (Patent Document 4).
[0003]
[Patent Document 1]
JP-A-11-52985
[Patent Document 2]
JP-A-61-55698
[Patent Document 3]
Japanese Patent No. 2048762
[Patent Document 4]
JP-A-2-45868
[0004]
[Problems to be solved by the invention]
The prior art cannot meet the demands of ensuring audibility, significantly shortening the playback time, and making the entire content of the played sound easier for the user to understand. That is, the above-described conventional techniques have the following problems.
(1) If the speed of the output sound is extremely high, the audibility is impaired.
(2) Only the length of the steady part included in the reproduced sound can shorten the reproduction time of the sound.
(3) In general, the temporal and linguistic divisions of audio do not match, so that the audio is difficult to understand even if the audibility is high.
(4) Since useful information for the user is not always at the beginning of a chapter or paragraph, it is difficult to make the whole sentence understood by this reproduction method.
[0005]
[Means for Solving the Problems]
The present invention has been made in view of the above problems, and finds a high importance portion by linguistically analyzing a text input to a speech synthesizer, and connects and reproduces only a high importance portion. Things. As a result, the user can easily understand the contents of the entire sentence and perform fast forward reproduction of the synthesized sound with high audibility.
[0006]
Also, by applying this method to each sentence of the input text and playing back the text in a sentence unit in the reverse direction, it is possible to realize a high-speed rewind playback of the synthesized sound that is easy for the user to understand.
[0007]
A first speech synthesizer according to the present invention comprises: a text storage unit for storing text information; a text acquisition unit for acquiring text information stored in the text storage unit; and a text information acquired by the text acquisition unit. A morphological analysis unit that performs a morphological analysis process on the basis of the text information acquired by the text acquisition unit and a result of the morphological analysis process by the morphological analysis unit, the text information corresponding to text information thinned out in morpheme units. And a speaker for converting an audio waveform generated by the synthesized audio waveform generating means into an audible sound wave and outputting the audible sound wave.
[0008]
The above speech synthesizer is characterized in that a morpheme unit is thinned out for a synthetic sound waveform generated by a synthetic sound waveform generating means using a morphological analysis result of text information to be subjected to voice synthesis. This makes it possible to increase the reproduction speed of the synthesized sound while ensuring that the user can hear the text information that is the source of the synthesized sound at least in units of morphemes.
[0009]
A second speech synthesizer according to the present invention comprises: a text storage unit for storing text information; a text acquisition unit for acquiring text information stored in the text storage unit; and a text information acquired by the text acquisition unit. And parsing the text information according to the syntactic importance based on the text information acquired by the text acquisition means and the result of the syntax analysis processing by the syntax analysis means. And a speaker for converting the speech waveform generated by the synthesized speech waveform generating means into an audible sound wave and outputting the audible sound wave.
[0010]
The above-mentioned speech synthesizer uses the syntax analysis result of text information to be subjected to speech synthesis to thin out a portion having a low syntactic importance to a synthetic sound waveform generated by a synthetic sound waveform generating means. Features. As a result, it is possible to increase the reproduction speed of the synthesized sound by cutting off the less important parts from the text information that is the source of the synthesized sound while removing the parts with the higher syntactic importance.
[0011]
A third speech synthesizer according to the present invention includes a text storage unit for storing text information, a text summarization unit for summarizing the text information stored in the text storage unit, and a text information summarized by the text summarization unit. And a speaker for converting the audio waveform generated by the synthesized audio waveform generator into an audible sound wave and outputting the audible sound wave.
[0012]
The speech synthesizer is characterized in that text information to be subjected to speech synthesis is first summarized, and then speech synthesis is performed. This makes it possible to perform a fast-forward operation that accurately represents the contents of the original text information by a relatively easy configuration change with respect to a normal speech synthesizer, as compared with a method using simple thinning of text information.
[0013]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to FIGS. 1 to 15.
[0014]
(1st Embodiment)
FIG. 1 shows the overall configuration of the speech synthesizer according to the first embodiment. This speech synthesizer performs fast-forward playback by thinning out morpheme units using the result of morphological analysis.
[0015]
The text information stored in the text storage device 100 is input by the text acquisition device 101 to the morphological analysis device 102 and the syntax analysis device 103, becomes the read information via the read analysis device 104, and is input to the synthetic sound waveform generation device 105. . At this time, the linguistic information stored in the linguistic information database 106 is used for analyzing the text information by the morphological analysis device 102, the syntax analysis device 103, and the reading analysis device 104.
