JP3540984B2 - Speech synthesis apparatus, speech synthesis method, and storage medium storing speech synthesis program - Google Patents

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a speech synthesizer, a speech synthesis method, and a storage medium storing a speech synthesis program, and in particular, sets a silence section caused by a difference in word order between languages in simultaneous interpretation to an appropriate pause time. Synthesizing apparatus and method for synthesizing speech by setting an appropriate pause time between silent sentences between sentences of the target language to be speech-synthesized so as to improve the naturalness of synthesized speech of the target language translated from And a storage medium storing a speech synthesis program.
[0002]
[Prior art]
A method for verifying whether or not the simultaneous interpreted voice is natural for the listener without discomfort and easy to understand the content, and a device for performing the same have not been developed so far.
At the time of simultaneous interpretation, if the translated speech obtained by translating the utterance sentence of the original language having a coherent meaning is separated by a long silent section, the listener may feel uncomfortable. This phenomenon can occur when one sentence of the source language is long, in the case of sequential translation that starts translating after the source language is uttered. Also, in the case of a gradual translation, in which the source language starts to be translated each time a phrase is uttered, for example, in the case of a Japanese-English interpreter, Japanese has a different word order from English, and the object precedes the verb. Until the last verb is uttered in Japanese, it cannot be translated into English, and the utterance timing eventually becomes equivalent to the sequential translation. Also in this gradual translation, when the Japanese object is long, the silent section becomes long, and the user may feel uncomfortable. In such a case, if the interpreter is a human, the verb may be predicted and translated from the context of the subject, the first phrase of the object, or the verb.
[0003]
[Problems to be solved by the invention]
According to the present invention, speech synthesis is performed such that a silence section caused by a difference in word order existing between languages in simultaneous interpretation is set to an appropriate pause time to improve the naturalness of synthesized speech of a target language translated from a source language. An object of the present invention is to provide a speech synthesizer, a speech synthesis method, and a storage medium storing a speech synthesis program which are used when a silent section between sentences of a target language to be performed is set to an appropriate pause time to perform speech synthesis. It is assumed that.
[0004]
[Means for Solving the Problems]
Claim 1: A source language database 11 storing information on the utterance start time and utterance end time of each utterance sentence in the source language, and a translation of each utterance in the source language, an utterance start time and an utterance end of the translated sentence A target language database 21 storing time information is acquired. The utterance end time of the source language utterance corresponding to each translated sentence in the target language database 21 is acquired from the source language database 11, and the utterance end time is obtained. An utterance start time assigning unit 31 that assigns a time obtained by adding a certain time to the translated sentence as an utterance start time of the translated sentence, and converts each translated sentence in the target language database 21 into an utterance start time assigning unit A speech synthesizing apparatus including a speech synthesizing unit 51 which is operated in response to the utterance start time obtained in 31 is constructed.
[0005]
Claim 2: a source language database 11 storing information on the utterance start time and utterance end time of each utterance sentence in the source language, and a translation of each utterance in the source language and an utterance start time of the translated sentence; A target language database 21 storing information on the utterance end time is provided. The utterance end time of the source language utterance corresponding to each translation in the target language database 21 is acquired from the source language database 11, and the utterance end is obtained. An utterance start time giving unit 31 is provided for giving a time obtained by adding a certain time to the time as the utterance start time of the translated sentence to the translated sentence, and the utterance time of each translated sentence in the target language database 21 is The utterance end time giving unit 3 calculates the utterance end time of the translated sentence from the utterance time and the utterance start time obtained by the utterance start time giving unit 31, and gives it to the translated sentence. And a speech synthesizing unit 51 which operates each translated sentence in the target language database 21 in response to the utterance start time obtained by the utterance start time giving unit 31 or the utterance end time giving unit 32. A speech synthesizer was constructed.
[0006]
Claim 3: In the voice synthesizing device according to claim 2, the speech ending time assigning unit 32 corresponds to a reversal between the speech ending time of the immediately preceding translation and the speech starting time of the subsequent translation. Thus, a speech synthesizer having a delay processing unit for delaying the utterance start time and the utterance end time of the later translated sentence is configured.
