JPH0282300A - Device for reading sentence aloud - Google Patents

Abstract

PURPOSE:To prevent error in reading out of numerals by previously imparting the symbols to indicate the kinds of the way how the numerals are to be read out if the numerals in sentences are desired to be specifically read out or if the way of reading is mistaken when speech synthesis is once executed. CONSTITUTION:The symbols to indicate the kinds of the way how the numerals are to be read out, for example, the symbols to instruct reading with units or straight reading without units are previously inputted if the numerals in sentences are desired to be specifically read out or if the way of reading is mistaken with the sentences synthesized by once executing language processing. Whether the instruction symbol is imparted to the numeral or not is confirmed by a signal processing part 5-1 for the kinds of the reading way of the numerals in the case of making sentence analysis in a language processing part 5. The kind of the reading way is set in accordance with the instruction and the speech output is executed if the instruction symbol is imparted to the numeral. The correct reading out of the numerals is executed in this way.

Description

[Detailed description of the invention] [Summary] Regarding a text reading device that inputs a sentence mixed with Kanji and Kana and performs speech synthesis, when a number word is present in the sentence mixed with Kanji and Kana, especially when there are multiple readings for a number word with the same notation. If possible,
With the aim of solving the problem that errors occur easily in the way the number words are read and unnaturalness occurs in the output synthesized speech, we have added a symbol that specifies the type of reading of the number words as auxiliary information in the input sentences containing kanji and kana. and a means for inputting the number, and when outputting audio, set the specified number word reading type for the part where the reading type is specified,
and means for synthesizing and outputting audio based on the settings.

[Industrial Field of Application] The present invention relates to a text reading device that inputs a sentence containing kanji and kana and synthesizes speech, and particularly relates to a text reading device that reads out number words in a sentence without error.

[Prior Art] An apparatus for inputting a sentence containing kanji and kana and synthesizing speech generally has a system configuration as shown in FIG.

FIG. 4 shows a block configuration diagram of a conventional text reading device, in which 51 is an input device, 52 is a main control section, 53 is an input control section, 54 is a text storage section, 55 is a language processing section, 5
6 represents a sound processing section, and 57 represents a speaker.

In the same figure, a sentence containing kanji and kana inputted from an input device 51 is stored in a sentence storage section 54, and is analyzed in a language processing section 55 using phonetic symbols and prosodic symbols (accent,
pose symbols, etc.).

After converting these phonetic symbols and prosodic symbols into audio parameters, the audio processing unit 56 synthesizes audio and outputs it from the speaker 57.

[Problems to be Solved by the Invention] In a text reading device, it is desired that the original characters be pronounced correctly and the meaning of the text can be accurately determined using the output audio.

However, in reality, when numerals exist in a sentence containing kanji and kana, it is quite difficult to assign correct pronunciation symbols and metrical symbols to them.

In particular, when the same number word can be read in multiple ways, for example, for the number word 211, "Zip code is 21".
It's 1. ”, it is read as “ni-iichii°chi” (゛ represents an accent), but when saying, “The answer is 211,” it is read as “nihyaku juichi” with a digit. , even if the notation is the same, there is no fixed way to read it.

In such cases, until now, once a certain reading method (for example, stick reading) was set at the initial stage of the system, the text would be read out in that reading method (for example, stick reading) from then on. I was unable to read the text aloud.

In addition, in order to obtain correct audio without errors, there was a method of analyzing the meaning of the sentences before and after the numeral part in the input sentence and then reading the sentence out loud, but this would complicate the system configuration. There were drawbacks.

The present invention has been made in view of the above-mentioned problems, and an object of the present invention is to provide a text reading device that can correctly read out numerical words with little effort when inputting a text.

[Means for Solving the Problems] According to the present invention, the above objects are achieved by the means described in the claims.

That is, the present invention provides a text-to-speech device that inputs a sentence containing kanji and kana, converts it into a string of phonetic symbols and prosodic symbols, and synthesizes speech.
a means for inputting a symbol specifying the type of reading of a number word as auxiliary information; and a means for inputting a symbol specifying the type of reading of a number word when outputting audio, setting the type of reading of the number word assumed for the part where the type of reading is specified;
The present invention is a text reading device that includes means for synthesizing and outputting audio based on the settings.

[For production]

When there are number words in a sentence containing kanji and kana, especially when the same number word can be read out in multiple ways, errors are likely to occur when reading out the number word. For,
When you want to read a specific reading, or if you misread a number word in a sentence synthesized by -degree error processing, you can add a symbol indicating the type of reading, such as a reading with a digit, to the number word. , enter symbols to indicate stick reading, etc.
When analyzing a sentence in the language processing unit, it is checked whether or not the above instruction symbol is attached to the number word, and if the instruction symbol is attached, the reading type is set based on the instruction. to output audio.

[Embodiment] Fig. 1 is a block configuration diagram of a text reading device showing an embodiment of the present invention, in which 1 is an input device and a display device (hereinafter simply referred to as "input/output display device"), and 2 is a main control unit. , 3 is a human output control section, 4 is a text storage section, 5 is a language processing section, 5-1 is a number word reading type symbol processing section in the language processing section, 6 is a sound processing section, and 7 is a speaker. In the figure, input/output display device 1
The sentences inputted from the kanji/kana mixed sentences are sent to the sentence storage section 4 by the human output control section 3 and the main control section 2, and are stored in the sentence storage section 4.

