JPH09281993A - Phonetic symbol forming device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain a speech output free from an odd feel by imparting correct phonetic symbols to the foreign language character strings among input character strings of a sentence-speech converter which converts the sentences to speeches receiving sentences. SOLUTION: The part of the character string of the language 2 included in input character string of inputted language 1 is specified by a language 2 character string decision section 103. The character string part of this language 2 is subjected to a morpheme analysis in a language 2 morpheme analysis section 102. The result thereof is subjected to a construction analysis by a language 2 construction analysis section 106, thereby, the result is subjected to the construction deformation of the pronunciation. The phonetic symbol string of the language 2 is formed in a language 2 phonetic symbol forming section 108 and converted into the phonetic symbol string of the language 1 by a language 2-language 1 phonetic symbol conversion section 109, thereby, the phonetic symbol of the language 1 correct for the language 2 character string parts included in the language 1 input character strings is formed.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention can be applied to a sentence-to-speech conversion device for converting an input sentence into a voice and outputting the voice, and can correctly read a foreign language character string contained in the input sentence. It relates to a phonetic symbol generation method necessary for giving.

[0002]

2. Description of the Related Art In a Japanese sentence-speech conversion apparatus, phonetic symbols necessary for performing linguistic analysis on an input sentence, determining reading of each word in the sentence, and outputting voice are roughly divided. It is generally composed of two parts: a language processing part for generating a sound signal and a sound processing part for synthesizing an actual voice signal based on the phonetic symbol information received from the language processing part. In the language processing unit, a morphological analysis unit that searches the input sentence for a dictionary and decomposes the input sentence into word strings, collects the word string that is the result of the morphological analysis into bunsetsu units,
It is generally composed of a syntactic analysis unit that performs pronunciation modification processing, a pronunciation symbol generation unit that generates pronunciation symbols from word readings, and an accent imparting unit that determines reading accents in word units or phrase units. . In the language processing section, particularly, a section related to determining a reading of a word and generating a phonetic symbol for the reading is a morphological analysis section, a syntactic analysis section, and a phonetic symbol generation section. The present invention relates to a processing method of these morphological analysis unit, syntactic analysis unit and phonetic symbol generation unit, and more particularly to handling of foreign language character strings included in an input sentence.

In the conventional phonetic symbol generation method, regarding the handling of a foreign language character string, a phonetic symbol assigned to a foreign language character string registered in the dictionary is generated, and it is not registered in the dictionary. Prepares rules for converting foreign language strings to foreign language phonetic symbols, and applies those rules to the detected foreign language strings to obtain the foreign language phonetic symbols. A method has been used in which the foreign language phonetic symbols are converted into Japanese phonetic symbols and generated (see, for example, Japanese Patent Laid-Open No. 5-298364, hereinafter referred to as system 1).

[0004]

However, the method 1
Then, there was a problem that the correct phonetic symbol could not be added to the foreign language character string that does not conform to the prepared conversion rule, that is, it becomes an exceptional pronunciation. Also,
In order to give a correct phonetic symbol to a foreign language character string that does not conform to the conversion rules, it can be handled only by registering the foreign language character string in the word dictionary, and it is possible that foreign words are mixed in the word dictionary. Then, there was a problem that it caused confusion in dictionary management. Further, there is a problem that the capacity of the word dictionary becomes unnecessarily large and the dictionary search processing time in the morpheme processing is increased. In method 1, since the conversion rule is applied to the foreign language character string for each word and the phonetic symbol is assigned, for example, "I have to go." And "I
There was a problem that the pronunciation of "have" could not be used properly for the two English sentences "have a car." On the other hand, regarding the phonetic symbol generation method in the European-speech converter, foreign language character strings, especially, There is a problem in that it is practically impossible to deal with the conversion rules for phonetic symbols when handling a Japanese character string by the method 1. In consideration of such a problem, the present invention takes an input sentence into consideration. It is intended to provide a method for giving a correct phonetic symbol to a foreign language character string inside.