[0016]
The synthetic sound waveform generation device 105 generates a synthetic sound waveform using the reading information input from the reading analysis device 104, the morphological analysis information input from the morphological analysis device 102, and the voice information stored in the voice information database 107. , To the waveform buffer device 108. The output synthesized sound waveform is reproduced as sound from the speaker device 109.
[0017]
Further, the apparatus has a speech synthesis control device 111 which can receive an input from a speech synthesis control switch 110 operated by a user, whereby the position of the text acquired by the text acquisition device 101 and the synthesized sound waveform generation device 105 Can control the fast-forward speed of the synthetic sound waveform generation processing by the above.
[0018]
FIG. 2 shows a processing flow until the text sentence 200 received by the text acquisition device 101 from the text storage device 100 is converted into the synthesized sound 204 by the synthesized sound waveform generation device 105.
[0019]
First, the morphological analyzer 102 analyzes the input text sentence 200 based on the information stored in the language information database 106, and obtains information obtained by dividing the text sentence 200 into morpheme units. Hereinafter, this is referred to as morphological analysis information 201.
[0020]
Further, using the morphological analysis information 201 and the text sentence 200, the syntactic analysis device 103 parses the text sentence 200 based on the information stored in the linguistic information database 106, and what is the text sentence 200 divided into morpheme units. Investigate whether it is structured. Hereinafter, the information obtained here is referred to as syntax analysis information 202.
[0021]
Next, the reading analysis device 104 analyzes the text sentence 200 using the linguistic information database 106, the morphological analysis information 201, and the syntax analysis information 202, and determines a phoneme sequence corresponding to the reading of the text sentence 200, an accent position, and the like. The reading information 203 for notifying the generating device 105 is generated.
[0022]
Finally, the synthesized sound waveform generating apparatus 105 outputs the synthesized sound waveform 204 corresponding to the reading of the text 200 using the reading information 203 and the voice information database 107.
[0023]
Here, since the synthesized sound waveform generating apparatus 105 knows which part of the reading information 203 corresponds to which part of the morphological analysis information 201, the synthesized sound waveform to be generated can be thinned out for each morpheme. FIG. 3 shows an example of processing for shortening the generated synthetic sound waveform by thinning out morpheme units. Here, when a synthesized sound waveform is generated for the entire text sentence 200, a synthesized sound waveform 204 is generated. However, the synthesized sound waveform 300 can be generated by thinning out morpheme units. In the processing shown in this figure, first, after removing punctuation marks and particles / auxiliary verbs from the original text sentence 200 subjected to morphological analysis, the remaining morphemes are taken out every two to generate a synthetic sound waveform 300. The synthesized sound waveform 300 may be generated by generating only a waveform of a necessary portion, or may be generated by first generating the synthesized sound waveform 204 and extracting a waveform of the required portion therefrom.
[0024]
The operation flow of this device is shown in FIGS.
[0025]
FIG. 4 shows the operation of the speech synthesis control device 111 in the device of the present embodiment. The speech synthesis control device 111 operates roughly as follows.
If there is an input from the voice synthesis control switch 110, a corresponding signal is transmitted to the text acquisition device 101 and the synthesized sound waveform generation device 105.
If the text acquiring device 101 can acquire the next text, a signal for acquiring the next text is transmitted to the text acquiring device 101.
When the reproduction of the end of the text information stored in the text storage device 100 is completed, a stop signal is transmitted to the text acquisition device 101 and the synthetic sound waveform generation device 105.
[0026]
The operation content of each step in the operation flow of FIG. 4 is as follows.
S400: Operation of the device starts.
S401: Various variables used inside the speech synthesis control device 111 are cleared.
S402: It is determined whether or not there is an input from the voice synthesis control switch 110 to stop the reproduction of the synthesized sound.
S403: It is determined whether or not there is an input from the voice synthesis control switch 110 to increase the reproduction speed of the synthesized sound by one step.
S404: It is determined whether or not there is an input from the voice synthesis control switch 110 to instruct to decrease the reproduction speed of the synthesized sound by one step.
S405: It is determined whether or not there is an input from the voice synthesis control switch 110 to instruct to return the reproduction speed of the synthesized sound to the initial value.
S406: It is determined whether the text acquisition device 101 is in a state where the next text information can be acquired. The text acquisition device 101 can acquire the next text information because the text acquisition device 101 has not yet acquired the text information from the text storage device 100 or the text information acquired by the text acquisition device 101 has already been acquired. This is the time when the data has been passed to both the morphological analyzer 102 and the syntax analyzer 103.
S407: A signal is transmitted to the synthetic sound waveform generating apparatus 105 to instruct to increase the reproduction speed of the synthetic sound by one step.