The language processing unit 5 reads the text from the text storage unit 4, divides it into analysis units, performs language analysis for each analysis unit, and converts the language into information on phonetic symbols and prosodic symbols.

The language processing unit 5 also includes a number word reading type symbol processing unit 5-
1, check whether a number word reading type symbol is attached to the number words in the sentence, set the specified number word reading for the number words with the above symbol, and set the specified number word reading. In the case of a number word that is not written, the general number word processing inferred from the notation is performed.

The acoustic processing section 6 sets various parameters for outputting synthesized speech based on the phonetic symbols and prosodic symbols set in the language processing section 5, and the synthesized speech is output from the speaker 7.

Note that the main control section 2 centrally controls the operation of the entire apparatus.

FIG. 2 is a flowchart showing an example of inputting a number word reading type symbol according to the present invention, and shows the procedure for inserting a number word reading type symbol into a sentence.

The sentences (including numerals) retrieved from the sentence storage section 4 are subjected to language processing in step 10 and converted into a string of phonetic symbols and prosodic symbols (accents, pause marks, etc.), and then subjected to acoustic processing in step 11. is performed to obtain various voice parameters for voice synthesis, and in step 12, synthesized voice is output.

In step 13, it is determined whether or not the sentence contains a numeral word. If the sentence does not contain a numeral word, the process moves to step 14, and a check is made to see if the sentence to be speech synthesized has finished. Ru.

In step 13, if the sentence contains number words, proceed to step 15, and if you want the number words to be read in a specific way, or if you try -sentence speech synthesis and the sentence is read incorrectly, Using the input/output display device 1, in step 16, a symbol indicating the type of reading of a numeral is input into a sentence containing kanji and kana.

Furthermore, when the text into which the reading type symbol is inserted is to be synthesized into speech, the procedure shown in FIG. 3 is used.

That is, in the language processing step indicated by 20, it is determined in step 21 whether the word is a numeral, and if it is not a numeral, the process moves to step 22, where a phonetic symbol and a prosodic symbol are added.

If the word is a numeral, the process moves to step 23, and it is determined whether or not the numeral is given a numeral reading type symbol.

If no number word reading type symbol is given, the process moves to step 24, where the type of reading inferred from the notation is set, and in step 25, phonetic symbols and prosodic symbols are set.

If it is determined in step 23 that a number word reading type symbol is assigned to the number word, the process moves to step 26, where the reading type specified by the number word reading type code is set, and the process moves to step 25. do.

In step 26, sound processing is performed, in step 27, synthesized speech is output, and in step 28, it is confirmed that the text to be output as audio is finished.

As explained above, in the present invention, for number words in sentences,
For those to which a number word reading type symbol is attached, synthesized speech is output based on the type specified by the symbol,
For other objects, the pronunciation type inferred from the notation is set and a synthesized speech is output.

By the way, there are no particular restrictions on what kind of symbol to use as this number word reading type symbol, and various symbols can be used. For example, when using stick reading, the symbol <B)
An example using symbols such as (K> for digit reading is shown below.

In other words, when you say "The postal code is <B) 211", specify the reading such as "Niebin Bangou Honey Yew Des", and say "The postal code is <K) 211".
is. ”, then “Nibin bangou ha”
It reads something like, ``Niihyakujiniiichi desu''.

〔Effect of the invention〕

As explained above, according to the present invention, if you want to read a numeral word in a sentence containing kanji and kana in a specific way, or if you make a mistake in the reading when performing -sentence speech synthesis, the numeral word can be read in advance. On the other hand, by adding symbols that indicate the type of reading of number words, there will be no errors when reading number words.
With a little effort, the naturalness of synthesized speech can be greatly improved.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of a text reading device according to an embodiment of the present invention. FIG. 2 is a flowchart showing an example of inputting a number word reading type symbol according to the present invention. FIG. 3 is a block diagram of a sentence reading device according to an embodiment of the present invention. FIG. 4 is a flowchart showing the procedure for speech synthesis of a text into which a type symbol has been inserted. FIG. 4 is a block diagram of a conventional text reading device. DESCRIPTION OF SYMBOLS 1...Input/output display device, 2...Main control section, 3...Human output control section, 4...Text storage section, 5... ...Language processing section, 5-1...
・Number reading type symbol processing unit in the Kogo processing unit, 6...
...Sound processing unit, 7...Speaker, 10-
16.20-28... each processing step on the flowchart

Claims (1)

[Claims] In a text reading device that inputs a sentence containing kanji and kana, converts it into a string of phonetic symbols and prosodic symbols, and synthesizes speech, the type of pronunciation of a number word is provided as auxiliary information in the input sentence containing kanji and kana. A means for inputting the specified symbol, and when outputting audio, set the specified number word reading type for the part where the above reading type is specified,
A text reading device comprising: means for synthesizing and outputting audio based on the settings.