[0005]

In a sentence-speech conversion device for inputting a sentence and outputting a voice, a language 1 morpheme analysis unit for performing a morpheme analysis of a first language and the language 1 morpheme analysis unit can be analyzed. And a language 2 morphological analysis unit that performs morphological analysis of the character string of the second language identified by the language 2 character string determination unit, and The language 2 syntactic analysis means for performing necessary transformations on the pronunciation information of the words included in the morpheme string output by the language 2 morphological analysis means, and the phonetic symbols of the second language from the reading output by the language 2 morphological analysis means. A phonetic symbol generating device comprising: a language-2 phonetic symbol generating means for outputting; and a language2-language-1 phonetic symbol converting means for converting the language-2 phonetic symbol generating means into phonetic symbols of language-1.

For each of the n types of languages i, the language i
A phonetic symbol generation device having a morpheme analysis means, a language i character string determination means, a language i syntax analysis means, and a phonetic symbol conversion means between languages.

FIG. 1 shows the principle of the present invention. In FIG.
The case where there are two language types is shown. In the figure, a language 1 morpheme analysis unit 101 and a language 2 morpheme processing unit 102 search an input character string of each language for a dictionary and divide it into morpheme strings. The language 2 character string determination unit 103 determines the language 1
The morphological analysis unit 101 identifies the second language character string that has failed in analysis. The language 1 character string determination unit 104 identifies the first language character string that the language 2 morphological analysis unit 102 failed to analyze. Language 1 syntax analysis unit 105, Language 2 syntax analysis unit 1
06 performs necessary transformation on the pronunciation information included in the morpheme string of each language. Language 1 Phonetic symbol generator 10
7. The language 2 phonetic symbol generator 108 generates phonetic symbols from the phonetic information of each language. The language 2-language 1 phonetic symbol conversion unit 109 converts the phonetic symbol string of the second language into the phonetic symbol string of the first language. The language 1-language 2 phonetic symbol conversion unit 110 converts phonetic symbol strings in the first language into phonetic symbol strings in the second language.

In FIG. 1, an input character string in the first language is passed through a language 1 morphological analysis processing unit 101, a language 1 syntactic analysis unit 105, and a language 1 phonetic symbol generation unit to a phonetic symbol string in the first language. Is converted to.

When the input character string of the first language includes the second language character string, when the processing of the language 1 morphological analysis unit 101 reaches the position of the second language character string, Since the analysis fails, the language 2 character string determination unit 103
Specify the language character string part of. The second language character string identified by the language 2 character string determination unit is the language 2 morphological analysis unit 102.
To the language 2 syntactic analysis unit 106, the language 2 phonetic symbol generation unit 108, and the language 2-language 1 phonetic symbol conversion unit.
Are converted to phonetic symbols in the language.

As described above, the foreign language character string for the first language, that is, the second language character string is subjected to the morphological analysis processing including the dictionary search by the language 2 morphological analysis unit, so that it is exceptional. It is possible to correctly assign phonetic symbols including pronunciation words.

The word information of the first language can be registered in the dictionary of the language 1 morpheme analysis unit 101, and the word information of the second language can be registered in the dictionary of the language 2 morpheme analysis unit 102. Because of the configuration, it is possible to avoid mixing word information of each language in a single dictionary.

Furthermore, the capacity of the dictionary for each language can be suppressed to the amount required for each language.

In the language 2 syntax analysis unit 106, since it corresponds to a change in pronunciation necessary for the syntax, for example, the pronunciation of "have" in two English sentences "I have to go." And "I have a car." It is possible to use them properly.

Further, even when Japanese, which is written in ideographs whose pronunciation cannot be derived from the notation, is included in a European sentence, morphological analysis and syntactic analysis are performed on the Japanese character string.
It is possible to correctly convert the phonetic symbol string.

[0015]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 2 shows a block diagram of a phonetic symbol generating apparatus according to an embodiment of the present invention. Japanese morphological analyzer 201
Divides the input character string in Japanese into a sequence of Japanese morphemes while searching the Japanese word dictionary 202. The Japanese syntactic analysis unit 207 performs necessary syntactic modification on the pronunciation information of the Japanese morpheme string output by the Japanese morphological analysis unit 201. The Japanese phonetic symbol generation unit 209 generates a Japanese phonetic symbol string based on the pronunciation information of the Japanese morpheme string received from the Japanese syntax analysis unit 207.