S408: A signal for instructing the synthetic sound waveform generating apparatus 105 to reduce the reproduction speed of the synthetic sound by one step is transmitted.
S409: A signal is transmitted to the synthesized sound waveform generating apparatus 105 to instruct to return the reproduction speed of the synthesized sound to the initial value.
S410: It is determined whether the text information acquired last by the text acquisition device 101 is at the end of the text information stored in the text storage device 100.
S411: The text acquisition device 101 transmits a signal instructing to acquire text information immediately after the text information acquired last.
S412: It is determined whether or not the reproduction of the synthesized sound from the waveform buffer device 108 by the speaker device 109 has been completed.
S413: A signal for instructing to stop the operation is transmitted to the text acquisition device 101 and the synthetic sound waveform generation device 105.
S414: The operation of the device stops.
[0027]
FIG. 5 shows the operation of the text acquisition device 101 in the device of the present embodiment. The text acquisition device 101 operates as follows.
When a signal for instructing acquisition of the next text information in the text storage device 100 is received from the speech synthesis control device 111, the instruction is followed.
When the text information is obtained from the text storage device 100, the text information is transmitted to the morphological analysis device 102 and then to the syntax analysis device 103.
When receiving a signal from the voice synthesis control device 111 to support the stop of the operation, follow the instruction.
[0028]
The operation content of each step in the operation flow of FIG. 5 is as follows.
S500: Operation of the device starts.
S501: Various variables used inside the text acquisition device 101 are cleared.
S502: It is determined whether or not a signal for instructing to stop the operation has been received from the speech synthesis control device 111.
S503: It is determined whether a signal instructing acquisition of the next text information has been received from the speech synthesis control device 111.
S504: In the text storage device 100, the text information held immediately after the text information obtained last is obtained. After starting the operation, if the text information has not been obtained at least once, the text information held at the head position in the text storage device 100 is obtained.
S505: It is determined whether the morphological analyzer 102 is in a state where the next text information can be acquired. The morpheme analyzer 102 can acquire the next text information because the morpheme analyzer 102 has not yet acquired text information from the text acquirer 101 or the morpheme of the text information acquired last by the morpheme analyzer 102. This is the time when the analysis processing has already been completed and the morphological analysis results have been passed to all of the syntax analysis device 103, the reading analysis device 104, and the synthesized sound waveform generation device 105.
S506: The currently held text information is transmitted to the morphological analyzer 102.
S507: It is determined whether or not the syntax analysis device 103 is in a state where the next text information can be acquired. It should be noted that the syntax analysis device 103 can acquire the next text information because the syntax analysis device 103 has not yet acquired any information from the text acquisition device 101 and the morphological analysis device 102. This is when the result of the performed syntax analysis processing has already been passed to the reading analysis apparatus 104.
S508: The currently stored text information is transmitted to the syntax analysis device 103. S509: The operation of the device stops.
[0029]
FIG. 6 shows the operation of the morphological analysis device 102 in the device of the present embodiment. The morphological analyzer 102 operates as follows.
Receives text information from the text acquisition device 101, performs morphological analysis processing, and transmits the morphological analysis information to the syntax analysis device 103, the reading analysis device 104, and the synthetic sound waveform generation device 105. In the morphological analysis process, the linguistic information stored in the linguistic information database 106 is used.
[0030]
The operation content of each step in the operation flow of FIG. 6 is as follows.
S600: Operation of the device starts.
S601: Various variables used inside the morphological analyzer 102 are cleared.
S602: It is determined whether the text acquisition device 101 has stopped operating.
S603: The text acquisition device 101 determines whether or not the next text information can be transmitted to the morphological analysis device 102.
S604: The text information transmitted from the text acquisition device 101 is received.
S605: The morphological analysis processing of the text information received in S604 is performed with reference to the language information database 106.
S606: It is determined whether or not the syntax analysis device 103 is in a state where the next morphological analysis information can be acquired. Note that the syntactic analysis device 103 can acquire the next morphological analysis information because the syntactic analysis device 103 has not yet obtained the morphological analysis information from the morphological analysis device 102 or has obtained the last morphological analysis information from the morphological analysis device 102. This is the time when the result of the syntax analysis performed using the morphological analysis information obtained has already been passed to the reading and analyzing apparatus 104.
S607: The morphological analysis information generated in S605 is transmitted to the syntax analysis device 103.