The source language database 11 stores information on the utterance start time and the utterance end time of each utterance sentence in the source language. A target language database 21 storing information on the utterance end time is provided. The utterance end time of the source language utterance corresponding to each translation in the target language database 21 is acquired from the source language database 11, and the utterance end is obtained. An utterance start time giving unit 31 is provided for giving a time obtained by adding a certain time to the time as the utterance start time of the translated sentence to the translated sentence, and the utterance time of each translated sentence in the target language database 21 is The utterance end time giving unit 32 calculates the utterance end time of the translated sentence from the utterance time and the utterance start time obtained by the utterance start time giving unit 31, and gives the utterance end time to the translated sentence. The length of a silent section between each sentence in the target language database 21 is calculated based on the time information obtained by the utterance start time giving unit 31 and the utterance end time giving unit 32, and the length is A modified utterance time providing unit 41 for providing a value obtained by correcting the utterance start time and the utterance end time earlier to a later translated sentence, A speech synthesizing apparatus comprising a speech synthesizing unit 51 that operates a translated sentence in response to the utterance start time obtained by the utterance start time giving unit 31, the utterance end time giving unit 32, or the corrected utterance time giving unit 41 did.
[0007]
Claim 5: In the voice synthesizing apparatus according to claim 4, the utterance end time giving unit 32 responds to the reversal of the utterance between the utterance end time of the immediately preceding translation and the utterance start time of the subsequent translation. Thus, a speech synthesizer having a delay processing unit for delaying the utterance start time and the utterance end time of the later translated sentence is configured.
Here, Claim 6: The source language database 11 storing information on the utterance start time and utterance end time of each utterance sentence in the source language, the translated sentence of each uttered sentence in the source language, the utterance start time of the translated sentence, and the utterance end Using the target language database 21 storing the time information, the utterance end time of the source language utterance corresponding to each translated sentence in the target language database 21 is acquired from the source language database 11 and the utterance end time is set to The time obtained by adding the time to the translation is added to the translation as the utterance start time of the translation, and the speech synthesis method is performed to synthesize the speech in the target language database 21 from the obtained utterance start time. did.
[0008]
Claim 7: The source language database 11 storing information on the utterance start time and utterance end time of each utterance sentence in the source language, the translation of each utterance in the source language, the utterance start time and the utterance end time of the translated sentence Of the utterance of the source language corresponding to each translated sentence in the target language database 21 from the source language database 11 using the target language database 21 storing the information of the target language. Is added to the translated sentence as the utterance start time of the translated sentence, and the utterance time of each translated sentence in the target language database 21 is calculated. The utterance end time of the translation is calculated from the utterance start time, and the utterance start time of the subsequent translation corresponding to the reversal of the utterance end time of the immediately preceding translation and the utterance start time of the subsequent translation When Allowed delay talk end time, each translation in the target language database 21, to constitute a speech synthesis method of speech synthesis from speech start time previously obtained or delay caused to be utterance start time.
[0009]
Claim 8: The source language database 11 storing information of the utterance start time and utterance end time of each utterance sentence of the source language, the translation of each utterance sentence of the source language, the utterance start time and the utterance end time of the translated sentence Of the source language corresponding to each translation in the target language database 21 is acquired from the source language database 11, and the target language database 21 is stored in the target language database 21. Is added to the translated sentence as the utterance start time of the translated sentence, and the utterance time of each translated sentence in the target language database 21 is calculated. The utterance end time of the translation is calculated from the utterance start time, and the utterance start time of the subsequent translation corresponding to the reversal of the utterance end time of the immediately preceding translation and the utterance start time of the subsequent translation When The speech end time is delayed, and the length of a silent section between each utterance sentence in the target language database 21 is calculated. If the length exceeds a certain value, the utterance start for a later translated sentence is started. The time and the end time of the utterance are corrected earlier, and each translated sentence in the target language database 21 is adjusted to the utterance start time obtained earlier, the utterance start time delayed, or the utterance start corrected earlier. A speech synthesis method for synthesizing speech from time was constructed.
[0010]
Here, claim 9: a source language database 11 storing information on the utterance start time and utterance end time of each utterance sentence in the source language, a translation of each utterance in the source language, and an utterance start time and utterance of the translated sentence A certain time is added to the utterance end time of the source language utterance corresponding to each translation in the target language database 21 storing the information of the end time and the utterance start time of the translation. A process of calculating the utterance time of each translation in the target language database 21, a process of calculating the utterance end time of the translation based on the utterance time and the utterance start time assigned to the translation, A process of calculating the length of a silent section between each utterance in the target language database 21 based on the obtained utterance start time and utterance end time information of each translation; If the length of the silence section exceeds a certain value, the process of correcting the utterance start time and the utterance end time of the translated sentence earlier and correcting each translated sentence in the target language database 21. And a storage medium for storing a speech synthesis program comprising a speech synthesis process from the utterance start time.
