JPS63140329A - Sentence reading system - Google Patents

Abstract

PURPOSE:To read aloud a sentence smoothly without being conscious of fluctuation in a naturalized word, by providing a conversion part which convert the naturalized word having the fluctuation to a standard for. CONSTITUTION:When the naturalized word exists in a text inputted from a sentence input part 11, and it is not the standard form, it is converted to the standard form at the conversion part 121, and a word dictionary 16 is retrieved. In other words, when the text inputted from the input part 11 is the naturalized word having the fluctuation, a dictionary retrieval part 12 sends it to the conversion part 121. The conversion part 121 converts it to the reference form by a prescribed conversion rule, and retrieves the word dictionary. A word identification part 13 selects an optimum word string based on a retrieved result, and generates a reading string by a reading string storage part 14, and performs voice synthesis at a voice output part, then, outputs voice. In such way, it is possible to read aloud the sentence without being conscious of the fluctuation of the naturalized word.

Description

[Detailed Description of the Invention] [Summary] The present invention uses fluctuations in a text reading method in a Japanese information processing device to facilitate word dictionary searches and word identification using foreign words with fluctuations in written sentences. A conversion unit is provided to convert a certain foreign word into a standard form, and a word dictionary search is performed on the foreign word converted to the standard form to identify the word.

[Industrial application field]

The present invention relates to a Japanese language information processing device, and more particularly to a text reading method that inputs written text, analyzes the text to identify words in the character string, and outputs the text aloud.

[Prior art and problems to be solved by the invention] In Japanese information processing, there is a word dictionary that stores the notation parts of words, etc. consisting of character strings such as words, phrases, phrases, etc., their pronunciation, grammar, etc. , a sentence reading method is already known in which a written sentence is referred to in this word dictionary, words, etc. are fixed, and a reading sequence is generated.

In this case, the written sentence may include katakana loanwords with so-called "fluctuations", such as "violin" for violin, "whiskey" for whiskey, etc. In other words, foreign words are mainly words that have been introduced into the Japanese language from Western languages, and the problem is how to write the original sounds that do not exist in Japanese, and the way they are written is not necessarily the same depending on the word. This is because it varies depending on the person.

As the Japanese language allows foreign words to be written in different ways, it is important to develop word processing technology for foreign words as early as possible for machine translation, text-mail reading, database search output, and text analysis for reading devices. This has become a major issue.

FIG. 32 is a block diagram of a conventional text reading system. In FIG. 3, 31 is a text input section, 32 is a dictionary search section,
33 is a word identification section, 34 is a reading sequence storage section, 35 is an audio output section, and 36 is a word dictionary. The dictionary search section 32 and the word identification section 33 constitute a text analysis section 37.

The text analysis unit 37 includes a text input unit 31 such as a keyboard, for example.
The word dictionary 36 is used to refer to the input sentence, and candidate words are picked up, and the word identification section 33 selects the optimal combination of words. The word dictionary 36 stores word notations, their pronunciations, grammar, accents, etc. From these, a pronunciation sequence is created by a pronunciation storage section 34, and a speech output section 35 performs speech synthesis and outputs speech.

As is clear from the above explanation, if all variations of a loan word are not registered in the word dictionary because the writing style is not fixed, if a loan word that is not registered in the dictionary exists in a sentence, word identification will be difficult. This makes it difficult to read sentences smoothly. For example, if the words ``data,''``whiskey,'' and ``variation'' are registered in the dictionary, but the sentences are written as ``data,''``whiskey,'' and ``variation,'' then Like this, word identification fails.

[Means for solving problems]

The object of the present invention is to provide a text reading method that solves the above-mentioned problems, and the means thereof is to set a standard form for writing katakana foreign words in advance, and read out katakana foreign words that have fluctuations in written sentences. The present invention is characterized in that it includes a conversion unit that converts the converted standard form into the standard form according to a predetermined conversion rule, and performs a word dictionary search on the converted standard form to identify words.

〔Example〕

FIG. 1 is a block diagram of a text reading method according to the present invention. In FIG. 1, 11 is a text input section, 12 is a dictionary search section, 13 is a word identification section, 121 is a foreign word standard form conversion section,
14 is a reading sequence storage section, 15 is an audio output section, and 16 is a word dictionary section. The dictionary search section 12, the word identification section 13, and the foreign word standard form conversion section 121 constitute the sentence analysis section 17.

In such a configuration, if there is a foreign word in a sentence inputted from the text input section 11 using a keyboard or the like and it is not in the prescribed standard form, the conversion section 121 converts it into the standard form and converts it into the word dictionary 16. search for. That is, text input section 1
For example, if the sentence input in step 1 is ``the sound of a violin on a computer,'' the dictionary search unit 12 sends the katakana notation in the sentence to the loan word standard form conversion unit 121, and converts it to the ``voice'' of ``violin.'''' as ``ha'' and the long part of ``computer'' as ``computer'', search the word dictionary 16, and pink-amplify candidate words. As shown in Figure 2, the word dictionary 16 includes "notation", "reading", "grammar", and "accent" that indicate the conversion result to the standard form.
etc. are stored.

The loanword standard form conversion unit 121 converts "Vua, Vui, Vu, Vue, Vuo" in the above example into [Ba, Bi, Bu, Be, Bo], "
"Excluding all long sounds" and "rai, uni, wo"
It has rules for converting fluctuations in written sentences, such as changing ``rai, uni, wo''.

This conversion is performed by software, for example, using IF statements, GOTO
Converted by sentences, etc.

After picking up the candidate words, the word identification section 13 selects the most suitable word string, and the reading string storage section 14 stores the reading string "Violin no Otoo Computer De" according to the determined word string.
The speech output unit 15 performs speech synthesis based on the reading sequence and outputs speech.

〔Effect of the invention〕

According to the present invention, in a text-to-speech method for generating pronunciations from sentences in a Japanese information processing device, smooth word identification processing and text analysis of Japanese are possible without worrying about fluctuations in foreign words. .

[Brief explanation of the drawing]

FIG. 1 is a configuration diagram of a text reading method according to the present invention, Figure 2 is a detailed diagram of the word dictionary in the configuration shown in Figure 1, and FIG. 3 is a conventional configuration diagram. (Explanation of symbols) 11, 31... text input section, 12, 32...dictionary search section, 13, 33...word identification section, 14.34... reading sequence storage section, 15.35...Audio output section, 16, 36...word dictionary, 121...Loan word standard form conversion unit. Configuration diagram of conventional text reading method Figure 3

Claims (1)

[Claims] 1. In the text reading method of a Japanese information processing device that inputs a written sentence, analyzes the text to identify the words in the string, and outputs the sound, a standard form for writing katakana foreign words is set in advance. , comprising a converting unit that converts a katakana foreign word with fluctuations in the written sentence into the standard form according to a predetermined conversion rule, and performs a word dictionary search on the converted standard form to identify the word. Characteristic text reading method.