JPS62208125A - Sentence reading device - Google Patents

Abstract

PURPOSE:To attain reading to be easily listened by inserting silent codes into a sentence reading string in accordance with previously set conditions. CONSTITUTION:When a sentence consisting of a character string is inputted from a document input part 11, a sentence analyzing part 12 analyzes the sentence while referring a reading dictionary 13 to execute word identification processing 121. Then, silent insertion condition deciding processing 122 is executed in accordance with previously set condition such as the continuation of the same vowel on a border of words or the continuation of specified word groups such as numerals and reading string forming processing 123 for a reading string into which a necessary silent part is inserted is executed. The reading string processed by said processing is outputted as voice through a reading string storing part 14, a voice output part 15, and so on. Thus, sentence reading to be easily listened can be attained by inserting required silent parts.

Description

[Detailed Description of the Invention] [Summary] In the text reading voice output from the text reading device,
When words are connected by vowels.

Since words may coalesce and become difficult to distinguish, it is possible to insert silent intervals between words as necessary.

[Industrial application field]

The present invention relates to a text reading device that inputs a sentence written in binary notation and outputs it aloud, and particularly relates to a reading string generation method that inserts silent intervals to clarify word divisions.

[Conventional technology]

Text-to-speech devices that input sentences written in the form of ordinary Japanese sentences and output the correct pronunciation aloud have recently been attracting attention as useful for many purposes such as checking and proofreading a single document. .

In Japanese sentences, a single kanji often has multiple readings, such as onkun, and the usage differs depending on the meaning and the relationship between consecutive words. Because changes may occur.

In order to determine the correct "reading" for a sentence, a multifaceted analysis of a single sentence is required.

FIG. 3 shows the basic configuration of a conventional text reading device.

In the figure, 31 is a text input section, and 32 is a text analysis section.

33 is a reading dictionary, 34 is a reading sequence storage section, and 35 is an audio output section.

The text input unit 31 inputs Japanese text using a keyboard or the like.

The text analysis unit 32 analyzes the input text data using the reading dictionary 33, identifies each word, creates a pronunciation sequence thereof, and stores the pronunciation sequence in the pronunciation sequence storage unit 34. Identification of words in a sentence is performed by extracting candidate words from the reading dictionary 33 and performing DP matching with one sentence data to select an optimal combination of words.

In the reading dictionary 33, the notation of the i-word, its pronunciation, usage, and other grammar are registered. The words are kanji words.

Includes katakana, kanji, kana, and other words.

The audio output unit 35 takes out the pronunciation of the sentence from the pronunciation storage unit 34, performs speech synthesis, and outputs the result as a voice.

In this way, if any written text is input into the reading device, an appropriate reading sequence will be automatically created and the text will be read aloud.

[Problem that the invention seeks to solve]

With conventional sentence reading devices, if there are consecutive vowels in a created reading sequence, for reasons of speech synthesis, they are uttered as long vowels during reading, such as the notation ``aa''. The utterances corresponding to the notation "ah" were all "ah", and it was not possible to distinguish and hear the difference between the nine notations. To give a specific example, "sugar seller" is pronounced as "satori." In addition, as another example, a notation with consecutive numbers, such as "r2222", is
There was a problem with the words being uttered as "nee nee nee nee" when being read aloud, making them difficult to distinguish.

[Means for solving problems]

The present invention makes it possible to insert silence codes into necessary parts of the reading sequence when analyzing two sentences, and sets silent intervals during reading aloud, thereby making it easier to distinguish between words.

FIG. 1 shows the basic configuration of the present invention.

In the figure, 11 is a text input section, and 12 is a text analysis section.

13 is a reading dictionary; 14 is a reading sequence storage unit; 15 is an audio output unit; 121 is a word identification process; 122 is a silence insertion condition determination process; and 123 is a reading sequence creation process.

The text input unit 11 inputs the notation of the text to be read out,
It is supplied to the text analysis section 12 as character string data of Kanji code.

The text analysis unit 12 divides the input character string data into various types, searches the reading dictionary 13, performs word identification processing 121, and analyzes the primary identified word string.

Silence insertion condition determination processing 122 is performed to identify a portion that matches a predetermined silence code insertion condition such as inter-word vowel connection/continuation or continuous digits, and to specify a silence code insertion position. In this manner, the reading sequence creation process 123 is performed according to the identified word sequence and the silent code insertion instruction.

The reading sequence storage unit 14 temporarily stores the reading sequence data created by the text analysis unit 12.

