JP3279261B2 - Apparatus, method, and recording medium for creating a fixed phrase corpus - Google Patents

Abstract

PROBLEM TO BE SOLVED: To efficiently create a fixed form sentence corpus to be accumulated in a speech data base used for generating a synthetic speech. SOLUTION: Fixed form words and a position of an arbitrary word to be inserted between the fixed form words are inputted from the fixed form word input circuit 1. The attribute of the arbitrary word inserted and the insertion position are inputted from an arbitrary word input circuit 2. An arbitrary word selection part 3 retrieves a word dictionary 4, and selects and extracts all the words having the same attribute as the inputted arbitrary word from the registered words. A fixed form sentence creation part 5 inserts each of the words extracted by the arbitrary word selection part 3 between the fixed form words respectively and generates a fixed form sentence. The created fixed form sentences are each outputted from a fixed form sentence output part 6, and accumulated in a data base 10 provided in the speech synthesis device connected through a LAN, etc.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus and method for creating a fixed-text corpus, and a recording medium on which a program for executing the method is recorded. It is suitable for collecting a sentence corpus.

[0002]

2. Description of the Related Art A speech synthesis system for outputting an arbitrary text as a synthesized speech is disclosed in Japanese Patent Laid-Open Publication No. 8-87297 (hereinafter, this speech synthesis system is referred to as Conventional Example 1). FIG. 6 is a block diagram showing a main configuration of a speech synthesis system according to Conventional Example 1, which includes a speech information search unit 21, a speech information database 22, a synthesized speech generation unit 23, and a synthesized speech generation rule 24. I have.

[0003] In the conventional speech synthesis system,
When a text or a phonetic symbol string is input, the natural voice is analyzed, and the voice information searching unit 21 stores the input text or pronunciation in a voice information database 22 storing the extracted voice feature amounts and the corresponding utterance contents. A search is made to see if there is any utterance content that matches the symbol string. If a matching utterance content is found, the speech information search unit 21 passes this to the synthesized speech generation unit 23. Then, the speech synthesis generation unit 23 performs a process according to the speech information to generate a synthesized speech. On the other hand, if no match is found,
The voice information search unit 21 passes the input text or phonetic symbol string as it is to the synthetic voice generating unit 23, and the synthetic voice generating unit 23 generates a synthetic voice based on the synthetic voice generation rule 24.

Japanese Patent Application Laid-Open No. 9-244680 discloses a prosody control device for synthesizing a sentence in which only a specific word of a fixed sentence changes, such as a response sentence of a telephone guidance system (for example). Hereinafter, this prosody control device is referred to as Conventional Example 2.) 2. Description of the Related Art Conventionally, a speech synthesizer has used natural speech as it is for synthesizing a fixed sentence portion, synthesizing only arbitrary words according to rules, and simply connecting them. On the other hand, the prosody control device of Conventional Example 2 controls the prosody pattern of an arbitrary word to match the prosody pattern of a fixed sentence.

[0005]

However, the above conventional example has the following problems. First, Conventional Example 1
When an appropriate speech waveform does not exist in the speech information database, the speech synthesis device generates a synthesized speech based on the generation rule by the synthesized speech generation unit 23 or the waveform generation unit (not shown). Will deteriorate. Therefore, unless you have a data-balanced audio information database that fits as much text as possible,
Practically, there is a problem that a synthesized sound that sounds natural can not be generated.

In the prosody control device of Conventional Example 2, the prosody pattern of an arbitrary word is controlled to match the prosody pattern of a fixed sentence. However, if the control exceeds an allowable range, the sound quality of the arbitrary word itself will be reduced. Will deteriorate. For this reason, there is a problem that a synthetic voice that sounds natural can not be generated practically if a prosody pattern sufficient to control an arbitrary word falls within an allowable range is prepared.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-described problems of the conventional example, and has a fixed form corpus which can be efficiently created in a fixed sentence corpus stored in a sound database used for generating a synthesized sound. An object of the present invention is to provide a sentence corpus creation apparatus and method, and a computer-readable recording medium on which a program for executing the method is recorded.

[0008] To achieve the above object, the first aspect of the present invention.
The fixed form corpus creation device according to the aspect of the present invention includes a dictionary means for storing a plurality of words and attributes of each word in association with each other, a fixed form word included in the fixed form sentence to be created, and represented by a desired word. A fixed word input means for inputting the position of an arbitrary word to be inserted between fixed words in the fixed sentence, and inputting the position of the arbitrary word to be inserted in the fixed sentence and the attribute of the arbitrary word to be inserted in association with each other An arbitrary word input means,
An optional word selecting means for selecting and extracting from the dictionary means a word corresponding to the attribute of the optional word input from the optional word input means; and an optional word at the position of the optional word input from the fixed form input means. A fixed sentence generating means for generating a fixed sentence by inserting each of the words extracted by the word selecting means ,
And the dictionary means corresponds to each of a plurality of words.
And remember the number of mora or syllables for each word
And the arbitrary word selecting means selects the word of the selected word.
The number of mora or syllables corresponding to each
Extracted together from the column and extracted by the optional word selection means
The number of mora extracted along with the extracted word
Or a word classification means for classifying according to the number of syllables
For example, the fixed sentence generating means may include
It is characterized in that a fixed phrase is generated by adding information about a class .

In order to achieve the above object, a second aspect of the present invention is provided.
The fixed phrase corpus creation device according to the aspect of
And a dictionary for storing attributes of each word in association with each other
Means and the words included in the fixed phrase to be created.
Between the fixed form word and the fixed form word in the fixed form sentence.
Fixed word input means for inputting the position of an arbitrary word to be input;
The position of the optional word to be inserted in the fixed phrase and the task to be inserted
An arbitrary word input means for inputting in association with a meaning attribute,
Corresponds to the attribute of an arbitrary word input from the arbitrary word input means
Arbitrary word selection for selecting and extracting words to be extracted from the dictionary means
Selection means, and an arbitrary word input from the fixed form input means.
In the position, the word extracted by the arbitrary word selecting means
A fixed phrase generation means for generating a fixed phrase by inserting each;
And the dictionary means corresponds to each of a plurality of words.
And memorize the accent position of each word.
And the optional word selecting means selects each of the selected words.
The position of the corresponding accent is also extracted from the dictionary means.
Out, the word extracted by the arbitrary word selecting means,
According to the position of the accent extracted along with the word
The standard sentence generating means,
Adds information on classification by the word classification means
And generating a fixed phrase.