S608: It is determined whether the reading analysis device 104 is in a state where the next morphological analysis information can be acquired. Note that the reading analysis apparatus 104 can acquire the next morphological analysis information because the reading analysis apparatus 104 has not yet acquired the morphological analysis information from the morphological analysis apparatus 102 or has obtained the last morphological analysis information from the morphological analysis apparatus 102. This is the time when the result of the reading analysis process performed using the obtained morphological analysis information has already been passed to the synthetic sound waveform generator 105.
S609: The morphological analysis information generated in S605 is transmitted to the reading analysis device 104.
S610: It is determined whether or not the synthesized sound waveform generating apparatus 105 is in a state where the next morphological analysis information can be acquired. Note that the synthesized morpheme analysis device 105 can acquire the next morphological analysis information only when the synthesized morpheme analysis device 105 has not yet acquired the morpheme analysis information from the morpheme analysis device 102. This is when the generation of the synthetic sound waveform using the morphological analysis information acquired last from the step has already been completed.
S611: The morphological analysis information generated in S605 is transmitted to the synthetic sound waveform generator 105.
S612: The operation of the device stops.
[0031]
FIG. 7 shows the operation of the syntax analysis device 103 in the device of the present embodiment. The syntax analyzer 103 operates as follows.
It receives text information from the text acquisition device 101 and morphological analysis information from the morphological analysis device 102, performs a syntax analysis process, and reads the syntax analysis information and sends it to the analysis device 104. In the syntax analysis processing, linguistic information stored in the linguistic information database 106 is used.
[0032]
The operation content of each step in the operation flow of FIG. 7 is as follows.
S700: The operation of the device starts.
S701: Various variables used inside the syntax analysis device 103 are cleared.
S702: It is determined whether the text acquisition device 101 has stopped operating.
S703: The text acquisition device 101 determines whether the next text information can be transmitted to the morphological analysis device 102.
S704: The text information transmitted from the text acquisition device 101 is received.
S705: The morphological analysis device 102 determines whether or not the next morphological analysis information can be transmitted to the syntax analysis device 103.
S706: The morphological analysis information transmitted from the morphological analyzer 102 is received.
S707: While referring to the language information database 106, a syntax analysis process of the text information received in S704 is performed using the morphological analysis information received from the morphological analysis device 102 in S706.
S708: It is determined whether the reading analysis device 104 is in a state where the next syntax analysis information can be acquired. Note that the reading analysis apparatus 104 can acquire the next syntactic analysis information because the reading analysis apparatus 104 has not yet obtained the syntactic analysis information from the syntactic analysis apparatus 103 or the syntactic analysis information is obtained last from the syntactic analysis apparatus 103. This is the time when the result of the reading analysis processing performed using the obtained syntactic analysis information has already been passed to the synthetic sound waveform generating apparatus 105.
S709: The syntax analysis information generated in S707 is transmitted to the reading analysis device 104.
S710: The operation of the device stops.
[0033]
FIG. 8 shows the operation of the reading analysis device 104 in the device of the present embodiment. The reading analysis device 104 operates roughly as follows.
Receiving text information from the text acquisition device 101, morphological analysis information from the morphological analysis device 102, and syntactic analysis information from the syntax analysis device 103, and reading the read information generated by referring to the language information database 106; Send to
[0034]
The operation content of each step in the operation flow of FIG. 8 is as follows.
S800: Operation of the device starts.
S801: Various variables used inside the reading analysis device 104 are cleared.
S802: It is determined whether the text acquisition device 101 has stopped operating.
S803: The text acquisition device 101 determines whether or not the next morphological analysis information can be transmitted to the reading analysis device 104.
S804: The text information transmitted from the text acquisition device 101 is received.
S805: The morphological analysis device 102 determines whether or not the next morphological analysis information can be transmitted to the reading analysis device 104.
S806: The morphological analysis information transmitted from the morphological analyzer 102 is received.
S807: The syntax analysis device 103 determines whether or not the next syntax analysis information can be transmitted to the reading analysis device 104.
S808: The syntax analysis information transmitted from the syntax analysis device 103 is received.
S809: The text information received from the text acquisition device 101 in S804, the morphological analysis information received from the morphological analysis device 102 in S806, and the syntax analysis information received from the syntax analysis device 103 in S808 while referring to the language information database 106. To perform reading analysis processing of the original text information to generate reading information.
S810: It is determined whether the synthesized sound waveform generating apparatus 105 is in a state where the next reading information can be acquired. It should be noted that the synthesized sound waveform generating apparatus 105 can acquire the next reading information only when the synthesized sound waveform generating apparatus 105 has not yet acquired the reading information from the reading analyzing apparatus 104 or when the last reading information is obtained from the reading analyzing apparatus 104. This is the case where the generation of the synthetic sound waveform using the read information acquired in step (1) has already been completed.
