JPH0232398A - Accent dictionary generating device - Google Patents

Abstract

PURPOSE:To facilitate the registration of an accent dictionary by setting an automatic utterance mode and a confirmation mode in accordance with whether accent information is inputted or not at the time of requesting a voice output. CONSTITUTION:A control part 3 sets automatically an automatic utterance mode for bringing word information whose accents are different to voice synthesizing output in order, and a confirmation mode for confirming a correct accent by the voice synthesizing output of the word information of an input accent type, in accordance with whether information of an accent type is inputted already or not, when a voice output request is given from a keyboard 1. In such a way, a dictionary of accent information related to the word information can be registered simply, efficiently and exactly.

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Industrial Field of Application) The present invention relates to an accent dictionary creation device that can easily and accurately create an accent dictionary regarding word information and register it in the dictionary.

(Conventional, conventional technology) Recently, character code strings indicating word information are manually generated, and this input character code string is analyzed to obtain its phonological parameter series and prosodic parameter series. Various speech synthesis devices have been developed that generate synthesized speech according to sequences and therefore predetermined rules. This type of speech synthesis device using rule synthesis has the advantage that it can easily generate synthesized speech representing any word or sentence, compared to a conventional speech synthesis device using a recording/editing method. For this reason, it is attracting attention as an important technology for realizing highly natural man-machine interfaces in combination with voice recognition technology.

However, in such rule-based synthesis of speech, it is important to accurately reproduce the word-specific accent of each word in order to obtain highly natural synthesized speech.

Therefore, conventionally, for each simple word or compound word made up of a combination of multiple words, information on the accent type unique to that word is registered as an accent dictionary, and during speech synthesis, the word information indicated by the input character code string is used. The accent dictionary is searched to find an accent unique to the input character code string.

However, accurately registering accent-type information regarding word information in a dictionary in advance requires a large processing load. Moreover, even for people with a certain level of experience, it is extremely difficult to always accurately register accent type information in a dictionary.

Conventionally, it has been considered that when word information to be registered in a dictionary is specified, synthesized speech with different accents for the word information is sequentially generated and output to encourage selection of the correct accent (accent type). On the other hand, it is being considered to actually synthesize word information into speech according to accent type information registered in a dictionary and prompt the user to confirm the information.

However, these functions have advantages and disadvantages depending on the proficiency level of the person registering the accent dictionary.For example, for beginners, it is difficult to select the correct accent (accent type) by sequentially generating and outputting synthesized voices with different accents. However, there is a problem in that a confirmation function using voice synthesis is easier for experts to use. Therefore, even if these functions were provided, there was a problem in that it was very difficult to select and specify the output form (mode) of the synthesized speech and use these functions properly.

(Problem to be solved by the invention) In this way, in the past, there was a function to sequentially generate and output synthesized speech with different accents and select the correct accent (accent type), and information on accent types registered in a dictionary. Accordingly, there have been problems such as that it is difficult to easily and accurately register accent information related to word information in a dictionary by properly using the function of synthesizing word information into speech and confirming it.

The present invention has been made in consideration of these circumstances, and its purpose is to provide an accent dictionary creation device that can easily and accurately register accent information related to word information in a dictionary. It is in.

[Structure of the Invention] (Means for Solving the Problems) The present invention relates to an accent dictionary creation device that registers accent type information with respect to word information in a dictionary and creates an accent dictionary regarding the word information. When a speech output request is given before, the word information is automatically synthesized and output in order with a different accent type, and when a speech output request is given after the accent type information is input, the input The present invention is characterized in that the word information is synthesized into speech and output according to the accent type information and the pronunciation of the word information.

In other words, when a voice output request is given, it is determined whether accent type information has already been input, and if the accent type information has not been input, the automatic voice mode is set and the accent is changed. The system is characterized in that the synthesized speech is output in sequence to prompt selection, and if the accent type information has been input, a confirmation mode is set to prompt confirmation of the human accent type information. .

(Function) According to the present invention, there is an automatic voice mode in which word information with different accents is sequentially synthesized and outputted depending on whether or not accent type information has already been input when a voice output request is given; Since the input accent type is automatically set to a confirmation mode to check the correctness of the accent by speech synthesis output of word information, you can register accent information related to word information in the dictionary without having to be aware of the selection settings of these modes. can be performed easily, efficiently, and accurately.

In other words, if you are a beginner, by instructing a voice output request, you can sequentially obtain synthesized voices of word information with different accents, so by selecting and instructing the synthesized voice with the correct accent from among them, you can easily , and can accurately register accent information in the dictionary. Furthermore, by inputting accent type information and instructing a voice output request, an expert can easily check the accuracy of the input accent information and register it in the dictionary.

In other words, by simply issuing a voice output request, it is possible to effectively select the correct accent using the automatic voice mode and confirming the correctness of the accent using the confirmation mode.