In order to achieve the above object, a third aspect of the present invention is provided.
The fixed phrase corpus creator according to the viewpoint includes a plurality of words.
And a dictionary for storing attributes of each word in association with each other
Means and the words included in the fixed phrase to be created.
Between the fixed form word and the fixed form word in the fixed form sentence.
Fixed word input means for inputting the position of an arbitrary word to be input;
The position of the optional word to be inserted in the fixed phrase and the task to be inserted
An arbitrary word input means for inputting in association with a meaning attribute,
Corresponds to the attribute of an arbitrary word input from the arbitrary word input means
Arbitrary word selection for selecting and extracting words to be extracted from the dictionary means
Selection means, and an arbitrary word input from the fixed form input means.
In the position, the word extracted by the arbitrary word selecting means
A fixed phrase generation means for generating a fixed phrase by inserting each;
And the dictionary means corresponds to each of a plurality of words.
And store information indicating the phonological environment of each word.
And the optional word selecting means selects each of the selected words.
The information indicating the corresponding phonological environment is added from the dictionary means.
Words extracted and extracted by the arbitrary word selecting means
In accordance with the information indicating the phoneme environment extracted along with the word.
Further comprising word classification means for classifying
Generating means for generating information on classification by the word classification means;
Is added to generate a fixed sentence. Said
The word classification means further receives the input from the fixed form input means.
Between the fixed phrase and the arbitrary word corresponding to the extracted word
Extracted by the optional word extracting means according to the relation
Words may be classified.

In order to achieve the above object, a fourth aspect of the present invention is provided.
The fixed phrase corpus creator according to the viewpoint includes a plurality of words.
And a dictionary for storing attributes of each word in association with each other
Means and the words included in the fixed phrase to be created.
Between the fixed form word and the fixed form word in the fixed form sentence.
Fixed word input means for inputting the position of an arbitrary word to be input;
The position of the optional word to be inserted in the fixed phrase and the task to be inserted
An arbitrary word input means for inputting in association with a meaning attribute,
Corresponds to the attribute of an arbitrary word input from the arbitrary word input means
Arbitrary word selection for selecting and extracting words to be extracted from the dictionary means
Selection means, and an arbitrary word input from the fixed form input means.
In the position, the word extracted by the arbitrary word selecting means
A fixed phrase generation means for generating a fixed phrase by inserting each;
And the dictionary means corresponds to each of a plurality of words.
In addition, information about the sound of each word is also stored,
The arbitrary word selecting means corresponds to each of the selected words.
Information relating to the sound to be extracted is also extracted from the dictionary means,
The word extracted by the arbitrary word selecting means is
Simply classify according to the information about the sound extracted along with
Further comprising word classification means, wherein the fixed phrase generation means comprises:
Adds information about the classification by the word classification means,
The method is characterized by generating a sentence.

In order to achieve the above object, a fifth aspect of the present invention is provided.
The fixed phrase corpus creation device according to the viewpoint
A fixed phrase included in the type sentence and represented by a desired word;
In the fixed phrase, the position of an arbitrary word to be inserted between fixed words and
Inputting a fixed form word, and in the fixed form sentence,
The position of the arbitrary word to be inserted and the attribute of the arbitrary word to be inserted are paired.
Inputting an arbitrary word in response to the inputting;
Word corresponding to the attribute of the arbitrary word entered in the force step
Is stored in association with a plurality of words and respective attributes.
An optional word extraction step of selecting and extracting from a dictionary
At the position of any word entered in the fixed word input step,
Insert each word extracted in the optional word extraction step
Anda fixed sentence generating step of generating a template sentence and,
The optional word extracting step includes: storing the plurality of words in the dictionary.
Of information about sound stored in association with
The information about the sound corresponding to the selected word
And the word extracted in the optional word extracting step
According to the information about the sound extracted along with the word.
A word classification step for classifying the fixed form sentence;
The forming step is related to the classification in the word classifying step.
In addition, a fixed phrase is generated by adding information to be executed .

[0013] To achieve the above object, a sixth aspect of the present invention.
Fixed text corpus creation method according to the aspect, the constant to be generated
A fixed phrase included in the type sentence and represented by a desired word;
In the fixed phrase, the position of an arbitrary word to be inserted between fixed words and
Inputting a fixed form word, and in the fixed form sentence,
The position of the arbitrary word to be inserted and the attribute of the arbitrary word to be inserted are paired.
Inputting an arbitrary word in response to the inputting;
Word corresponding to the attribute of the arbitrary word entered in the force step
Is stored in association with a plurality of words and respective attributes.
An optional word extraction step of selecting and extracting from a dictionary
At the position of any word entered in the fixed word input step,
Insert each word extracted in the optional word extraction step
A fixed phrase generation step of generating a fixed phrase by performing
The notation dictionary is further associated with each of a plurality of words.
The accent position of the word is stored, and the arbitrary word extraction is performed.
The exit step is an action corresponding to each of the selected words.
The cent position is also extracted from the dictionary, and the arbitrary word is extracted.
The word extracted in the extraction step is added to the word.
Words to be classified according to the position of the extracted accent
Type step, wherein the fixed phrase generation step includes:
Adding information on classification by the word classification step
And generating a fixed phrase .

In order to achieve the above object, a seventh aspect of the present invention is provided.
The fixed phrase corpus creation method according to the viewpoint
A fixed phrase included in the type sentence and represented by a desired word;
In the fixed phrase, the position of an arbitrary word to be inserted between fixed words and
Inputting a fixed form word, and in the fixed form sentence,
The position of the arbitrary word to be inserted and the attribute of the arbitrary word to be inserted are paired.
Inputting an arbitrary word in response to the inputting;
Word corresponding to the attribute of the arbitrary word entered in the force step
Is stored in association with a plurality of words and respective attributes.
An optional word extraction step of selecting and extracting from a dictionary
At the position of any word entered in the fixed word input step,
Insert each word extracted in the optional word extraction step
And a fixed sentence generating step of generating a fixed sentence, before
The notation dictionary is further associated with each of a plurality of words.
Information indicating a phonological environment of the word is stored, and
The extracting step includes the sound corresponding to each of the selected words.
The information indicating the rhyme environment is also extracted from the dictionary, and
The word extracted in the meaning extraction step is added to the word
Words classified according to the information indicating the extracted phoneme environment
Type step, wherein the fixed phrase generation step includes:
Adding information on classification by the word classification step
And generating a fixed phrase . The word classification
The step is further input in the fixed form input step.
Between the fixed form word and the arbitrary word corresponding to the extracted word
Word extracted in the optional word extraction step according to
May be classified.