S811: The reading information generated in S809 is transmitted to the synthetic sound waveform generating device 105.
S812: The operation of the device stops.
[0035]
FIG. 9 shows the operation of the synthetic sound waveform generating device 105 in the device of the present embodiment. The synthetic sound waveform generator 105 operates roughly as follows.
Receives morphological analysis information from the morphological analysis device 102 and read information from the read analysis device 104, and writes the synthesized sound waveform generated with reference to the voice information database 107 to the waveform buffer device 108.
Receives a signal sent from the voice synthesis control device 111, and stops the operation and controls the fast-forward speed of the synthesized sound according to the signal.
[0036]
The operation content of each step in the operation flow of FIG. 9 is as follows.
S900: Operation of the device starts.
S901: Various variables used inside the synthetic sound waveform generating apparatus 105 are cleared. At this time, the reproduction speed of the synthesized sound is initialized to zero.
S902: It is determined whether or not a signal supporting the stop of the operation has been received from the speech synthesis control device 111.
S903: It is determined whether or not a signal instructing to increase the reproduction speed of the synthesized sound by one step from the voice synthesis control device 111 is received.
S904: Increase the reproduction speed of the synthesized sound by one. However, the reproduction speed of the synthesized sound should not exceed 5.
S905: It is determined whether or not there is an input from the speech synthesis control device 111 to instruct to decrease the reproduction speed of the synthesized sound by one step.
S906: Decrease the synthetic sound reproduction speed by one. However, the reproduction speed of the synthesized sound should not fall below 0.
S907: The morphological analyzer 102 determines whether or not the next morphological analysis information can be transmitted to the synthetic sound waveform generator 105.
S908: The morphological analysis information transmitted from the morphological analyzer 102 is received.
S909: The reading analysis device 104 determines whether or not the next reading information can be transmitted to the synthetic sound waveform generating device 105.
S910: Reading information transmitted from the reading analysis device 104 is received.
S911: A synthetic sound waveform is generated using the morphological analysis information received from the morphological analyzer 102 in S908 and the reading information received from the reading analyzer 104 in S910 while referring to the voice information database 107. At this time, if the reproduction speed of the synthetic sound is 0, a synthetic sound waveform is generated for all of the passed reading information. If the reproduction speed of the synthetic sound is 1 or more, the method shown in FIG. After removing punctuation marks and particles / auxiliary verbs, morphemes are further thinned out, and only the synthesized sound waveform corresponding to the remaining morphemes is output. For example, if the reproduction speed of the synthetic sound is 1, a synthetic sound is generated by removing every other morpheme after removing punctuation marks and particles / auxiliary verbs, and if the reproduction speed of the synthetic sound is 2, Similarly, every third morpheme is taken out.
S912: It is determined whether the waveform buffer device 108 is empty. It is to be noted that the waveform buffer device 108 being empty means that the synthesized waveform has not been written to the waveform buffer device even once after the operation of the device has started, or the reproduction of the last synthesized waveform has already been completed. State.
S913: The synthesized sound waveform generated in S911 is written in the waveform buffer device 108.
S914: The operation of the device stops.
[0037]
When the synthesized sound waveform is written in the waveform buffer device 108, the speaker device 109 reproduces the sound waveform. When the reproduction is completed, the speaker device 109 repeats waiting for the next synthesized sound waveform to be written in the waveform buffer device 108 again.
[0038]
With the above configuration and procedure, it is possible to realize a speech synthesizer that can reproduce text information stored in the text storage device 100 as a synthesized sound waveform while performing fast-forwarding by thinning out morpheme units.
[0039]
In this method, by using the result of the morphological analysis performed when generating the synthesized sound waveform, the synthesized sound is thinned out by a very simple algorithm, thereby minimizing the amount of calculation processing newly required for fast-forwarding. It is possible to fast-forward a synthetic sound that is highly audible and easy to understand while suppressing it. This method is suitable for applications such as reading out a mail on a mobile phone or reading out road information using a car navigation system and performing an early listening in an environment where resources available for speech synthesis are limited.
[0040]
In this embodiment, the morpheme thinning is controlled by, for example, a method of “thinning out one morpheme every third”, for example, by starting the reproduction of the synthesized sound at the current fast-forward speed. The morpheme is decimated when the sum of the time lengths of the synthesized sounds that have been reproduced after that exceeds half that of normal playback without fast-forwarding. It is also possible to control.