The audio output unit 15 reads out the reading sequence data stored in the reading sequence storage unit 14, synthesizes the data, and outputs the voice.

At this time, if a silence code exists in the pronunciation sequence data, an appropriate fixed silence interval is inserted at that position to delimit the audio output.

[Effect]

According to the present invention, by setting in advance the conditions for inserting a silent section when reading aloud, a silent code can be inserted in the reading sequence for any written text, and the silent section can be used as a delimiter during the reading. can be entered.

The conditions for inserting a silent code are when the same vowel is in contact with the same vowel at the boundary between successive words in the identified word string, or when pre-specified word groups such as numbers are consecutive. , any other conditions can be used. When a word string matching these conditions is detected, a silence code is inserted between the corresponding words.

As a way to specify a specific group of words.

A method of making it identifiable using flags, etc. in the reading dictionary.

Alternatively, there is another way to hold it on a table.

In addition, to insert a silence code, read two types of words for which a silence code may be inserted, one with and without a silence code, and register them in the dictionary, and then judge the conditions for inserting a silence code. The simplest method is to select one of the words according to the results, but it is also possible to insert only words without silence codes in the reading dictionary and add silence codes by program according to the results of the 9-condition judgment. be.

〔Example〕

FIG. 2 shows the configuration of one embodiment of the present invention.

The configuration shown in FIG. 2 is based on the configuration in FIG. 1, and the same reference numerals are used. However, because it is necessary to explain at a concrete level, 1 sentence analysis section 1
2, additional detailed steps are shown.

Reading Dictionary 13 includes sand, sugar, desert... and numbers 1, 2, etc., although illustrations are omitted. ..., for words that should be pronounced with a 9th grade (silence).

Two types are prepared in advance, one without a silent code and one with one, and one of them can be selected. Of course, for specific words that must be uttered with breaks, only those with silence codes can be used.

The character string data based on the notation of the sentence inputted from the sentence input section 11 is first classified into various types in the word identification process 121 of the two-sentence analysis section 12, and then the reading dictionary 13 is searched.

As a result of the search, a plurality of candidate words are obtained, and these search result words are sequentially stored in a table called a word lattice table.

Next, the words in the word lattice table are compared to determine the optimal word string corresponding to the input sentence.

In the next silence insertion condition determination process 122, conditions for inserting a silence code are checked for the determined word string, and between words that satisfy one condition.

Instructs to insert a silence code.

In the next pronunciation sequence creation process, pronunciation data is created based on the word string accompanied by a silent code insertion instruction. Here, the silence code is inserted at the position in the reading sequence where the silence code insertion instruction was given. For example, the reading sequence that corresponds to the notation "sugar counter" is:

Sato, silent code, and Uriba.

The created pronunciation data is once stored in the pronunciation storage unit 14, and then is synthesized into speech in the audio output unit 15, and a silent section (pause) is set in the silent code and output as audio. For example, in the “sugar counter” example above, “
A silent section is placed between "Sato" and "Uriba" to separate the words.

The length of the silent section corresponding to the silent code is set to an appropriate length in advance. However, by continuously inserting a plurality of silence codes, it is possible to extend the silence interval arbitrarily. It is also possible to provide a plurality of types of silence codes so that the length of the silence section can be specified arbitrarily.

〔Effect of the invention〕

According to the present invention, it becomes easy to distinguish between words in a reading sequence outputted from a single-sentence reading device, and work efficiency and reliability can be improved when the reading sequence is used for reading proofreading.

[Brief explanation of drawings]

FIG. 1 is a basic configuration diagram of the present invention, FIG. 2 is a configuration diagram of one embodiment of the present invention, and FIG. 3 is a basic configuration diagram of a conventional device. In Figure 1. 11: Text input unit 12: Text analysis unit 13: Reading dictionary 14: Reading sequence storage unit 15: Audio output unit 121: Word identification process 122: Silence insertion condition determination process 123: Reading sequence creation process

Claims (1)

[Scope of Claims] A pronunciation dictionary that provides the pronunciation of a written word and a sentence analysis unit are provided, and the text analysis unit receives the written text as input and analyzes the text using the reading dictionary, and generates a pronunciation sequence of the sentence. In the text reading device that generates and outputs audio, the text analysis unit includes means for inserting a silence code into the pronunciation of the word determined as a result of the text analysis, and inserts silence into the pronunciation sequence according to the determination result of the silence code insertion condition. A text reading device that is characterized by inserting a code.