(Embodiment) An embodiment of the accent dictionary creation device according to the present invention will be described below with reference to the drawings.

FIG. 1 shows the concept of processing in the apparatus of the embodiment, and FIG. 2 is a schematic diagram of the main parts of the apparatus of the embodiment. This embodiment device is incorporated as a part of a speech synthesis device using a rule synthesis method, for example, and as shown in FIG. 2, word information and accent type information.

A keyboard 1 for inputting command codes, etc., a control unit 3 for registering accent information in the accent dictionary memory 2 according to information input from the keyboard 1, and displaying various messages during the accent information registration process. It is configured with a display 4 for

In addition, the accent type table 5 registers information on accent types according to the number of moras in the word information, and the speech rule synthesizer 6 registers the word information (character code string) given from the control unit 3 and the accent type information. The synthesized speech of the word information is generated by rule synthesis according to the information.

The control unit 3 basically controls entry words of word information to be registered in the dictionary from the keyboard 1 to the accent dictionary memory 2;
Information on the pronunciation 1 part of speech is input, and accent type information for specifying the accent unique to the word is input manually, and this is registered in the accent dictionary memory 2 to create an accent dictionary. The accent dictionary created in this way is
It is used for rule-based synthesis of synthesized speech according to input character code strings.

By the way, when creating such an accent dictionary, the control unit 3, as shown in the control procedure in FIG. Step B) determines whether or not accent type information has already been input.According to the determination result, if accent type information has not been input, automatic voice mode is set (Step C), and if accent type information has already been input. If so, the confirmation mode is set (step D).

Therefore, when setting the automatic voice mode, the control unit 3 sequentially obtains accent type information according to the number of moras from the accent type table 5 for the word information input from the keyboard 1, and together with this accent type information, the word information give the reading information to the speech rule synthesizer 6,
Synthesized speech of the word information with different accents is sequentially generated and output (step E). Selection of the correct accent (accent type) is encouraged by the synthesized speech that is sequentially synthesized and output with different accents according to rules.

In addition, when the confirmation mode is set, the control unit 3 provides the already input accent type information and the pronunciation of the word information to the speech rule synthesizer 6 to generate actual synthesized speech according to the accent type information. Output (step F). The synthesized speech that is actually synthesized and output according to the rules prompts confirmation as to whether or not the correct accent type has been assigned to the word information.

As described above, the accent information (accent type) to be registered in the accent dictionary is selected and confirmed, and the accent dictionary is created while outputting synthesized speech that is actually synthesized according to the accent type information. In this device, the present invention is characterized by automatically changing the output mode of the synthesized speech depending on whether or not accent type information has already been given when the above-mentioned voice output request instruction is given. mode or confirmation mode. In other words, instead of leaving the setting of the mode to the user (accent dictionary registrant), it is automatically set according to the input state of the information (accent type information), and the synthesis is performed by sequentially changing the accents to select the accent type information. It is characterized by automatically outputting voice or synthesized voice to confirm the correctness of the input accent type.

FIG. 3 shows an example of a processing procedure in the control unit 3 that realizes such a processing function. First, the entry word, pronunciation, and part of speech of the word information to be registered in the accent dictionary memory 2 are prompted. (step a). By this input processing.

For example, when registering accent information for the word "Tokyo" in a dictionary, the following information is entered using the keyboard: headword; Tokyo reading; Tokyo part of speech - noun (place name).

Thereafter, the control parameter X for the dictionary registration process is initialized to [1] (step b), and the type of key manual information is determined after waiting for information input from the keyboard l (step C). This key manual determination process is performed by checking whether the key information input from the keyboard 1 is numeric information that specifies the accent type, or a special symbol (for example, #) that indicates a voice output request. .

However, if the key input indicates a voice output request, since accent type information using numerical information has not been input at this point, this is a request to output synthesized voice in a different accent using automatic voice mode. , and sets the automatic voice mode. By setting this automatic voice mode, the output of synthesized speech with different accents is controlled in the following manner.

That is, first, the accent type table 5 is referred to, and the Xth accent type y corresponding to the number of moras of the pronunciation of the word information to be registered is determined (step d). Then, it is confirmed that the y-th character of the pronunciation indicated by the accent type y is not a long sound, a cursive sound, or a consonant sound (step e), and the information [y] of the accent type y is displayed on the display 4 (step f). Then, the pronunciation information and the accent type y are given to the speech rule synthesizer 6 to generate and output synthesized speech with the accent indicated by the accent type y (step g). Then, the control parameter
Increment (+1) L (step h), confirm that the parameter value X does not exceed the number of moras, and repeat the processing from step d described above (step i).

Furthermore, if the y-th character in the pronunciation indicated by accent type y is a long sound, a cursive sound, or a consonant, it is impossible for an accent to occur at that position, so the generation of the synthesized speech shown by the above accent type y is as follows. Omitted.