[0041]
(Second embodiment)
FIG. 10 shows the overall configuration of the speech synthesizer according to the second embodiment. This speech synthesizer performs fast-forward playback by thinning out morpheme units using the result of syntax analysis (syntax analysis information). FIG. 11 shows a data flow in the speech synthesizer. The difference between the apparatus configuration shown in FIG. 1 and the flow of data shown in FIG. 2 is that morphological analysis information is not directly sent from the morphological analyzer 102 to the synthetic sound waveform generator 105, and The point is that syntax analysis information is directly sent to the synthetic sound waveform generating apparatus 105.
[0042]
FIG. 12 shows an example of a shortening process of a synthesized sound waveform by thinning out morpheme units using a result of syntax analysis in the present embodiment. Here, when a synthetic sound waveform is generated for the entire text sentence 200, a synthetic sound waveform 204 is generated. In consideration of the syntax analysis information, a portion having low importance is preferentially thinned out in units of morphemes. A sound waveform 1200 can be generated. At this time, the synthetic sound waveform 1200 may be generated by generating only a waveform of a necessary portion, or may be generated by first generating the synthetic sound waveform 204 and extracting a waveform of the necessary portion therefrom.
[0043]
In the processing of this figure, first, the subject and predicate clauses considered to be essential for the composition of the sentence are extracted from the original text sentence 200 that has been subjected to morphological analysis and syntax analysis, and information on the interdependence between the clauses is further extracted. At a glance, a phrase considered to be highly important is extracted to generate a synthetic sound waveform 1200. A phrase that is considered to be of high importance is, for example, a phrase that modifies a predicate of the entire sentence and modifies the subject or predicate of the entire sentence, or a phrase that ends with a particle such as a Phrases.
[0044]
The flowchart of each component in the present embodiment will be described in the form of a difference from the flowchart of the first embodiment.
[0045]
The operation of the speech synthesis control device 111 is the same as that shown in FIG. 4 of the first embodiment.
[0046]
The operation of the text acquisition device 101 is the same as that shown in FIG. 5 of the first embodiment.
[0047]
The morphological analyzer 102 in the present embodiment performs exactly the same operation as the first embodiment except that it does not need to transmit the morphological analysis information to the synthetic sound waveform generator 105. Therefore, S610 and S611 are omitted from the operation flow shown in FIG. 6, and the process shifts to S602 after the end of S609 is the operation flow of the morphological analyzer 102 in the present embodiment.
[0048]
The parsing device 103 according to the present embodiment performs exactly the same operation as that of the first embodiment except that the parsing information needs to be transmitted to the synthetic sound waveform generating device 105. Therefore, immediately after S709 of the operation flow shown in FIG. 7, a step for further determining whether or not the synthesized sound waveform generating apparatus 105 can receive the syntax analysis information, and transmitting the syntax analysis information to the synthesized sound waveform generating apparatus 105 The operation flow of the syntax analysis device 103 according to the present embodiment is obtained by adding a step for performing the operation.
[0049]
The operation of the reading analysis device 104 is the same as that shown in FIG. 8 of the first embodiment.
[0050]
Compared with the first embodiment, the synthetic sound waveform generating apparatus 105 according to the present embodiment is different from the first embodiment in that, instead of receiving morphological analysis information from the morphological analyzer 102, it receives syntactic analysis information from the syntactic analyzer 103. The flow is exactly the same. Therefore, in the operation flow shown in FIG. 9, S907 for determining whether morphological analysis information can be received from the morphological analysis device 102 and S908 for receiving morphological analysis information, respectively, the syntax analysis information is received from the syntax analysis device S103. The operation flow of the synthesized sound waveform generating apparatus 105 according to the present embodiment has been changed to a step of determining whether or not it can be performed and a step of receiving syntax analysis information. The synthesized sound waveform generating apparatus 105 generates a synthesized sound waveform using the received syntax analysis information in S911. At this time, if the reproduction speed of the synthesized sound is 0, a synthesized sound waveform is generated for all of the passed read information. However, if the reproduction speed of the synthesized sound is 1 or more, as shown in FIG. The synthetic sound waveform 1200 is generated in such a manner that the number of morphemes to be thinned out increases as the reproduction speed of the synthetic sound increases.
[0051]
With the above configuration and procedure, it is possible to realize a speech synthesizer that can reproduce text information stored in the text storage device 100 as a synthesized sound waveform while performing fast-forwarding by thinning out morpheme units based on syntax analysis information.
[0052]
In this method, by using the result of the syntax analysis processing obtained by using the result of the morphological analysis processing, it becomes possible to perform the fast-forward processing reflecting the importance of each morpheme as compared with the method of the first embodiment.
