JPH086586A - Accent imparting device and speech synthesizer - Google Patents

Abstract

(57) [Abstract] [Purpose] To provide an accent imparting device and a voice synthesizing device capable of coping with a style including abbreviations of dialogues, headings and the like, and capable of highly recognizable voice synthesis. [Structure] A style setting unit 1e that sets a style of a text input from a text input terminal 0, a morpheme processing unit 1a that divides the input text into morphemes, and a syntax analysis that performs syntax analysis based on the divided morphemes. The section 1c and the reading and accent of the analyzed text are written in the style setting section 1
The reading / accent processing unit 1d is provided according to the style set by e.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a language processing device used for synthesizing a sentence or the like into a voice, and more particularly to an accent imparting device and a voice synthesizing device.

[0002]

2. Description of the Related Art As a conventional speech rule synthesizer, for example, Furui: Digital Speech Processing p.146 (Tokai University Press)
The device shown in 1985) is known. FIG.
It is a block diagram which shows the structure of this conventional speech synthesizer. In FIG. 4, a kanji / kana mixed sentence is input to the character string input terminal 0. In the morpheme processing unit 1a, the input kanji / kana mixed sentence is divided into morphemes using the dictionary 1b,
The reading, accent type, part of speech, etc. of each morpheme are added.
The syntactic analysis unit 1c uses the information of each morpheme obtained by the morpheme processing unit 1a to determine a phrase and performs a dependency analysis between the phrases. The reading / accent processing unit 1d determines accent phrases, accent positions, poses and intonation repositioning positions, and reading. The morpheme processing unit 1a, the dictionary 1b, the syntax analysis unit 1c, and the reading / accent processing unit 1d constitute the language processing unit 1.

The acoustic processing unit 2 creates a synthesis parameter based on the reading and accent information obtained by the language processing unit 1. The synthesis parameters are an amplitude that determines the volume of a voice and a vocal tract description parameter (P that determines the state of the vocal tract).
(ARCOR coefficient, formant frequency, etc.), voiced / unvoiced determination flag that determines the state of the vocal cords, and basic frequency that determines the pitch of the voice. The synthesis processing unit 3 converts the synthesis parameter sequence obtained by the acoustic processing unit 2 into a voice waveform, and obtains a voice waveform at the synthesized voice output terminal 4.

[0004]

Such a speech rule synthesizing device is being used in various fields such as reading WP sentences and public guide broadcasting. From the viewpoint of intelligibility that individual syllables of synthesized speech can be understood, it is approaching a practical level. However, accent punctuation and position are decided uniformly, and no consideration is given to the type of style, and it is relatively possible to read aloud editorials, but omission of attached words such as dialogues and headings. There is a problem that the style including "is not supported".

In consideration of such a problem of the conventional speech synthesizer, the present invention can cope with a style including an abbreviation of adjectives such as dialogues and headings, and an accent imparting apparatus capable of highly recognizable speech synthesis. It is an object of the present invention to provide a speech synthesizer.

[0006]

According to the present invention of claim 1, text input means for inputting text, style setting means for setting the style of the input text, and the input text are divided into morphemes. The accent imparting device is provided with a morpheme analyzing means and a reading / accent imparting means for imparting a reading and an accent of a text divided into the morphemes according to a set style.

According to a second aspect of the present invention, a text input means for inputting text, a style determination means for determining the style of the input text, a morpheme analysis means for dividing the input text into morphemes, and the same. An accent imparting device comprising: a reading / accent imparting means for imparting a reading or an accent of a text divided into morphemes according to a determined style.

According to a fifth aspect of the present invention, a text input means for inputting text, a style determination means for determining the style of the input text, a morpheme analysis means for dividing the input text into morphemes, and a determination are made. According to the written style, the accent imparting device is provided with a sentence shaping means for adding an adjunct word necessary for the text and a reading / accent imparting means for imparting a reading or an accent to the text to which the adjunct word is added.

According to a sixth aspect of the present invention, an accent imparting device according to any one of the first to fifth aspects, and a voice synthesizing means for synthesizing a text to which a reading or an accent is imparted by the accent imparting device into a voice. It is a speech synthesizer equipped with.

[0010]

According to the present invention of claim 1, the style setting means sets the style of the text input by the text input means,
The morpheme analysis means divides the input text into morphemes, and the reading / accenting means adds the reading and accent of the morpheme-divided text according to the set style.