In order to achieve the above object, an eighth aspect of the present invention is provided.
The fixed phrase corpus creation method according to the viewpoint
A fixed phrase included in the type sentence and represented by a desired word;
In the fixed phrase, the position of an arbitrary word to be inserted between fixed words and
Inputting a fixed form word, and in the fixed form sentence,
The position of the arbitrary word to be inserted and the attribute of the arbitrary word to be inserted are paired.
Inputting an arbitrary word in response to the inputting;
Word corresponding to the attribute of the arbitrary word entered in the force step
Is stored in association with a plurality of words and respective attributes.
An optional word extraction step of selecting and extracting from a dictionary
At the position of any word entered in the fixed word input step,
Insert each word extracted in the optional word extraction step
And a fixed sentence generating step of generating a fixed sentence, before
The notation dictionary is further associated with each of a plurality of words.
Stores information about the sound of words and extracts the arbitrary words
The steps relate to the sound corresponding to each of the selected words.
Information to be extracted from the dictionary, and the arbitrary word extraction
The words extracted by the step are extracted together with the words.
A word classification step that classifies according to the information about the emitted sounds.
The fixed form sentence generation step,
Add information about the classification by the word classification step
It is characterized by generating a type sentence .

In order to achieve the above object, a ninth aspect of the present invention is provided.
The computer-readable recording medium according to the aspect is
Included in the fixed phrase to be generated, represented by the desired word
The fixed phrase and the task to insert between the fixed words in the fixed phrase.
A fixed word input step of inputting a position of a meaning word;
Arbitrary words to be inserted in the type sentence and arbitrary words to be inserted
An optional word input step of inputting in association with the attribute of
The attribute of the arbitrary word input in the optional word input step is
Corresponding words, multiple words and their attributes are associated
Word extraction step to select and extract from the dictionary
In the fixed word input step
At the position, the word extracted in the optional word extraction step is
A fixed phrase generation step of generating a fixed phrase by inserting each
And the optional word extracting step includes:
The sound stored in association with each word
Information related to the sound corresponding to the selected word.
Information to be extracted at the same time, and
The output word is related to the sound extracted along with the word.
Professional to perform word classification step to classify according to information
Further recording the gram, and the fixed form generation step
Adds information about classification in the vocabulary classification step
And record the program that executes the process to generate the fixed phrase
It is characterized by doing.

To achieve the above object, the tenth aspect of the present invention
Readable recording medium according to the aspect of the present invention
Is included in the fixed phrase to be generated, and is represented by the desired word.
Inserted between the fixed phrase and the fixed phrase in the fixed phrase.
Inputting the position of an arbitrary word to be input
The position of the optional word to be inserted in the fixed phrase and the task to be inserted
Arbitrary word input step of inputting in association with the meaning attribute
And the attribute of the arbitrary word input in the optional word input step
The word corresponding to is paired with multiple words and their attributes.
Arbitrary word extraction that selects and extracts from dictionaries stored and associated
Step and any of the input in the fixed form input step
In the position of the word, the word extracted in the optional word extraction step
To generate a fixed phrase by inserting
And the dictionary includes a pair for each of a plurality of words.
And memorize the accent position of each word.
The optional word extracting step includes selecting each of the selected words.
The corresponding accent position is extracted from the dictionary.
And the words extracted by the optional word extraction step
According to the position of the accent extracted along with the word.
Further comprising a word classification step for classifying
The generation step is related to the classification by the word classification step.
Perform processing to generate a fixed phrase by adding information
It is characterized by recording a program.

To achieve the above object, an eleventh aspect of the present invention is provided.
Readable recording medium according to the aspect of the present invention
Is included in the fixed phrase to be generated, and is represented by the desired word.
Inserted between the fixed phrase and the fixed phrase in the fixed phrase.
Inputting the position of an arbitrary word to be input
The position of the optional word to be inserted in the fixed phrase and the task to be inserted
Arbitrary word input step of inputting in association with the meaning attribute
And the attribute of the arbitrary word input in the optional word input step
The word corresponding to is paired with multiple words and their attributes.
Arbitrary word extraction that selects and extracts from dictionaries stored and associated
Step and any of the input in the fixed form input step
In the position of the word, the word extracted in the optional word extraction step
To generate a fixed phrase by inserting
Comprising a flop, wherein the dictionary pairs in each of a plurality of words
And memorize the information indicating the phonological environment of each word
And the optional word extraction step includes the step of extracting the selected word.
Information indicating the phonological environment corresponding to each is combined from the dictionary.
Extracted in the optional word extraction step
In accordance with the information indicating the phoneme environment extracted along with the word.
Further comprising a word classification step of classifying
In the sentence generation step, the classification by the word classification step is performed.
Process to generate a fixed phrase by adding information about
The program is characterized by recording a program . For the word
The type step includes the fixed form input in the fixed form input step.
The positional relationship between the type word and the arbitrary word corresponding to the extracted word
Therefore, the words extracted in the optional word extraction step are separated.
You may make it similar.

To achieve the above object, the present invention provides a twelfth aspect.
Readable recording medium according to the aspect of the present invention
Is included in the fixed phrase to be generated, and is represented by the desired word.
Inserted between the fixed phrase and the fixed phrase in the fixed phrase.
Inputting the position of an arbitrary word to be input
The position of the optional word to be inserted in the fixed phrase and the task to be inserted
Arbitrary word input step of inputting in association with the meaning attribute
And the attribute of the arbitrary word input in the optional word input step
The word corresponding to is paired with multiple words and their attributes.
Arbitrary word extraction that selects and extracts from dictionaries stored and associated
Step and any of the input in the fixed form input step
In the position of the word, the word extracted in the optional word extraction step
To generate a fixed phrase by inserting
Comprising a flop, wherein the dictionary pairs in each of a plurality of words
And also memorize information about the sound of each word.
The optional word extracting step includes selecting each of the selected words.
Extracts information about the sound corresponding to the sound from the dictionary
And the words extracted by the optional word extraction step
According to the information about the sound extracted along with the word.
The method further comprises a word classification step of classifying the fixed form sentence.
The forming step relates to the classification by the word classifying step.
By adding that information, up to execute a process of generating template sentence
Recording the program.

In the above-mentioned recording medium, the arbitrary word extracting step
The multiple words used in the and the attributes of each word
A dictionary to be stored in association with the dictionary may be further recorded.

[0021]

Embodiments of the present invention will be described below with reference to the accompanying drawings.

[First Embodiment] FIG. 1 is a block diagram showing a configuration of a fixed-form sentence corpus creating apparatus according to this embodiment. As shown in the figure, this fixed phrase corpus creation device includes a fixed phrase input unit 1, an optional word input unit 2, an optional word selection unit 3, a word dictionary 4, a fixed phrase generation unit 5, a fixed phrase output unit. 6 is provided. Further, the fixed form corpus creation device is connected to a speech database of a speech synthesis system having a database 10 via a LAN (Local Area Network) or the like.

By designing a corpus necessary for generating or modeling a prosody pattern of a synthesized speech in advance using this fixed-sentence corpus creating apparatus, a fixed-sentence sentence such as a response sentence of a telephone guidance system can be obtained. The prosody pattern necessary to generate a sentence in which only certain words change in a certain word can be stored in the speech database. Further, with this fixed-sentence corpus creation device, it is possible to prepare a sufficient prosodic pattern in the speech database so that the control of the prosodic pattern of an arbitrary word falls within an allowable range.