[0053]
(Third embodiment)
FIG. 13 shows the overall configuration of the speech synthesizer according to the third embodiment. This device performs fast-forward playback by summarizing text information to be subjected to speech synthesis processing in advance. FIG. 14 shows the flow of data in this speech synthesizer. The difference between the configuration of the apparatus shown in FIG. 1 and the flow of data shown in FIG. 2 is that in this speech synthesis apparatus, the synthesized speech generation apparatus 105 does not directly use the morphological analysis information 201 but is included in the reading information 203. The synthesis sound waveform 204 is generated using only the information. Instead, a text summarization device 1300 is provided, whereby the text information 200 is first summarized into the text information 1400, and then the speech synthesis process is performed. Is to be shortened.
[0054]
In this configuration, first, the text information input is summarized by the text summarization device 1300 to shorten the length of the synthesized sound waveform that is output. Therefore, in the processing after the text acquisition device 101, the shortening of the synthesized sound waveform is not considered. There is an advantage that it may be.
[0055]
FIG. 15 shows an operation flow of the text summarizing apparatus 1300. The text summarizing apparatus 1300 operates as follows.
The text information is received from the text storage device 100, subjected to a summarization process to reduce the length to less than the original text, and transmitted to the text acquisition device 101.
Receives a signal sent from the voice synthesis control device 111, and stops the operation and controls the fast-forward speed of the synthesized sound according to the signal.
As the fast-forward speed of the synthesized sound specified by the voice synthesis control device 111 increases, the length of the summarized text information for the input text information monotonically decreases.
[0056]
The operation content of each step in the operation flow of FIG. 15 is as follows.
S1500: Operation of the device starts.
S1501: Various variables used inside the text summarization apparatus 1300 are cleared. At this time, the reproduction speed of the synthesized sound is initialized to zero.
S1502: It is determined whether or not a signal supporting stop of the operation has been received from the speech synthesis control device 111.
S1503: It is determined whether or not a signal instructing to increase the reproduction speed of the synthesized sound by one step from the voice synthesis control device 111 is received.
S1504: The reproduction speed of the synthesized sound is increased by one. However, the reproduction speed of the synthesized sound should not exceed 5.
S1505: It is determined whether or not there is an input from the speech synthesis control device 111 to instruct to reduce the reproduction speed of the synthesized sound by one step.
S1506: Decrease the reproduction speed of the synthesized sound by one. However, the reproduction speed of the synthesized sound should not fall below 0.
S1507: It is determined whether or not a signal for instructing acquisition of the next text information has been received from the speech synthesis control device 111.
S1508: In the text storage device 100, the text information held immediately after the text information obtained last is obtained. After starting the operation, if the text information has not been obtained at least once, the text information held at the head position in the text storage device 100 is obtained.
S1509: Summarize the text information acquired in S1508.
S1510: The text acquisition device 101 determines whether the summarized text information generated in S1509 can be acquired. The text acquisition device 101 can acquire text information because the text acquisition device 101 has not yet acquired text information from the text summarization device 1300 or the text information acquired last from the text summarization device 1300 is , Which has already been passed to the morphological analysis device 102, the syntax analysis device 103, and the reading analysis device 104.
S1511: The summarized text information generated in S1509 is transmitted to the text acquisition device 101.
S1512: Operation of the device stops.
[0057]
Hereinafter, the operation flow of each device other than the text summarization device 1300 will be described in the form of a difference from the flowchart of the first embodiment.
[0058]
The speech synthesis control device 111 performs the processing performed on the text acquisition device 101 in S406, S410, S411, S412, and S413 of the operation flow illustrated in FIG. Do. Further, in the present embodiment, the processing that has been performed on the synthesized sound waveform generating apparatus 105 in S407, S408, and S409 is similarly performed on the text summarizing apparatus 1300.
[0059]
The text acquisition device 101 replaces the operation of waiting for a signal from the speech synthesis control device 111 in S503 of the operation flow illustrated in FIG. 5 with the operation of waiting for transmission of text information from the text summarization device 1300 in the present embodiment. I do. Also, the place where the text information was obtained from the text obtaining apparatus 100 in S504 is obtained from the text summarizing apparatus 1300.
[0060]
The morphological analyzer 102 performs the same operation as that of the second embodiment.
[0061]
The operation of the syntax analyzer 103 is the same as that shown in FIG. 7 of the first embodiment.
[0062]
The operation of the reading analysis device 104 is the same as that shown in FIG. 8 of the first embodiment.