[0011]
BEST MODE FOR CARRYING OUT THE INVENTION
An embodiment of the present invention will be described first with reference to FIGS.
In the embodiment of FIG. 1, 11 is a source language database, 21 is a target language database, 31 is an utterance start time giving unit, 32 is an utterance end time giving unit, 41 is a corrected utterance time giving unit, and 51 is a speech synthesis unit. . FIG. 2 is a diagram showing a relationship between utterance times between the source language and the target language. Hereinafter, a specific description will be given also with reference to FIGS.
[0012]
The source language database 11 stores information on each utterance sentence in the source language and its utterance start time and utterance end time. FIG. 3 is a diagram showing an example of the source language database 11, in which an utterance sentence in the source language and information on the utterance start time and utterance end time are stored. The utterance sentence in the source language may be text data or voice data.
The target language database 21 stores a translated sentence corresponding to each uttered sentence in the source language and information on the utterance start time and the utterance end time. FIG. 4 is a diagram showing an example of the target language database, in which a translated sentence and information on its utterance start time and utterance end time are stored. The translation may be text data or audio data. The target language database 21 stores (a) a translated sentence that has been subjected to correction processing such as a translation, in addition to (a) a translated sentence of a direct translation expression, and (c) a normal tone. The translated voice data, and (d) voice data translated in a tone reflecting the utterance content of the source language including emotions such as joy, sadness, and anger can also be stored.
[0013]
The utterance start time giving unit 31 acquires the utterance end time of the utterance sentence in the source language corresponding to each translation in the target language database 21 from the source language database 11, and adds a certain time to the utterance end time. Is given as the utterance start time of the translated sentence. In FIG. 4, a value obtained by adding a certain time, for example, 1 msec to the utterance end time of the utterance sentence in the source language corresponding to the utterance start time of each translated sentence is displayed.
[0014]
The utterance end time assigning unit 32 calculates the utterance time of each translated sentence in the target language database 21 and, based on the uttered time and the utterance start time obtained by the utterance start time assigning unit 31, ends the utterance of the translated sentence. The time is calculated and given. By the way, when the translated sentence is text data, the speech time of the translated sentence can be obtained by synthesizing the translated sentence by the speech synthesizer 51 and measuring the reproduction time of the speech. In the case of text data, the speech synthesizer 51 decomposes the text into phonemes, and estimates the utterance time without reproducing the speech using the phoneme length information stored in the speech synthesizer 51 in advance. It can also be obtained by calculating a value. When the translated sentence is voice data, information on the utterance time may be incorporated in advance. The utterance time can also be obtained by reproducing the voice data by the voice synthesis unit 51 and measuring the reproduction time. In FIG. 4, a value obtained by adding the utterance time calculated as described above to the utterance start time is shown as the utterance end time.
[0015]
Here, when giving the utterance end time by the utterance end time giving unit 32, the utterance end time of a certain translated sentence may be later than the utterance start time of the next translated sentence. In this case, there are a method in which the utterances are overlapped and the time is not specifically corrected, and a method in which the utterances are not overlapped and the time is corrected. In the following method, the start of the utterance of the later translated sentence is immediately after the end of the utterance of the preceding translated sentence. That is, the utterance end time giving unit 32 delays the utterance start time and the utterance end time of the next translated sentence in response to the reversal between the utterance end time of the immediately preceding translated sentence and the utterance start time of the subsequent translated sentence. And a delay processing unit.
[0016]
When N is the total number of sentences, the utterance time is corrected by the following routine.
Routine contents (1) i = 1.
(2) If i <N, proceed to (3). If i = N, the process ends.
(3) If the utterance start time of the (i + 1) th translated sentence is f and the utterance end time of the ith translated sentence is (f + e) (e> 0), the utterance start time of the (i + 1) th translated sentence And the utterance end time is delayed by e.
(4) Set i = i + 1. Proceed to (2).