The fixed phrase input unit 1 includes an input device such as a keyboard, and inputs a fixed phrase and a position of an arbitrary word to be inserted between the fixed phrases. For example, the position of an arbitrary word inserted between fixed words and fixed words is, for example, “[A] is now in [B].”
And so on. In this case, “just now”,
“Ha” and “middle” are fixed words, and [A] and [B] are insertion positions of arbitrary words.

The arbitrary word input unit 2 includes an input device such as a keyboard, and inputs an attribute of an arbitrary word to be inserted and its insertion position. The attribute of the arbitrary word to be inserted and its insertion position are
In the above example, the input is, for example, “person name [A], operation [B]”.

The optional word selection unit 3 searches the word dictionary 4 and
All words having attributes that match the attributes of the arbitrary word input from the arbitrary word input unit 2 are selected and extracted. When there are a plurality of arbitrary words to be inserted between the fixed words, the arbitrary word selection unit 3 selects and extracts all the words corresponding to each. The arbitrary word selection unit 3 also extracts other information stored in association with each word.

The word dictionary 4 is a dictionary used for extracting words by the arbitrary word selection unit 3, and includes, for example, headings, readings, mora numbers, accent positions, parts of speech, word attributes, frequency information, etc. of each word. They are stored in association with each other.
A configuration example of the word dictionary 4 will be described later in detail.

The fixed phrase generator 5 inserts each word extracted by the optional word selector 3 between fixed words input from the fixed word input unit 1 to generate a fixed phrase. Fixed phrase generator 5
Generates a fixed phrase in which, when there are a plurality of arbitrary words to be inserted, words extracted corresponding to each of the words are inserted between the fixed words in all combinations thereof.

The fixed phrase output unit 6 outputs the fixed phrase generated by the fixed phrase generation unit 5 onto a LAN, and stores the fixed phrase as a fixed phrase corpus in a database of a speech synthesis system connected to the LAN.

FIG. 2 is a diagram showing a specific example of the word dictionary 4 of FIG. In this example, in the word dictionary 4, headings (words), readings, mora, accents, parts of speech, and attributes are registered in association with each other.
"Ryoko", "Aichi", "Taxi", "Bus", "Yamagata", "Train", "Subway", "Yokohama", "Yui",
"Love" and "Airplane" are registered.

For example, words whose attributes are names include "Akemi", "Ryoko", "Yui", and "love". Among them, "Akemi" and "Ryoko" have a mora number of "3", "Yui" and "love"
Means that the mora number is "4". In addition, words whose attributes are vehicles include “taxi”, “bus”, “train”, “subway”, and “airplane”. Of these, “taxi” and “bus” have an accent prefix (number “ 1), "train" and "subway" have no accent (indicated by the number "0"), and "airplane" has accents in the word (indicated by the number "2").

Hereinafter, the operation of the fixed phrase corpus creating apparatus according to this embodiment will be described with reference to specific examples. In the following description, the word dictionary 4 is based on the example shown in FIG.

First, the operator inputs "[A]" as a fixed word from the fixed word input unit 1 and an arbitrary word to be inserted between the fixed words.
How are you? And input "person name [A]" as an attribute of an arbitrary word from the arbitrary word input unit 2. Thereby, a personal name is supplied to the arbitrary word selection unit 3 as an attribute corresponding to [A]. .

When the arbitrary word selection unit 3 searches the word dictionary 4, the words "Akemi", "Ryoko", "Yui", and "Love" are searched as words whose attributes are personal names. Select and extract. And an arbitrary word selection unit 3
Supplies these four words to the fixed phrase generator 5.

"[A] -chan, how are you?" Input from the fixed-form word input unit 1 is supplied to the fixed-form sentence generation unit 5, and the optional word selection unit 3 is added to [A] in this sentence. Insert each of the four words provided by "Akemi-chan,
How are you. "" How are you, Ryoko-chan? "" How are you, Yui-chan? "Ai-chan, how are you doing? Is generated.

The four fixed phrases generated in this way are output from the fixed phrase output unit 6 to the LAN, and are stored as a fixed phrase corpus in the database 10 provided in the speech synthesis system connected to the LAN.

As described above, according to the apparatus for creating a fixed phrase corpus according to the present embodiment, the position of an arbitrary word to be inserted between fixed words and the attribute and position of the arbitrary word can be simply input. You can create various fixed phrases,
A sentence pattern required for use as a speech database in a speech synthesis system can be efficiently created.

[Second Embodiment] FIG. 3 is a block diagram showing the configuration of a typical sentence corpus creating apparatus according to this embodiment. This fixed phrase corpus creation device differs from that of the first embodiment (FIG. 1) in further including a mora comparison unit 7.

The mora comparison unit 7 classifies the words extracted from the word dictionary 4 by the arbitrary word selection unit 3 into mora groups according to the number of mora, and forms the fixed phrase generation unit 5 in order for each mora group. To supply. Further, the fixed phrase generation unit 5 generates a fixed phrase to which information on the classified mora group is added. The fixed phrase output unit 6 outputs a fixed phrase in association with information on the mora group,
It is stored in the database 10.

The operation of the fixed-text corpus creation apparatus according to this embodiment will be described below based on a specific example. In the following description, it is assumed that the word dictionary 4 follows the example shown in FIG. 2, as in the first embodiment.

In this embodiment, the arbitrary word selecting section 3
It is assumed that the process up to the extraction of a word by is the same as that of the first embodiment. However, the arbitrary word selection unit 3 supplies the extracted four words and the number of mora extracted for each word together with the four words to the mora comparison unit 7 instead of the fixed phrase generation unit 5.

Next, the mora comparing section 7 classifies the extracted four words into mora groups according to the number of mora. Here, "Akemi" and "Ryoko" having the number of mora of 3 are included. In the mora group “M3”, the number of mora is 2
Are classified into the mora group “M2”. Then, in order for each Mora group,
Each word is supplied to the fixed phrase generation unit 5.

The standard sentence generating unit 5 is supplied with “[A], how are you?” Input from the standard word input unit 1. First, an arbitrary word selecting unit is added to [A] in this sentence. Insert “Akemi” and “Ryoko” classified into the mora group “M3” supplied from 3, respectively, and add “Akemi-chan,
How are you. "" How are you, Ryoko-chan? Then, the fixed phrase generator 5 adds information “M3” to the two fixed phrases.