On the other hand, the English morpheme analysis unit 203 divides the English input character string into English morpheme strings while searching the English word dictionary 204. English parser 20
Reference numeral 8 applies a syntactically necessary modification to the pronunciation information of the English morpheme string output by the English morpheme analysis unit 203. The English phonetic symbol generation unit 210 generates an English phonetic symbol string based on the pronunciation information of the English morpheme string received from the English syntax analysis unit 208.

When an English character string is included in the Japanese input character string, when the Japanese morphological analysis unit 201 fails in the analysis, the English character string determination unit 205 identifies the English character string portion, After converting the English character string portion into an English phonetic symbol string through the English morpheme analysis unit 203, the English syntax analysis unit 208, and the English phonetic symbol generation unit 210, the English-Japanese phonetic symbol conversion unit 211 converts the Convert to phonetic symbol string to obtain Japanese phonetic symbol string for English character string part.

When a Japanese character string is included in the English input character string, the Japanese character string determination unit 206 identifies the Japanese character string portion when the analysis by the English morphological analysis unit 203 fails. After converting the Japanese character string portion into a Japanese phonetic symbol string through the Japanese morpheme analysis unit 201, the Japanese syntax analysis unit 207, and the Japanese phonetic symbol generation unit 209, the Japanese-English phonetic symbol conversion unit. At 213, it is converted into an English phonetic symbol string to obtain an English phonetic symbol string for the Japanese character string portion.

The English-Japanese phonetic symbol conversion unit 211 and the Japanese-English phonetic symbol conversion unit 213 respectively convert the English-Japanese phonetic symbol conversion table 2 when converting phonetic symbols.
12. The phonetic symbols are converted by referring to the Japanese-English phonetic symbol conversion table 214. FIG. 3 shows an example of the English-Japanese phonetic symbol conversion table. Japanese phonetic symbols are written in Roman letters. In FIG.
The left side is a vowel conversion table and the right side is a consonant conversion table. In the vowel conversion table, '-' indicates that the preceding vowel is pronounced as a long vowel. In the consonant conversion table, the part enclosed in parentheses indicates that the consonant is converted to the phonetic symbol when it is the last in the phonetic symbol string. FIG. 4 shows an example of the Japanese-English phonetic symbol conversion table. In FIG. 4, five'a ',' i ',' u ',' e ', and'o' from the upper left side are vowel conversion tables, and the other five are consonant conversion tables.

In the above embodiment, the target languages are two types, Japanese and English, but the target languages may be two or more types. In that case, correct phonetic symbols can be added to the character strings of a plurality of languages included in the input character string, as in the above embodiment.

[0021]

As is apparent from the above description,
The present invention assigns a correct phonetic symbol to a foreign language character string included in a sentence even when it has an exceptional pronunciation when converting a sentence of a language into a phonetic language. It has the advantage that it can be converted into a phonetic symbol string of, and the listener can hear the converted voice without any discomfort.

[Brief description of drawings]

FIG. 1 is a diagram showing the operation of the present invention.

FIG. 2 is a diagram showing a configuration of an embodiment of the present invention.

FIG. 3 is a diagram showing an example of an English-Japanese pronunciation conversion table.

FIG. 4 is a diagram showing an example of a Japanese-English pronunciation conversion table.

Claims (2)

[Claims] 1. A sentence-speech conversion device for inputting a sentence and outputting a voice, wherein a language 1 for performing morphological analysis of a first language.
The morpheme analysis means, the language 2 character string determination means for identifying the character string of the second language that the language 1 morpheme analysis means could not analyze, and the second identified by the language 2 character string determination means.
Language 2 morpheme analysis means for performing morpheme analysis of a character string of the language, and language 2 syntactic analysis means for performing necessary transformations on pronunciation information of words included in the morpheme sequence output by the language 2 morpheme analysis means, Language 2 phonetic symbol generating means for outputting phonetic symbols of the second language from the reading output by the language 2 syntax analyzing means, and language 2-language 1 pronunciation for converting the language 2 phonetic symbol generating means into the phonetic symbols of language 1. A phonetic symbol generating apparatus having a symbol converting means. 2. A language i morpheme analysis means, a language i character string determination means, a language i syntax analysis means, and a phonetic symbol conversion means between each language for each of the n kinds of languages i.
Description phonetic symbol generator.