According to the second aspect of the present invention, the style determination unit determines the style of the text input by the text input unit, the morpheme analysis unit divides the input text into morphemes, and the reading / accenting unit. Adds the reading and accent of the text divided into morphemes according to the determined style.

According to the fifth aspect of the present invention, the style determination unit determines the style of the text input by the text input unit, the morpheme analysis unit divides the input text into morphemes, and the sentence shaping unit, In accordance with the determined style, a necessary accessory word is added to the text, and the reading / accenting means adds a reading or an accent to the text to which the accessory word is added.

According to the sixth aspect of the present invention, the voice synthesizing means synthesizes the text to which the reading or accent is added by the accent adding device according to any one of the first to fifth aspects.

[0014]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below with reference to the drawings showing its embodiments. (Embodiment 1) FIG. 1 is a block diagram showing the configuration of a speech synthesizer according to a first embodiment of the present invention. That is, the speech synthesizing device is based on the output data of the language processing unit 1 and the language processing unit 1 that determines the reading, etc., so that the Kanji / Kana mixed sentence that is the text input from the character string input terminal 0 can be speech-synthesized. In addition, the sound processing unit 2 that creates a synthesis parameter and the synthesis processing unit 3 that converts the created synthesis parameter into voice.

The language processing unit 1 includes a morpheme processing unit 1a for performing a morpheme analysis of an input text, a dictionary 1b used for the morpheme analysis, a syntactic analysis unit 1c for performing a syntactic analysis based on the processing result of the morpheme processing unit 1a, and an input. A style setting unit 1e for setting the style of text, according to the set style,
It is composed of a reading / accent processing unit 1d for determining reading / accent and the like for the analysis result of the syntax analysis unit 1c.

Of the above, the character string input terminal 0 is text input means, the style setting unit 1e is style setting means,
The morpheme processing unit 1a and the syntax analysis unit 1c constitute a morpheme analysis unit, the reading / accent processing unit 1d constitutes a reading / accenting unit, and the sound processing unit 2 and the synthesis processing unit 3 constitute a speech synthesizing unit. are doing.

Next, the operation of the speech synthesizer of the first embodiment will be described with reference to the drawings.

First, the kana-kana mixed sentence input from the character string input terminal 0 is processed by the morphological processing unit 1a into the dictionary 1b.
Is used to divide into morphemes, and the pronunciation, accent type, part of speech, etc. of each morpheme are added. Next, the syntactic analysis unit 1c determines a phrase using the information of each morpheme obtained by the morpheme processing unit 1a, and performs a dependency analysis between the phrases.

On the other hand, the style of the input text is set from the style setting terminal 5, and the reading / accent processing unit 1d determines the accent phrase and the accent position by using the result of the syntax analysis according to the set style. , The pose and the intonation upright position, and the reading are determined.

Next, the sound processing unit 2 creates a synthesis parameter based on the reading and accent information obtained by the language processing unit 1. This synthesis parameter is an amplitude that determines the volume of the voice, a vocal tract description parameter that determines the state of the vocal tract (PARCOR coefficient, formant frequency, etc.), a voiced / unvoiced determination flag that determines the state of the vocal cord, and a basic that determines the pitch of the voice. Frequency etc. Furthermore, the synthesis processing unit 3
Converts the synthesis parameter sequence obtained by the sound processing unit 2 into a voice waveform, and outputs the voice waveform from the synthesized voice output terminal 4.
Here, in the present embodiment, the processing methods of the sound processing unit 2 and the synthesis processing unit 3 are not particularly limited.

Next, the details of each process will be described using an actual example. Now, let us consider a case where the sentence "new VTR release" is entered. This is a sentence that "the new VTR is released (was done)" is omitted from the part of the adjunct word (the ending "suru" of the Sahen noun "release" is also treated as an adjunct word). " It is an expression often used in written words such as newspaper headlines and examples. In addition to this, there are also adjective abbreviations that are often used in dialogues, such as "I will go."

The style setting unit 1e sets whether the input sentence is a "general sentence" or an "adjunct abbreviation". When "adjunct abbreviation" is set, the reading / accent processing is performed. In the part 1d, the phrase at the end of the phrase is divided.