Next, the fixed-form sentence generating unit 5 sends [A] in the sentence "[A], how are you?"
In addition, the mora group “M” supplied from the arbitrary word selection unit 3
3 ”,“ Ui ”and“ Ai ”are inserted, respectively, to generate four fixed phrases“ Ui-chan, how are you? ”And“ Ai-chan, how are you? ”. Then, the fixed phrase generation unit 5 adds information “M2” to the two fixed phrases.

The four fixed sentences generated in this way are output from the fixed sentence output unit 6 to the LAN together with information indicating the classification of the mora group “M3” or “M2” added to each of the fixed sentences. The data is stored as a fixed sentence corpus in the database 10 provided in the connected speech synthesis system.

As described above, in the fixed sentence corpus creating apparatus according to this embodiment, words extracted from the word dictionary 4 as arbitrary words to be inserted are classified according to the number of mora, and the fixed form is further generated. Sentences are also classified according to the number of mora. Therefore, a sentence pattern including words having a desired number of mora required for use as a speech database in the speech synthesis system can be efficiently created.

By the way, when a sentence in which a word having a different number of mora from a word included in the fixed sentence is inserted based on the fixed sentence stored in the voice information database, the prosody pattern is also controlled. Required. The fixed-sentence corpus creation device according to the present embodiment can easily create a plurality of fixed-sentences classified by the number of mora, so that many fixed-sentences are stored in the audio information database so that control of the prosodic pattern becomes unnecessary. It is also effective in the sense that it is stored.

[Third Embodiment] FIG. 4 is a block diagram showing the configuration of a fixed phrase corpus creating apparatus according to the third embodiment. This fixed-form sentence corpus creation device differs from that of the first embodiment (FIG. 1) in further including an accent comparison unit 8.

The accent comparing section 8 classifies the words extracted from the word dictionary 4 by the optional word selecting section 3 into accent groups corresponding to the respective accent positions.
It is supplied to the fixed phrase generation unit 5 in order for each accent group. Further, the fixed phrase generation unit 5 generates a fixed phrase while retaining information on the classified accent groups.
The fixed phrase output unit 6 outputs a fixed phrase in association with information on the accent group, and stores the fixed phrase in the database 10.

Hereinafter, the operation of the fixed phrase corpus creating apparatus according to this embodiment will be described based on a specific example. In the following description, it is assumed that the word dictionary 4 follows the example shown in FIG. 2, as in the first embodiment.

First, the operator inputs, as an arbitrary word to be inserted between the standard word and the standard word from the standard word input unit 1, "It takes 30 minutes to the destination [A]." , "Vehicle [A]" is input as an attribute of an arbitrary word. Thereby, the vehicle is supplied to the arbitrary word selection unit 3 as an attribute corresponding to [A].

When the optional word selecting section 3 searches the word dictionary 4, five words having the attribute "personal name" such as "taxi", "bus", "train", "subway" and "airplane" are searched. Select and extract all these. Then, the arbitrary word selection unit 3 supplies the extracted five words and the positions of the accents of the respective words extracted together with the five words to the accent comparison unit 8.

Next, the accent comparison unit 8 extracts the extracted 4
The two words are classified into an accent group according to the position of the accent. In this case, “taxi” and “bus” having the accent position at the beginning of the word belong to the accent group “A1”, and “train” and “ “Subway” is classified into the accent group “A0”, and “airplane” with the accent position in the word is classified into the accent group “A2”. Then, the words are supplied to the fixed phrase generation unit 5 in order for each accent group classification.

The fixed phrase generator 5 is supplied with “It takes 30 minutes to [A] to the destination” input from the fixed phrase input unit 1. First, an arbitrary word is added to [A] in the sentence. Insert "taxi" and "bus" classified into the accent group "A1" supplied from the selection unit 3, respectively,
It generates a fixed phrase "It takes 30 minutes by taxi to the destination.""It takes 30 minutes by bus to the destination." Then, the fixed phrase generation unit 5 adds information “A1” to the two fixed phrases.

Next, the fixed phrase generator 5 adds the accent group “A0” supplied from the optional word selector 3 to [A] in the supplied “It takes 30 minutes to [A] to the destination.” Insert the "train" and "subway" classified as "", respectively, and generate a fixed phrase "It takes 30 minutes by train to the destination.""It takes 30 minutes by subway to the destination." Then, the fixed phrase generation unit 5 adds information “A0” to the two fixed phrases.

Furthermore, the fixed phrase generation unit 5 supplies the accent group “A2” supplied from the optional word selection unit 3 to [A] in the supplied “It takes 30 minutes to [D] to the destination.” Insert "Airplane" classified as
Generate a fixed phrase saying "It takes 30 minutes by air to reach the destination." Then, the fixed phrase generation unit 5 adds information “A2” to the fixed phrase.

The five fixed phrases generated in this way are respectively assigned “A1”, “A0” or “A2”.
Along with the information indicating the classification of the accent group, the standard sentence output unit 6 outputs the information to the LAN, and is stored as a standard sentence corpus in the database 10 provided in the speech synthesis system connected to the LAN.

As described above, in the fixed sentence corpus creating apparatus according to this embodiment, the words extracted from the word dictionary 4 as arbitrary words to be inserted are classified according to the position of the accent, and the fixed form is further generated. Sentences are also classified according to the position of the accent. Therefore, it is possible to efficiently create a sentence pattern including a word having a desired accent position required for use as a speech database in the speech synthesis system.

By the way, the pitch pattern is considerably different between words having different accent positions, so that substitution is difficult. The fixed-sentence corpus creation device according to this embodiment can easily create a plurality of fixed-sentences classified according to accent positions, thereby avoiding a situation in which speech synthesis is performed based on fixed-sentences including words that are difficult to substitute. In a sense, it is effective.

[Fourth Embodiment] FIG. 5 is a block diagram showing the configuration of a fixed-form sentence corpus creation apparatus according to this embodiment. This fixed phrase corpus creation device differs from that of the first embodiment (FIG. 1) in further including a phoneme environment comparison unit 9.

The phoneme environment comparison unit 9 classifies the words extracted from the word dictionary 4 by the arbitrary word selection unit 3 into phoneme groups corresponding to the combination of the beginning and / or the end phoneme based on each reading. It is supplied to the fixed phrase generation unit 5 in order for each phoneme group. The fixed sentence generation unit 5 generates a fixed sentence while retaining information on the classified phoneme groups. The fixed phrase output unit 6 outputs a fixed phrase in association with information about the phoneme group, and stores the fixed phrase in the database 10.

The method of grouping performed by the phoneme environment comparison unit 9 differs depending on whether the insertion position of the arbitrary word is at the beginning of a sentence, in a sentence, or at the end of a sentence. In other words, if the insertion position of the arbitrary word is the beginning of the sentence, it follows the reading of the beginning of the sentence, and if it is in the sentence, it follows the reading of both the beginning and the end of the reading,
If it is at the end of the sentence, the speech is grouped into phoneme groups according to the ending reading.