When registering the accent dictionary for "Tokyo" as described above using the above processing procedure, ■ "Lo-Kiyo"
■Synthesized voices with different accents such as "To-shita-yo" are output in sequence. At the same time, information on the accent type y for the output synthesized speech is displayed on the display 4. Therefore, the user (
The dictionary registrant) listens to the synthesized speech in the accent specified by the accent type y while checking the information on the accent type y on the display 4. Furthermore, when registering the word information indicating the person's name "Takahashi" in the accent dictionary, synthesized voices with different accents such as ■ "7 Kahashi" ■ "Tagahashi" ■ "Takaneshi" ■ "Takahafu" will be output in order.

When the output of a series of synthesized voices with different accents is completed, the processing procedure returns to step b described above, and the system is again in a state of waiting for human input.

If a voice output request is issued again in this state, the synthesized voice with different accents as described above will be repeatedly output, but in general, a number is used to specify the accent type y according to the listening results. Information is input. In addition, if the user (dictionary registrant) is an expert, do not output synthesized speech with different accents as described above (directly input numerical information that specifies accent type y). In such a case, since input of accent type information is confirmed by this key human power determination process (step C), the processing form for subsequent voice output requests is changed.

In this state, it waits for a key input from the keyboard l, and determines the type of input key (step j).

In this determination process, it is determined whether the type of the input key is the aforementioned key that indicates a voice output request (e.g., #) or the key that instructs dictionary registration of the accent information (e.g., *).
This is done by determining whether

However, when a voice output request is issued, since the information for accent type y has already been input, the confirmation mode is set, and the input pronunciation information and the information for accent type y that is being input are output. The speech rule synthesizer 6 generates and outputs synthesized speech with an accent indicated by the accent type y (step k). By outputting this synthesized speech, it is confirmed whether the set accent type indicates the correct accent specific to the word information.

Thereafter, the system waits again for key input and determines the type of input key (step m). In this judgment, whether the key type is a number and the accent type information has been changed,
This is also performed by determining whether a voice output request has been given again or whether a dictionary registration instruction has been given. Then, according to this determination result, the process from step j or the process from step k described above is repeatedly executed, the instruction for dictionary registration is confirmed, and the information on the accent way y set at that time is used as the accent regarding the word information. When the word information is registered as a dictionary in the accent dictionary memory 2 (step n), the accent dictionary creation process (dictionary registration) for the word information is completed.

According to this device configured in this way, by simply giving a voice output request without setting the synthesized voice output mode one by one, automatic voice production is performed depending on whether or not accent type y information has already been input. mode or confirmation mode can be automatically set. When in automatic voice mode, synthesized voices with different accents are output in order to prompt the selection, and when setting confirmation mode, synthesized voices with accents according to the input accent type y are outputted to confirm the selection. It becomes possible.

As a result, it is possible to easily and effectively use these two types of synthesized speech output functions to accurately create an accent dictionary regarding word information and register it in the dictionary.

Note that the present invention is not limited to the embodiments described above. For example, when outputting synthesized speech with different accents using automatic voice mode, if information specifying the accent type y is manually generated, the synthesized speech with different accents can be stopped and checked, or registered mode (processing from step j) You may move to In addition, after the input accent information y is canceled in the confirmation or registration mode,
When a voice output request is made, it is also possible to return to the automatic voice mode described above. In addition, the present invention can be implemented with various modifications without departing from the gist thereof.

[Effects of the Invention] As explained above, according to the present invention, the automatic voice mode and the confirmation mode are automatically switched depending on whether or not accent information is input at the time of a voice output request. This makes it possible to facilitate (simplify) accent dictionary registration, and has great practical effects such as making it possible to easily and accurately register and set the accent information. It will be done.

[Brief explanation of the drawing]

The figures show an embodiment of the accent dictionary creation device according to the present invention. FIG. 1 is a diagram showing the processing concept in the embodiment device, FIG. 2 is a schematic configuration diagram of the embodiment device, and FIG. 3 is an implementation example. FIG. 3 is a diagram showing an example of a specific processing procedure in the example device. l...keyboard, 2...accent dictionary memory,
3... Control unit, 4... Display, 5... Accent type table, 6... Speech rule synthesizer, A...
Voice output request, B: Judgment processing, C: Automatic voice mode, D: Confirmation mode.

Claims (1)

[Claims] In an accent dictionary creation device that registers accent type information for word information in a dictionary and creates an accent dictionary regarding the word information, when a voice output request is given before accent type information is input, an accent that changes the accent type is generated. means for sequentially synthesizing and outputting word information in order, and when a voice output request is given after inputting accent type information, synthesizing and outputting the word information according to the input accent type information and the pronunciation of the word information. An accent dictionary creation device characterized by comprising means.