The input sentence is morpheme-divided by the morpheme processing unit 1a as follows to obtain accent and reading information. The accent type of each word indicates a syllable position with an accent. Furthermore, by the syntax analysis unit 1c,
Dependency between clauses (distance between clauses) is determined, but in this example, since the adjunct word is omitted, it is treated as one clause. The reading / accent processing unit 1d determines accent phrases and accent positions, and determines reading. At the time of setting a general style (conventional method), this example is treated as one phrase, and it becomes an accent combination of nouns.
An accent is set for the syllable “ha”. On the other hand, when the abbreviated style is set, when the final morpheme of the phrase is a sahen noun, the accent phrase is set immediately before it. As a result, this example sentence is determined to be a two-accent phrase, and the accent is added as follows.

(Input sentence) “New VTR release” (morpheme division) New / VTR / release.

(Reading) Synthetic Earl Hatsui (Accent)-50 (Part of speech) Prefix Noun Sahen noun ---- When setting general style ---- (Section) New VTR released.

(Accent phrase) New VTR released.

(Language processing unit output) shi nn bu i tii ia
a ru ha1 tsu ba i ---- When setting abbreviated style ---- (section) New VTR / release.

(Accent phrase) New VTR / release.

(Language processing unit output) shi nn bu i tii ia
1 a ru / ha tsu ba i The number [1] after the romaji syllable indicates an accented syllable, and [/] indicates accent phrase segmentation.

As described above, according to the present embodiment, it is possible to give a correct accent by setting in advance that the input sentence is a sentence in which the attached word is omitted. (Embodiment 2) FIG. 2 is a block diagram showing the configuration of a speech synthesizer according to Embodiment 2 of the present invention. The same elements as those of the first embodiment are designated by the same reference numerals. In the present embodiment, instead of the style setting terminal 5 and the style setting section 1e, a style determining section 1f, which is a style determining means for automatically determining the style using information of a morphological analysis result or a syntactic analysis result, is provided. ing.

The style determining unit 1f of this embodiment determines the style of the input sentence by the following method.

Independent word + Sahen noun + Punctuation (bunsetsu punctuation) → adjunct abbreviation style For example, in "new VTR released." Become.

The operation after the style determination unit 1f determines the style is the same as that in the first embodiment, and the description thereof is omitted.

As described above, according to the present embodiment, a correct accent can be given by determining the style of the input sentence from the very simple information of the morpheme and the punctuation mark and giving the accent.

(Embodiment 3) Regarding the configuration of this embodiment,
This is the same as the second embodiment shown in FIG. 2, but the style determination method by the style determination unit 1f is different. The style determination unit 1 of this embodiment
In f, the style of the input sentence is determined by the following method.

The dictionary 1b used for morphological analysis is provided with information regarding case analysis. For example, a verb describes an intransitive verb / transitive verb, and the possibility of taking "wa case" and "de case". In the noun, information about whether or not the person can be "wa" and "wa" is described. Hereinafter, an example will be described. "New / VTR / Release" VTR: Noun- "wakaku" Release: Sahen noun (action) -takes "wakaku" Therefore, this example sentence is judged as "wakaku abbreviation sentence" and the accent phrase is divided. It "Limited release" Limited: Sahen noun (action) -does not become "wa case",-takes "wa case" Release: Sahen noun (action) -takes "wa case" Therefore, in this example, Since there is no case, it is not judged as an adjunct abbreviation, it is processed as one accent phrase, and "ge nn te ih
a1 tsu ba i ”.

The operation after the style judgment by the style judgment unit 1f is the same as that in the first embodiment, and therefore its explanation is omitted.

As described above, by holding the information for case analysis in the dictionary and using the case analysis information to judge the style of the input sentence and give the accent, the correct accent can be given, It is possible to provide an intelligible rule-synthesized sound. (Embodiment 4) FIG. 3 is a block diagram showing the configuration of a speech synthesizer according to a fourth embodiment of the present invention. The same elements as those in the second embodiment are designated by the same numbers. In the present embodiment, a sentence shaping unit 1g, which is a sentence shaping unit that generates and gives an abbreviated auxiliary word by using the determination result by the style determination unit 1f, is provided. The sentence shaping unit 1g supplements the omitted auxiliary words to generate a general sentence according to the case relationship obtained by the style determination unit 1f. Further, since the reading / accent processing unit 1d only needs to perform speech synthesis on a general sentence to which an attached word is added, it is sufficient that the reading / accent processing for the general sentence can be performed. "New / VTR / Release" VTR: Noun- "wakaku" Release: Sahen noun (action) -takes "wakaku" Therefore, this example sentence is judged as "wakaku abbreviated sentence" and "wa" is supplemented Then, it is formed into the sentence "Sale new VTR". "Limited Release" Limited: Sahen noun (action) -does not become "wa case" -takes "wa case" Release: sahen noun (action) takes "wa case" Therefore, this sentence is "abbreviated" It is judged that it is not a sentence ”and is not changed.