Hereinafter, the operation of the fixed phrase corpus creation apparatus according to this embodiment will be described with reference to specific examples. In the following description, the word dictionary 4 contains
Headlines (words) include "Yamaoka", "Masuoka", "Yamada"
It is assumed that "Mita" and "Yamamoto" are registered. The readings corresponding to these headings are "Yamaoka",
"Masuoka", "Yamada", "Mita", "Yamamoto". These parts of speech and attributes are all proper nouns and surnames.

Here, as the arbitrary words inserted by the operator from the fixed word input unit 1 and between the fixed words, (1) “Mr. [A], how are you?” (2) “[ A] Introduce the teacher. ”(3) Consider the case where“ My maiden name is [A] ”. In any case, the operator inputs “surname [A]” as an attribute of an arbitrary word from the arbitrary word input unit 2.

In any of the cases (1) to (3), the arbitrary word selection unit 3 searches the word dictionary 4 and finds that the attribute is a person name as “Yamaoka”, “Masuoka”, “Yamada”,
Since “Mita” and “Yamamoto” are searched, all of them are selected and extracted. Then, the optional word selection unit 3
These five words are supplied to the fixed phrase generator 5. The arbitrary word selection unit 3 supplies the extracted five words and the reading of each word extracted together with the five words to the phoneme environment comparison unit 9.

Next, the phoneme environment comparing section 9 classifies the extracted five words into groups according to the phoneme environment based on the reading, and in the above cases (1) to (3), The way is different.

In the case of the above (1), the arbitrary word is at the head of the sentence, and the connection point with the fixed form word is only the ending. As a result, the phoneme environment comparison unit 9 classifies “Yamaoka” and “Masuoka” having the same ending reading among the extracted five words into the same group. Since the remaining three words do not have the same ending reading, the phoneme environment comparison unit 9 classifies the three words into one group for each word.

In the case of the above (2), the arbitrary word is in the middle of the sentence, and the connection point with the fixed form word is both the beginning and the end of the sentence. The phonological environment comparison unit 5 classifies the five words as one group for each word, because none of the five extracted words has the same reading at both the beginning and the end.

In the case of the above (3), the arbitrary word is at the end of the sentence, and the connection point with the fixed form word is only the beginning of the word. As a result, the phoneme environment comparison unit 9 determines that “Yamaoka”, “Yamada”, and “Yamamoto”, which have the same initial reading among the five extracted words,
Are classified into the same group. In the remaining three words,
Since no ending reading is the same, the phoneme environment comparison unit 9
Classifies the two words as one group for each word.

The subsequent processes are substantially the same as those in the second and third embodiments, and
In either case of (3), the generated fixed phrase is stored in the database 10 as a fixed phrase corpus together with information indicating the group classification.

As described above, in the fixed-sentence corpus creating apparatus according to this embodiment, words extracted from the word dictionary 4 as arbitrary words to be inserted are classified according to the phonological environment of the ending and the beginning of the word, and further generated. The fixed phrases to be sent are also classified according to the phonetic environment of the ending and the beginning. Therefore, a sentence pattern to be strengthened in the speech database can be efficiently created. Further, since the manner of grouping is changed depending on the presence or absence of fixed words before and after the arbitrary word, fine classification can be performed according to the insertion position of the arbitrary word.

[Modifications of Embodiment] The present invention is not limited to the above-described first to fourth embodiments, and various modifications and applications are possible. Hereinafter, modifications of the above-described embodiment applicable to the present invention will be described.

In the second embodiment, the words extracted from the word dictionary are classified into groups according to the number of mora. However, the groups may be classified according to the number of syllables. Further, the number of moras (or the number of syllables) described in each of the second to fourth embodiments,
The group may be classified by combining the position of the accent and the phonological environment. Further, the extracted words may be classified into groups according to the information about the sound when each of the other words is generated.

In the first to fourth embodiments, the arbitrary word selection unit 3, the fixed phrase generation unit 5, the mora comparison unit 7, the accent comparison unit 8, and the phonological environment comparison unit 9 operate under program control. By doing so, he created a fixed phrase corpus. On the other hand, a program for realizing such a function may be stored in a computer-readable recording medium such as a CD-ROM, distributed, installed in a general-purpose computer, and executed. In this case, the recording medium may also store and provide the word dictionary 4.

[0075]

As described above, according to the present invention,
Various fixed phrases can be created simply by inputting the position of an arbitrary word to be inserted between the fixed words and the attribute and position of the arbitrary word. For this reason, for example, a sentence pattern required for use as a speech database in a speech synthesis system can be efficiently created.

Further, classification is performed in accordance with information on the sound of each word, such as the number of mora or the number of syllables, the position of accents, or the phonological environment, thereby creating a fixed sentence. Can be created efficiently.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a fixed-form sentence corpus creation apparatus according to a first embodiment of the present invention.

FIG. 2 is a diagram showing an example of the word dictionary of FIG.

FIG. 3 is a block diagram showing a configuration of a fixed phrase corpus creation device according to a second embodiment of the present invention.

FIG. 4 is a block diagram showing a configuration of a fixed phrase corpus creation device according to a third embodiment of the present invention.

FIG. 5 is a block diagram showing a configuration of a fixed phrase corpus creation device according to a fourth embodiment of the present invention.

FIG. 6 is a block diagram showing a configuration of a speech synthesis system of Conventional Example 1.

[Explanation of symbols]

 Reference Signs List 1 fixed phrase input unit 2 optional word input unit 3 optional word selection unit 4 word dictionary 5 fixed phrase generation unit 6 fixed phrase output unit 7 mora comparison unit 8 accent comparison unit 9 phonological environment comparison unit 10 database

Claims (16)

