JP2746880B2 - Compound word division method - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a compound word dividing method, and more particularly to an accent unit extracting method for text-to-speech synthesis. 2. Description of the Related Art In order to add natural prosody in speech synthesis, it is essential to set the position and level of an accent nucleus. In order to do so, first, an input text is converted into an accent unit (a morpheme having only one accent nucleus). Column). At this time, the accent unit usually consists of one or more clauses. When a long compound word forms one clause, it is necessary to divide this compound word (clause) into two or more accent units. As a method for this, there is a method of performing dependency analysis in a compound word, and a method of dividing a compound word into 1 /
A method of dividing at position 2 has been proposed. In the former method, it is necessary to perform dependency analysis without information on particles and auxiliary verbs, and the processing becomes complicated. The latter method has a problem that improper division is often caused when a very long compound word of 5 words or more, a word string divided into two or an odd word is used. Object The present invention has been made in view of the above-mentioned circumstances, and in particular, in a text synthesis of speech, a long compound word which becomes unnatural when spoken as one accent unit is formed in a plurality of appropriate positions. This is done for the purpose of dividing into accent units. Configuration In order to achieve the above object, the present invention extracts morphological analysis information from input text by morphological analysis processing,
Based on the morphological analysis information, a phoneme symbol string is obtained according to a phoneme rule and a prosody symbol string is obtained according to a prosody symbol derivation rule. A parameter sequence of a prepared phoneme segment is read out according to the phoneme symbol string, and a prosody is added according to a combination rule. In the text-to-speech synthesis apparatus, when there is a long compound word having a suffix inserted in the input text, and there is a suffix present in the suffix dictionary in the compound word, a part immediately after this suffix is divided. Position, and if there is a prefix present in the prefix dictionary in the compound word, the position immediately before this prefix is used as a division position, and the compound word is divided into accent units to produce speech in one accent unit. The feature is that unnatural long compound words are uttered more naturally. Hereinafter, a description will be given based on examples of the present invention. FIG. 1 is a flow chart for explaining one embodiment of the present invention. As will be described below, the present invention uses a method of focusing a long compound word on a plurality of accents by focusing on affixes in the compound word. It is divided into units. In FIG.
First, if an affix exists in a compound word, this is a large semantic break. Therefore, a suffix is searched from a compound ending and a prefix is searched from a compound head. If a suffix is found here, the part immediately after the suffix is found, and if a prefix is found, the part immediately before the prefix is used as the dividing position to divide the word into two compound words. FIG. 2 is an example in which the suffix B is found, and FIG. 3 is an example in which the prefix C is found (where A is an accent unit). If the divided compound words are still long, this process is further repeated. Considering the average mora length of the words that make up the compound word, 1
It is appropriate that one accent unit is composed of two or three words (one in this case, one affix). Therefore, if neither a suffix nor a prefix exists in the compound word, a division position is set every two words from the compound word head as shown in FIG. Effects As is clear from the above description, according to the present invention, a phrase can be divided into a plurality of accent units from a long compound word by a method focusing on an affix sandwiched in the compound word.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a flowchart for explaining one embodiment of the present invention;
2 to 4 are diagrams showing examples of compound word division. A: Accent unit, B: Suffix, C: Prefix.

Claims (1)

(57) [Claims] Morphological analysis information is extracted from the input text by morphological analysis processing, a phonemic symbol sequence is obtained by a phonemic rule based on the morphological analysis information, a prosodic symbol sequence is obtained by a prosodic symbol derivation rule, and a parameter sequence of a phoneme segment prepared in advance is obtained.
In a text-to-speech synthesizer that reads in accordance with the phoneme symbol string and adds a prosody according to a combination rule, when a long compound word having an affix is inserted in the input text, the compound word is included in a suffix dictionary in the compound word. If there is a suffix, the division position is set immediately after this suffix, and if there is a prefix existing in the prefix dictionary in the compound word,
A compound word characterized in that immediately before the prefix is used as a division position, and the compound word is divided into accent units so that an unnatural long compound word can be uttered more naturally when uttered in one accent unit. Split method.