If the sentence is an abbreviated sentence, an adjunct word is added, and if the sentence is not an abbreviated sentence, the reading / accent processing unit 1d reads and accents the input sentence.
After that, voice synthesis is performed as in the first embodiment.

As described above, according to this embodiment, the supplementary words omitted according to the information obtained by the case analysis are supplemented to clarify the contents, and then the speech is converted into a voice. A synthetic sound can be provided.

As described above, by setting in advance that the input sentence is a sentence in which the attached words are omitted, a correct accent can be given. Also,
It is possible to provide a correct accent and provide a highly intelligible rule-synthesized sound by automatically determining the style of a sentence from simple information such as part of speech and punctuation marks or case-related information and imparting an accent. Alternatively, it is possible to provide a highly intelligible rule-synthesized voice by estimating an abbreviated word omitted by the case relation analysis, forming a sentence supplementing the affixed word, and converting the sentence into synthetic speech.

In each of the above-described embodiments, each processing unit is configured by dedicated hardware, but instead of this, the same function may be realized by software using a computer.

In the fourth embodiment, the omitted adjectives are added after the syntactic analysis. However, instead of this, the attached sentences may be added to the input sentence.

[0044]

As is apparent from the above description, the present invention provides a style setting means for setting the style of input text, a morpheme analysis means for dividing the input text into morphemes, and a morpheme division thereof. Yomi, which gives the reading and accent of the specified text according to the set style.
Since it is provided with an accent imparting means, it has an advantage that it can correspond to a style of sentence including omission of attached words such as dialogues and headings, and can perform highly intelligible speech synthesis.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a speech synthesizer according to a first exemplary embodiment of the present invention.

FIG. 2 is a block diagram showing a configuration of a speech synthesizer according to a second exemplary embodiment of the present invention.

FIG. 3 is a block diagram showing a configuration of a speech synthesizer according to a fourth exemplary embodiment of the present invention.

FIG. 4 is a block diagram showing a configuration of a conventional speech synthesizer.

[Explanation of symbols]

 1 Language Processing Section 1a Morphological Processing Section 1b Dictionary 1c Syntax Analysis Section 1d Reading / Accent Processing Section 1e Stylistic Setting Section 1f Stylistic Judgment Section 1g Text Forming Section 2 Acoustic Processing Section 3 Synthesis Processing Section 5 Stylistic Setting Section

Claims (6)

[Claims] 1. A text input means for inputting text, a style setting means for setting a style of the input text, a morpheme analysis means for dividing the input text into morphemes, and a morpheme An accent imparting device comprising: a reading / accent imparting means for imparting text reading and accent according to the set style. 2. A text input means for inputting text, a style determination means for determining the style of the input text, a morpheme analysis means for dividing the input text into morphemes, and a morpheme divided into the morphemes. An accent imparting device comprising: a reading and accent imparting means for imparting text reading and accent according to the determined style. 3. The accent applying device according to claim 2, wherein the style determining unit determines the style using the phrase segmentation information of the text and the part-of-speech information of the morpheme immediately preceding the phrase segmentation information. 4. The accent applying device according to claim 2, wherein the style determining unit determines the style using the structure of the input text. 5. A text input means for inputting text, a style determination means for determining the style of the input text, a morphological analysis means for dividing the input text into morphemes, and a determined style. An accent imparting device comprising: a sentence shaping means for adding an adjunct word necessary for the text; and a reading / accent imparting means for imparting a reading or an accent to the text to which the adjunct word is added. 6. The accent imparting device according to claim 1, and a voice synthesizing unit for synthesizing a text to which a reading or an accent is imparted by the accent imparting device into a voice. Characteristic speech synthesizer.