(57) [Claims] 1. A dictionary means for storing a plurality of words and attributes of each word in association with each other; a fixed form word included in a fixed form sentence to be created and represented by a desired word; Fixed word input means for inputting the position of an arbitrary word to be inserted between fixed words, and arbitrary word input means for inputting the position of the arbitrary word to be inserted in the fixed phrase and the attribute of the arbitrary word to be inserted in association with each other An optional word selecting means for selecting and extracting a word corresponding to an attribute of an arbitrary word input from the optional word input means from the dictionary means; and, at a position of the arbitrary word input from the fixed form input means, and a fixed sentence generating means for generating a standard sentence by inserting each word extracted by any word selection means, said dictionary means, in association with each of a plurality of words
And memorize the number of mora or syllables for each word.
The optional word selecting means corresponds to each of the selected words.
The number of mora or syllables
And extracting the word extracted by the arbitrary word selecting means into the word
According to the number of mora or syllables extracted along with
The system further comprises word classifying means for classifying , wherein the fixed form sentence generating means
A fixed form corpus creating apparatus characterized in that a fixed form sentence is generated by adding information relating to the fixed form. 2. The method according to claim 1 , wherein the plurality of words and the attribute of each word are
Dictionary means for storing in association with each other, and a dictionary means which is included in a fixed phrase to be created and is represented by a desired word.
The fixed phrase and the task to insert between the fixed words in the fixed phrase.
A fixed word input means for inputting the position of a meaning word, and the position of an arbitrary word to be inserted in the fixed phrase and the position to be inserted
Arbitrary word input means for inputting in association with the attribute of an arbitrary word
And the attribute of the arbitrary word input from the arbitrary word input means.
Any word election to the word to be extracted by selecting from the dictionary hand stage
Selection means and an arbitrary word position input from the fixed form input means,
Insert words extracted by the arbitrary word selection means
And a fixed phrase generating means for generating a fixed phrase by inputting , wherein the dictionary means further stores the position of the accent of each word in association with each of a plurality of words; and A word classification for extracting the position of an accent corresponding to each of the selected words from the dictionary means, and classifying the words extracted by the arbitrary word selecting means according to the positions of the accents extracted in accordance with the words; further comprising means, the typical sentence generating means, and characterized in that by adding information on the classification by the word classification means, to generate a template sentence
To create a fixed phrase corpus. 3. The method according to claim 1 , wherein the plurality of words and the attribute of each word are respectively
Dictionary means for storing in association with each other, and a dictionary means which is included in a fixed phrase to be created and is represented by a desired word.
The fixed phrase and the task to insert between the fixed words in the fixed phrase.
A fixed word input means for inputting the position of a meaning word, and the position of an arbitrary word to be inserted in the fixed phrase and the position to be inserted
Arbitrary word input means for inputting in association with the attribute of an arbitrary word
And the attribute of the arbitrary word input from the arbitrary word input means.
Arbitrary word selection for selecting and extracting words to be extracted from the dictionary means
Selection means and an arbitrary word position input from the fixed form input means,
Insert words extracted by the arbitrary word selection means
And a fixed-form sentence generating means for generating a fixed-form sentence , wherein the dictionary means further stores information indicating a phonological environment of each word in association with each of a plurality of words; The means further extracts information indicating a phonological environment corresponding to each of the selected words from the dictionary means, and indicates the words extracted by the arbitrary word selecting means to the phonological environments extracted together with the words. further comprising a word classification means for classifying according to the information, the typical sentence generating means, and characterized in that by adding information on the classification by the word classification means, to generate a template sentence
To create a fixed phrase corpus. 4. The word categorizing means further comprises: extracting a word extracted by the arbitrary word extracting means in accordance with a positional relationship between a fixed word input from the fixed word input means and an arbitrary word corresponding to the extracted word. The fixed-form sentence corpus creation device according to claim 3 , wherein classification is performed. 5. The method according to claim 1 , wherein the plurality of words and the attribute of each word are
Dictionary means for storing in association with each other, and a dictionary means which is included in a fixed phrase to be created and is represented by a desired word.
The fixed phrase and the task to insert between the fixed words in the fixed phrase.
A fixed word input means for inputting the position of a meaning word, and the position of an arbitrary word to be inserted in the fixed phrase and the position to be inserted
Arbitrary word input means for inputting in association with the attribute of an arbitrary word
And the attribute of the arbitrary word input from the arbitrary word input means.
Arbitrary word selection for selecting and extracting words to be extracted from the dictionary means
Selection means and an arbitrary word position input from the fixed form input means,
Insert words extracted by the arbitrary word selection means
A fixed-form sentence generating means for generating a fixed-form sentence , wherein the dictionary means further stores information on the sound of each word in association with each of a plurality of words, and A word classification for extracting information on sounds corresponding to each of the selected words from the dictionary means together, and classifying the words extracted by the arbitrary word selection means according to the information on the sounds extracted along with the words; further comprising means, the typical sentence generating means, and characterized in that by adding information on the classification by the word classification means, to generate a template sentence
To create a fixed phrase corpus. 6. A fixed form word input step of inputting a fixed form word included in a fixed form sentence to be generated and represented by a desired word, and a position of an arbitrary word inserted between the fixed form words in the fixed form sentence, An arbitrary word input step of inputting the position of the arbitrary word to be inserted in the fixed phrase and the attribute of the arbitrary word to be inserted, and a plurality of words corresponding to the attribute of the arbitrary word input in the arbitrary word input step. An arbitrary word extraction step of selecting and extracting from the dictionary the words and their attributes are stored in association with each other, at the position of the arbitrary word input in the fixed form input step,
Anda fixed sentence generating step of generating a fixed sentence and the words extracted by the arbitrary word extraction step respectively inserted, said any word extraction step, the plurality of words in the dictionary
Of information about sound stored in association with
The information about the sound corresponding to the selected word
The word extracted in the arbitrary word extraction step as the word
Words to be classified according to the information about the extracted sounds
The method further includes a classification step, wherein the fixed phrase generation step includes the step of:
A method for creating a fixed sentence corpus, characterized in that a fixed sentence is generated by adding information on classifications to be made. 7. A desired word included in a fixed phrase to be generated,
Between the fixed phrase represented by the fixed phrase and the fixed phrase in the fixed phrase.
Input step to input the position of the arbitrary word to be inserted into the
And the position of an arbitrary word to be inserted in the fixed phrase
Arbitrary word input step for inputting in association with the attribute of the arbitrary word
And the attribute of the arbitrary word input in the optional word input step.
Corresponding words, multiple words and their attributes are associated
Word extraction step to select and extract from the dictionary
At the position of an arbitrary word input in the fixed word input step,
Insert the words extracted in the optional word extraction step
And generating a fixed phrase by inputting the fixed phrase, wherein the dictionary is further associated with each of a plurality of words.
The position of the accent of each word is stored, and the optional word extracting step includes:
Extract the corresponding accent position from the dictionary
And extracting the word extracted in the arbitrary word extracting step
Classify according to position based on accents extracted along with words
Further comprising a word classifying step, wherein the fixed form sentence generating step includes:
Creates a fixed phrase corpus by generating fixed phrases by adding information on classification
How . 8. A desired word included in a fixed phrase to be created,
Between the fixed phrase represented by the fixed phrase and the fixed phrase in the fixed phrase.