[0063]
The operation of the synthetic sound waveform generation apparatus 105 in the present embodiment does not require the processing of S903, S904, S905, S906, S907, and S908 of the operation flow shown in FIG. In the present embodiment, after the process of S902, the determination process of S909 is performed, and as a result of the determination, the process branches to S910 if the next reading information can be transmitted by the reading analysis device 104, and branches to S902 if it cannot. Perform processing.
[0064]
With the above configuration and procedure, it is possible to realize a speech synthesizer that can reproduce text information stored in the text storage device 100 as a synthetic sound waveform while performing shortening processing by the text summarization device 1300 that has been turned into a black box.
[0065]
This embodiment has an advantage that the operation of a part that performs speech synthesis processing from text information can be simplified by leaving the text information shortening to the separately prepared text summarizing apparatus 1300. In addition, the text information summarization process requires more resources than simple thinning processing in morpheme units. However, the text summarization device 1300 analyzes the meaning of the original text information and performs an appropriate summarization process. It is possible to perform fast-forward processing of synthesized sounds that are semantically more accurate than performing simple thinning processing of morphemes and that are well-formed as text.
[0066]
It should be noted that it is also possible to realize a speech synthesizer obtained by combining any two or three of the speech synthesizers described in the first to third embodiments.
[0067]
【The invention's effect】
As described above, according to the present invention, the audibility and ease of understanding of a synthesized sound can be improved by thinning the input text information in units of linguistic delimiters and summarizing the text information in consideration of the meaning. The fast forward processing of the synthesized sound can be performed while ensuring the sound quality. This makes it possible to easily perform operations such as "early listening" when the text is read aloud by the synthetic sound, and "skip" for searching for the target text information from the text storage means.
[Brief description of the drawings]
FIG. 1 is a block diagram illustrating a configuration of a speech synthesis device according to a first embodiment.
FIG. 2 is a diagram showing a data flow in the speech synthesizer shown in FIG.
FIG. 3 is a schematic diagram illustrating an example of a synthetic sound thinning process.
FIG. 4 is a flowchart showing the operation of the speech synthesis control device shown in FIG.
FIG. 5 is a flowchart illustrating an operation of the text acquisition device illustrated in FIG. 1;
FIG. 6 is a flowchart showing an operation of the morphological analyzer shown in FIG. 1;
FIG. 7 is a flowchart showing an operation of the syntax analysis device shown in FIG. 1;
FIG. 8 is a flowchart showing an operation of the reading analysis apparatus shown in FIG. 1;
FIG. 9 is a flowchart showing an operation of the synthetic sound waveform generating device shown in FIG. 1;
FIG. 10 is a block diagram illustrating a configuration of a speech synthesis device according to a second embodiment.
FIG. 11 is a diagram showing a data flow in the speech synthesizer shown in FIG.
FIG. 12 is a schematic diagram illustrating an example of a thinning-out process of a synthesized sound in the speech synthesizer illustrated in FIG. 10;
FIG. 13 is a block diagram illustrating a configuration of a speech synthesis device according to a third embodiment.
FIG. 14 is a diagram showing a data flow in the speech synthesizer shown in FIG.
FIG. 15 is a flowchart showing the operation of the text summarizing apparatus shown in FIG.

Claims (3)

Text storage means for holding text information;
Text acquisition means for acquiring text information held in the text storage means,
A morphological analysis unit that performs a morphological analysis process on the text information acquired by the text acquisition unit,
Synthetic sound waveform generating means for generating a speech waveform corresponding to text information obtained by thinning out the text information in morpheme units based on the text information obtained by the text obtaining means and the result of the morphological analysis processing by the morphological analysis means When,
A speaker for converting an audio waveform generated by the synthetic sound waveform generator into an audible sound wave and outputting the audible sound. Text storage means for holding text information;
Text acquisition means for acquiring text information held in the text storage means,
Syntax analysis means for performing syntax analysis processing on the text information acquired by the text acquisition means,
Based on the text information acquired by the text acquisition means and the result of the syntax analysis processing by the syntax analysis means, an audio waveform corresponding to the text information obtained by thinning out the text information according to the syntactic importance is generated. Synthetic sound waveform generating means;
A speaker for converting an audio waveform generated by the synthetic sound waveform generator into an audible sound wave and outputting the audible sound. Text storage means for holding text information;
Text summarization means for summarizing text information held in the text storage means,
Synthetic sound waveform generating means for generating a speech waveform corresponding to the text information summarized by the text summarizing means,
A speaker for converting an audio waveform generated by the synthetic sound waveform generator into an audible sound wave and outputting the audible sound.

Images