JPH01204100A - Text speech synthesis system - Google Patents

Abstract

PURPOSE:To make a synthetic speech easy to hear by give variation to the rhythm of the synthetic speech for voicing a character string with emphasized information when the emphasized information is included in an input text. CONSTITUTION:Morpheme information is extracted from the input text by a morpheme analytic part 1 and a phoneme and rhythm symbol generation part 1 finds a phoneme symbol sequence and a rhythm symbol sequence according to the information. Further, when the input text contains the emphasized information, an emphasized symbol generation part 5 generates an emphasized symbol according to the information. Then emphasized rhythm generated an emphasized rhythm generation part 6 is added by an adder 8 to the rhythm generated by the rhythm generation part 3. Thus, a speech is synthesized with the rhythm to which the emphasized rhythm is added and the monotoneousness of the synthetic tone is reduced to offer important information in an easy-to-hear state for a listener.

Description

[Detailed description of the invention] Technical field The present invention relates to a text-to-speech synthesis method.

Prior Art When a long sentence is synthesized using text-to-speech synthesis, the synthesized speech is very monotonous compared to human speech, and listening to it for a long time is painful.

In addition to text information, written text contains text information.

It usually includes various emphasis information such as sidelines, dots, angle brackets, bold type, enlarged text, shading, modified fonts, etc.
This provides important information in an easy-to-understand format for readers. However, with conventional text-to-speech synthesis methods,
This emphasized information was ignored and only character information was used as input information. For this reason, there is no distinction between emphasized parts and non-emphasized parts in the output synthesized speech, and one synthesized speech becomes even more monotonous.

Purpose The present invention was made in view of the above-mentioned circumstances.
In particular, in text synthesis of speech, the aim is to make the synthesized speech, which tends to be monotonous, easier to listen to by emphasizing the important information contained therein.

Structure In order to achieve the above object, the present invention extracts morphological information from an input text by morphological analysis processing, and based on this extracts phonological symbol strings according to phonological rules and prosodic symbol strings according to prosodic symbol derivation rules, and prepares them in advance. In a text-to-speech synthesizer that reads out parameter sequences of speech segments according to phonetic symbol strings, connects them according to connection rules, and adds prosody from prosodic symbol strings according to prosodic rules, it is necessary to use a text-to-speech synthesizer that reads out parameter sequences of speech segments according to phonetic symbol strings, connects them according to connection rules, and adds prosody from prosodic symbol strings according to prosodic rules. When the character string emphasized by the emphasis information is uttered, the prosody of the synthesized speech is changed. Hereinafter, the present invention will be explained based on examples.

FIG. 1 is a block diagram for explaining one embodiment of the present invention, in which 1 is a morphological analysis section, 2 is a phonological and prosodic symbol generation section, 3 is a prosody generation section, 4 is a prosody table, and 5 is a block diagram for explaining an embodiment of the present invention. 1 is an emphasis symbol generation unit, 6 is an emphasis prosody generation unit, 7 is an emphasis bias table, 8 is an adder, 9 is a phonological parameter generation unit, 10 is a speech segment file, and 11 is a speech synthesizer. Morphological information is extracted from the text by morphological analysis processing, and based on this, phonological symbol strings are determined using phonological rules and prosodic symbol strings are obtained using prosodic symbol derivation rules.

In a text-to-speech synthesizer that reads out a parameter sequence of speech segments prepared in advance according to a phonetic symbol string, connects them according to a combination rule, and adds prosody using a prosodic symbol string according to a prosodic rule, it is possible to detect some emphasis information in the input text. If included, the prosody of the synthesized speech is changed when the character string emphasized by the emphasis information is uttered. For example, it changes the prosody of the synthesized voice.
Bold.

By changing the prosody of one emphasized part of a sentence that includes various emphasis information such as enlarged letters, shading, and modified fonts, the monotony of synthesized speech is reduced and important information is easier to understand for the listener. It is provided in the form of

Figures 2 to 4 are diagrams for explaining the present invention in detail, and each shows an example of uttering "This is the meat called 'marbled'" and "marbled". In the embodiment shown in FIG. 2, the fundamental frequency of the part emphasized by the emphasis information is given a certain bias with respect to that of the non-emphasized part. It was made more expensive. In addition, the embodiment shown in FIG.
The embodiment shown in FIG. 4, which increases the power of a part emphasized by emphasis information with a certain bias compared to that of a non-emphasis part, increases the power of a part emphasized by emphasis information. It is slower than that of the emphasized part, that is, the duration length of each phoneme of the emphasized part is longer than that of the non-emphasized part.

Effects As is clear from the above explanation, (1) according to the present invention, the monotony of synthesized speech can be reduced, (1) important information can be emphasized and uttered, and in particular, monotony can be reduced in speech-to-text synthesis. Synthesized speech can be made easier to listen to by emphasizing the important information contained therein.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an example of a text-to-speech synthesis device used to implement the present invention, and FIGS. 2 to 4 are diagrams for explaining the present invention in detail. l...Morphological analysis unit, 2...Phonological and prosodic symbol generation unit,
3... Prosodic generation unit, 4... Prosodic table, 5...
Emphasis symbol generation unit, 6: Emphasis prosody generation unit, 7: Emphasis bias table, 8: Adder, 9: Phonological parameter generation unit, 10: Speech element file, 11.
...Speech synthesizer. Figure 2 Figure 6

Claims (1)

[Claims] 1. Extract morphological information from the input text by morphological analysis processing, and based on this, obtain a phonetic symbol string using phonological rules and a prosodic symbol string using prosodic symbol derivation rules. In a text-to-speech synthesizer that reads the text according to the rules, connects it according to the combination rules, and adds prosody using the prosodic rules from the prosodic symbol string, if the input text contains some emphasis information, it is emphasized by the emphasis information. A text-to-speech synthesis method characterized by varying the prosody of synthesized speech when uttering a character string.