Input step to input the position of the arbitrary word to be inserted into the
And the position of an arbitrary word to be inserted in the fixed phrase
Arbitrary word input step for inputting in association with the attribute of the arbitrary word
And the attribute of the arbitrary word input in the optional word input step.
Corresponding words, multiple words and their attributes are associated
Word extraction step to select and extract from the dictionary
At the position of an arbitrary word input in the fixed word input step,
Insert the words extracted in the optional word extraction step
And generating a fixed phrase by inputting the fixed phrase , wherein the dictionary is further associated with each of a plurality of words.
Stores the indicate to information phoneme environment of each word, the arbitrary word extraction step, each of the selected word
Information indicating the corresponding phonological environment is also extracted from the dictionary
The word extracted in the arbitrary word extraction step is added to the word.
Classify according to the extracted information indicating the phonetic environment
The method further includes a word classification step, wherein the fixed sentence generation step includes the step of:
Generating fixed phrases by adding information about classifications
A method for creating a fixed-text corpus . 9. The word classification step further includes the step of extracting the word extracted in the arbitrary word extraction step according to a positional relationship between the fixed word input in the fixed word input step and an arbitrary word corresponding to the extracted word. 9. The method according to claim 8 , wherein classification is performed. 10. A desired word included in a fixed phrase to be created,
And the fixed form word in the fixed form sentence
A fixed word input step to input the position of an arbitrary word to be inserted between
And the position of an arbitrary word to be inserted in the fixed phrase
Arbitrary word input step for inputting in association with the attribute of the arbitrary word
And the attribute of the arbitrary word input in the optional word input step.
Corresponding words, multiple words and their attributes are associated
Word extraction step to select and extract from the dictionary
At the position of an arbitrary word input in the fixed word input step,
Insert the words extracted in the optional word extraction step
And generating a fixed phrase by inputting the fixed phrase , wherein the dictionary is further associated with each of a plurality of words.
The information on the sound of each word is stored, and the arbitrary word extracting step includes:
Information on the corresponding sound is also extracted from the dictionary , and the word extracted in the arbitrary word extraction step is
Classify according to information about sounds extracted along with words
Further comprising a word classification step , wherein the fixed sentence generation step is performed by the word classification step.
Generating fixed phrases by adding information about classifications
A method for creating a fixed-text corpus . 11. A computer-readable recording medium.
There are, included in the template text to be generated, the fixed word represented by the desired word, the fixed word input step of inputting the position of any words to be inserted between fixed word in the template text, the template text An arbitrary word input step of inputting the position of the arbitrary word to be inserted and the attribute of the arbitrary word to be inserted in association with each other, and converting the word corresponding to the attribute of the arbitrary word input in the arbitrary word input step into a plurality of words And an optional word extraction step of selecting and extracting from a dictionary storing the respective attributes in association with each other, at the position of the arbitrary word input in the fixed form input step,
And a fixed sentence generating step of generating a fixed sentence and the words extracted by the arbitrary word extraction step respectively inserted, said any word extraction step, the plurality of words in the dictionary
Of information about sound stored in association with
The information about the sound corresponding to the selected word
The word extracted in the arbitrary word extraction step as the word
Words to be classified according to the information about the extracted sounds
The program further includes a program for executing a classification step, wherein the fixed sentence generation step includes the step of:
Process for generating fixed phrases by adding information on
A computer-readable recording medium for recording a program for executing processing . 12. A computer-readable recording medium.
There are, included in the template text to be generated is represented by the desired word
The fixed phrase and the task to insert between the fixed words in the fixed phrase.
A fixed word input step of inputting the position of the meaning word, and the position of an arbitrary word to be inserted in the fixed sentence and the word to be inserted
Arbitrary word input step for inputting in association with the attribute of the arbitrary word
And the attribute of the arbitrary word input in the optional word input step.
Corresponding words, multiple words and their attributes are associated
Word extraction step to select and extract from the dictionary
At the position of an arbitrary word input in the fixed word input step,
Insert the words extracted in the optional word extraction step
And generating a fixed phrase by inputting the fixed phrase , wherein the dictionary is further associated with each of a plurality of words.
The position of the accent of each word is stored, and the optional word extracting step includes:
Extract the corresponding accent position from the dictionary
And extracting the word extracted in the arbitrary word extracting step
Classify according to position based on accents extracted along with words
Further comprising a word classifying step, wherein the fixed form sentence generating step includes:
Processing to generate fixed phrases by adding information about classification
A program for recording a program for executing
A computer-readable recording medium. 13. A computer-readable recording medium.
There are, included in the template text to be generated is represented by the desired word
The fixed phrase and the task to insert between the fixed words in the fixed phrase.
A fixed word input step of inputting the position of the meaning word, and the position of an arbitrary word to be inserted in the fixed sentence and the word to be inserted
Arbitrary word input step for inputting in association with the attribute of the arbitrary word
And the attribute of the arbitrary word input in the optional word input step.
Corresponding words, multiple words and their attributes are associated
Word extraction step to select and extract from the dictionary
At the position of an arbitrary word input in the fixed word input step,
Insert the words extracted in the optional word extraction step
And generating a fixed phrase by inputting the fixed phrase , wherein the dictionary is further associated with each of a plurality of words.
Information indicating a phonemic environment of each word is stored, and the optional word extracting step includes:
Information indicating the corresponding phonological environment is also extracted from the dictionary
The word extracted in the arbitrary word extraction step is added to the word.
Classify according to the extracted information indicating the phonetic environment
The method further includes a word classification step, wherein the fixed sentence generation step includes the step of:
Processing to generate fixed phrases by adding information about classification
A program for recording a program for executing
A computer-readable recording medium. 14. The word classification step, further comprising:
Fixed words input in the word input step and extracted words
The arbitrary word extraction according to the positional relationship with the arbitrary word corresponding to
Classifying the words extracted in the output step
14. The recording medium according to claim 13, wherein: 15. A computer-readable recording medium.
There are, included in the template text to be generated is represented by the desired word
The fixed phrase and the task to insert between the fixed words in the fixed phrase.
A fixed word input step of inputting the position of the meaning word, and the position of an arbitrary word to be inserted in the fixed sentence and the word to be inserted
Arbitrary word input step for inputting in association with the attribute of the arbitrary word
And the attribute of the arbitrary word input in the optional word input step.
Corresponding words, multiple words and their attributes are associated
Word extraction step to select and extract from the dictionary
At the position of an arbitrary word input in the fixed word input step,
Insert the words extracted in the optional word extraction step
And generating a fixed phrase by inputting the fixed phrase , wherein the dictionary is further associated with each of a plurality of words.
The information on the sound of each word is stored, and the arbitrary word extracting step includes:
Information on the corresponding sound is also extracted from the dictionary , and the word extracted in the arbitrary word extraction step is
Classify according to information about sounds extracted along with words
Further comprising a word classification step , wherein the fixed sentence generation step is performed by the word classification step.
Processing to generate fixed phrases by adding information about classification
A program for recording a program for executing
A computer-readable recording medium. 16. A compound used in the optional word extracting step.
Number of words and attributes of each word are stored in association with each other
12. The dictionary according to claim 11, further comprising:
Computer readable according to any one of the above items 15 to 15
Recording media.