[0017]
Referring to FIG. 2, based on the time information obtained by the utterance start time provision unit 31 and the utterance end time provision unit 32, the modified utterance time provision unit 41 performs silence between the translations in the target language database 21. The length of the section is calculated, and if this length exceeds a certain value, a value obtained by correcting the utterance start time and the utterance end time earlier is added to a later sentence. The solid line indicates the utterance section of each sentence, and the numbers i and (i + 1) attached to the solid lines indicate the corresponding source language sentence and target language sentence. m represents the time to be added to the utterance end time of the utterance sentence in the source language in the utterance start time providing unit 31. However, when the utterance start and end times of the target language sentence are corrected by the corrected utterance time providing unit 41, the value of m is not always constant.
[0018]
When N is the total number of sentences, the time correction in the corrected utterance time giving unit 41 is performed by the following routine.
Routine contents (1) i = 1.
(2) If i <N, proceed to (3). If i = N, the process ends.
(3) If the length of y obtained by subtracting the silence time x between the source language sentence i and (i + 1) from the silence time between the target language sentences i and (i + 1) is equal to or greater than a certain threshold k, the target language sentence (i + 1) The utterance start and end times are shifted forward by the same time so that the value of y becomes k. However, as a result, if the utterance start time of the target language sentence (i + 1) is earlier than the utterance start time of the source language sentence (i + 1), the utterance start time of the target language sentence (i + 1) becomes the source language sentence (i + 1). The utterance start and end times of the target language sentence (i + 1) are shifted forward by the same time so that the utterance start times are equal to the utterance start times.
(4) Set i = i + 1. Proceed to (2).
[0019]
As described above, based on the time information obtained by the utterance start time provision unit 31 and the utterance end time provision unit 32, the length of the silent section between each utterance sentence in the target language database 21 is calculated. Is greater than a certain value, the utterance start time and the utterance end time of the later sentence are provided with values that are corrected earlier, so that the sentence of the target language to be speech-synthesized is provided. It is possible to perform voice synthesis by setting a silence section between pauses to an appropriate pause time.
[0020]
FIG. 5 shows, in the target language database 21, the correction times given by the modified utterance time giving unit 41, separately from the utterance start and end times given by the utterance start time giving unit 31 and the utterance end time giving unit 32. ing. The fourth sentence, the Japanese sentence, takes a long time from the long object to the last verb being uttered. Accordingly, since the silent section between the third and fourth sentence of the English sentence as the target language becomes longer, the utterance start and end times of the fourth sentence are corrected forward. In this case, x = 5 seconds, y = 40 seconds, and k = 10 seconds.
[0021]
Here, the speech synthesis unit 51 will be described. The speech synthesis unit 51 synthesizes each translated sentence in the target language database 21 from the speech start time obtained by the speech start time giving unit 31 or the speech end time giving unit 32. Alternatively, a speech synthesis is performed from the utterance start time obtained by the corrected utterance time giving unit 41. The speech synthesis unit 51 may also have a configuration in which each speech sentence in the source language database 11 is speech-synthesized from its utterance start time, and the source language sentence and the target language sentence are simultaneously speech-synthesized.
[0022]
The above-described speech synthesis unit 51 synthesizes each translated sentence in the target language database 21 from the utterance start time obtained by the utterance start time giving unit 31 or the utterance end time giving unit 32, so that a sequentially translated speech is obtained. Can be If one sentence in the source language is long, the silence interval between the translated sentence and the immediately preceding translated sentence becomes longer depending on the sequential translation. It is considered a gradual translation that incorporates Therefore, by making each sentence in the target language database 21 speech-synthesized from the utterance start time obtained by the corrected utterance time giving unit 41, a progressively translated speech incorporating the predicted translation is obtained.
[0023]
By the way, in the operation of this embodiment, a storage medium storing a speech synthesis program is prepared, and although not shown, the speech synthesis program is installed from this storage medium by the CPU, and the source language database 11 and the target language are stored. This is performed with reference to the database 21. This speech synthesis program includes a source language database storing information on the utterance start time and utterance end time of each utterance sentence in the source language, a translation of each utterance sentence in the source language, and a translation A target language database storing information of the utterance start time and the utterance end time, and adding a certain time to the utterance end time of the utterance sentence of the source language corresponding to each translated sentence in the target language database, and uttering the translated sentence A process for obtaining a start time, a process for calculating the utterance time of each translation in the target language database, a process for obtaining the utterance end time of the translation based on the utterance time and the utterance start time assigned to the translation, A process of calculating the length of a silent section between each utterance in the target language database based on the utterance start time and utterance end time information of each translated sentence obtained If the length of the silence interval exceeds a certain value, the process of correcting the utterance start time and the utterance end time earlier for the later translated sentence, and correcting each translated sentence in the target language database. It is composed of a process of synthesizing voice from the utterance start time.
[0024]
Here, it is assumed that a certain utterance sentence in the source language sentence is "Why do you make such a gap?" As the corresponding translation, a translation of the following direct translation expression and a translation that has been subjected to a meaning translation and other correction processing are stored in the target language database 21.
Source language utterances: How did such a gap occur?
Translation of a literal translation: What is the reason of this gap?
Translated translations: How can we have this large gap between the foreigners and Japanese understanding?
The translated sentence subjected to the above-mentioned translation and other processing is translated in view of the context of the utterance contents up to that time. The speech synthesis unit 51 performs speech synthesis on the translated sentence of the direct translation expression and the translated sentence subjected to the meaning translation and other processes.
[0025]
【The invention's effect】
As described above, according to the present invention, by adopting the configuration for reproducing the sequential translation speech or the progressive translation speech, a more natural speech synthesis of the target language can be performed. Then, it is possible to verify a difference in discomfort between the sequentially translated speech and the progressively translated speech incorporating the predicted translation. In addition, it is possible to verify a difference in ease of content understanding between a translation speech of a direct translation expression and a translation speech of a linguistic expression. Further, the present invention can be used by a person who intends to be an interpreter as a learning material for simultaneous interpretation. It can also be used as a research tool by researchers in simultaneous translation and machine translation.
[Brief description of the drawings]
FIG. 1 illustrates an embodiment.
FIG. 2 is a diagram showing a mutual relationship between utterance times of a source language and a target language.
FIG. 3 is a diagram showing an example of a source language database.
FIG. 4 is a diagram showing an example of a target language database.
FIG. 5 is a diagram showing an assigned correction time.
[Explanation of symbols]
11 Source language database 21 Target language database 31 Utterance start time giving unit 32 Utterance end time giving unit 41 Modified utterance time giving unit 51 Voice synthesis unit

Claims (9)

A source language database storing information on the utterance start time and utterance end time of each utterance sentence in the source language;
A target language database storing translations of each utterance sentence in the source language and information on the utterance start time and utterance end time of the translated sentence;
The utterance end time of the utterance of the source language corresponding to each translation in the target language database is obtained from the source language database, and the time obtained by adding a certain time to the utterance end time is the utterance start time of the translation. An utterance start time giving unit that gives the translation as the time to the translation,
A speech synthesizer comprising: a speech synthesis unit that operates each translated sentence in a target language database in response to an utterance start time obtained by an utterance start time providing unit. A source language database storing information on the utterance start time and utterance end time of each utterance sentence in the source language;
A target language database storing translations of each utterance sentence in the source language and information on the utterance start time and utterance end time of the translated sentence;
The utterance end time of the utterance of the source language corresponding to each translation in the target language database is obtained from the source language database, and the time obtained by adding a certain time to the utterance end time is the utterance start time of the translation. An utterance start time giving unit that gives the translation as the time to the translation,
The utterance time of each translated sentence in the target language database is calculated, and the utterance end time of the translated sentence is calculated from the uttered time and the utterance start time obtained by the utterance start time assigning unit, and is attached to the translated sentence. An utterance end time providing unit,
A speech synthesizer comprising: a speech synthesis unit that operates each translated sentence in a target language database in response to an utterance start time obtained by an utterance start time giving unit or an utterance end time giving unit. The speech synthesizer according to claim 2,
The utterance end time assigning unit delays the utterance start time and utterance end time of the subsequent translation in response to the reversal between the utterance end time of the immediately preceding translation and the utterance start time of the subsequent translation. A speech synthesizer characterized by having: A source language database storing information on the utterance start time and utterance end time of each utterance sentence in the source language;
A target language database storing translations of each utterance sentence in the source language and information on the utterance start time and utterance end time of the translated sentence;
The utterance end time of the utterance of the source language corresponding to each translation in the target language database is obtained from the source language database, and the time obtained by adding a certain time to the utterance end time is the utterance start time of the translation. An utterance start time giving unit that gives the translation as the time to the translation,
The utterance time of each translated sentence in the target language database is calculated, and the utterance end time of the translated sentence is calculated from the uttered time and the utterance start time obtained by the utterance start time assigning unit, and is attached to the translated sentence. An utterance end time providing unit,
Based on the time information obtained by the utterance start time provision unit and the utterance end time provision unit, the length of a silent section between each utterance sentence in the target language database is calculated, and the length exceeds a certain value. In the case, there is provided a corrected utterance time providing unit that adds a value obtained by correcting the utterance start time and the utterance end time earlier to a later translated sentence,
A speech synthesis unit is provided which operates each translated sentence in the target language database in response to the utterance start time obtained by the utterance start time giving unit, the utterance end time giving unit or the corrected utterance time giving unit. Speech synthesizer. The speech synthesizer according to claim 4,
The utterance end time assigning unit delays the utterance start time and utterance end time of the subsequent translation in response to the reversal between the utterance end time of the immediately preceding translation and the utterance start time of the subsequent translation. A speech synthesizer characterized by having: A source language database storing information on the utterance start time and utterance end time of each utterance in the source language, and a translation of each utterance in the source language and information on the utterance start time and utterance end time of the translated sentence are stored. Using the target language database,
The utterance end time of the source language utterance corresponding to each translation in the target language database is obtained from the source language database, and a time obtained by adding a certain time to the utterance end time is set as the start of the utterance of the translation. Time is assigned to the translation as
A speech synthesis method characterized by performing speech synthesis from each utterance start time at which each translated sentence in a target language database is obtained. A source language database storing information on the utterance start time and utterance end time of each utterance in the source language, and a translation of each utterance in the source language and information on the utterance start time and utterance end time of the translated sentence are stored. Using the target language database,
The utterance end time of the source language utterance corresponding to each translation in the target language database is obtained from the source language database, and a time obtained by adding a certain time to the utterance end time is used as the utterance of the relevant translation. Assigned to the translation as a start time,
Calculate the utterance time of each translation in the target language database, calculate the utterance end time of the translation from the utterance time and the utterance start time obtained earlier,
The utterance start time and the utterance end time of the subsequent translation are delayed corresponding to the reversal of the utterance end time of the immediately preceding translation and the utterance start time of the subsequent translation,
A speech synthesis method comprising: synthesizing each translated sentence in a target language database from a speech start time obtained earlier or a speech start time delayed. A source language database storing information on the utterance start time and utterance end time of each utterance in the source language, and a translation of each utterance in the source language and information on the utterance start time and utterance end time of the translated sentence are stored. Using the target language database,
The utterance end time of the utterance of the source language corresponding to each translation in the target language database is obtained from the source language database, and the time obtained by adding a certain time to the utterance end time is the utterance start time of the translation. Time is assigned to the translation as
Calculate the utterance time of each translation in the target language database, calculate the utterance end time of the translation from the utterance time and the utterance start time obtained earlier,
The utterance start time and the utterance end time of the subsequent translation are delayed corresponding to the reversal of the utterance end time of the immediately preceding translation and the utterance start time of the subsequent translation,
Calculate the length of the silent section between each utterance in the target language database, and if the length exceeds a certain value, set the utterance start time and utterance end Modified to
A speech synthesis method comprising: synthesizing each translated sentence in a target language database from a speech start time obtained earlier, a speech start time delayed, or a speech start time corrected earlier. . A source language database storing information on the utterance start time and utterance end time of each utterance sentence in the source language;
A target language database storing translations of each utterance in the source language and information on the utterance start time and utterance end time of the translation;
A process of adding a certain time to the utterance end time of the utterance sentence of the source language corresponding to each translation in the target language database to obtain an utterance start time of the translation;
A process of calculating the utterance time of each translated sentence in the target language database;
A process of obtaining the utterance end time of the translation based on the utterance time and the utterance start time given to the translation;
A process of calculating a length of a silent section between each utterance in the target language database based on the obtained utterance start time and utterance end time information of each translation;
A process of correcting the utterance start time and the utterance end time earlier for the later translated sentence if the length of the calculated silent section exceeds a certain value;
A storage medium storing a speech synthesis program, characterized by comprising a process of synthesizing a speech from a corrected utterance start time of each translated sentence in a target language database.

Images