JP2004037528A - Information processor and information processing method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To solve a problem wherein, according to the method of producing recognition vocabulary only by inputting a sentence containing the recognition vocabulary, it is impossible to fabricate a speaking dictionary for speech synthesis requiring accents and parts of speech and, also, it is difficult to edit the erroneous reading automatically attached to the recognition vocabulary. <P>SOLUTION: The orthographic information of a word is inputted (S211) or is edited (S213) and, on the basis of the inputted and edited orthographic information, the word information indicating reading, accent and part of speech is complemented (S214), the complemented word information is corrected as necessary (S215-S217) and the produced word information is outputted (S219). <P>COPYRIGHT: (C)2004,JPO

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to an information processing apparatus and method, for example, to information processing for efficiently editing and creating word information used in speech recognition and speech synthesis.
[0002]
[Prior art]
Due to recent improvements in device performance and software technology, systems that recognize voices uttered by users and use the commands as commands for the devices have been developed. Many speech recognition devices used in these systems need to register a recognizable vocabulary (recognized vocabulary) in advance. The registration of the recognition vocabulary is performed by text including text or utterance content (for example, katakana indicating reading of a word). In addition, when the recognition vocabulary is registered by voice, there is a disadvantage that the recognition rate is reduced if the registered user and the user are different. For this reason, at present, there are many devices for registering recognition vocabulary by text.
[0003]
Speech synthesis techniques for converting digitized text into speech information have also been put to practical use. Speech synthesis basically does not require the user to register words, but requires a vocabulary dictionary of vocabulary if the vocabulary unknown to the device is to be uttered accurately. The utterance dictionary may include, in addition to the vocabulary reading, accent information (where the accent goes up or down in the reading) and part of speech information of the word.
[0004]
[Problems to be solved by the invention]
To use speech recognition, a recognition vocabulary is required as described above. Further, in order to accurately utter words in speech synthesis, an utterance dictionary is necessary as described above. These data (hereinafter collectively referred to as a “word dictionary”) need only be created once when the system is built unless the recognized vocabulary and utterance content change, but if the dialog content of the system changes, the word The dictionary needs to be maintained.
[0005]
Japanese Patent Application Laid-Open No. 2002-41081 discloses a technique for generating a recognition vocabulary for speech recognition by receiving a character string as input, dividing the character string into constituent words, and automatically complementing reading information of each word. It has been disclosed. According to this technique, the recognition vocabulary is generated only by inputting a sentence including the recognition vocabulary, so that the load on the user is small. However, with this method, it is not possible to create an utterance dictionary for speech synthesis requiring accents, parts of speech, etc., and it is also difficult to edit erroneous readings automatically added to the recognized vocabulary.
[0006]
The present invention is to solve the above-mentioned problems individually or collectively, and aims to efficiently edit and create word information used in speech recognition and speech synthesis.
[0007]
Another object of the present invention is to enable correction of information automatically added to word information.
[0008]
[Means for solving the problems]
The present invention has the following configuration as one means for achieving the above object.
[0009]
An information processing apparatus according to the present invention includes an input / editing unit that inputs or edits notation information of a word, and a complementing unit that complements word information indicating the reading, accent, and part of speech of the word based on the input or edited notation information. And correction means for correcting the word information complemented by the interpolation means as necessary, and output means for outputting the created word information.
[0010]
The information processing method according to the present invention inputs or edits notation information of a word, and based on the input or edited notation information, complements word information indicating the reading, accent, and part of speech of the word, and complements the complemented word information. Is corrected as necessary, and the created word information is output.
[0011]
Preferably, the method further comprises setting a usage mode of the word information, and outputting the word information according to the usage mode.
[0012]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, an information processing apparatus for editing a word dictionary (hereinafter, referred to as a “word dictionary editing apparatus”) according to an embodiment of the present invention will be described in detail with reference to the drawings.
[0013]
[Constitution]
FIG. 1 is a block diagram showing a configuration example of a device for editing a word dictionary.
[0014]
A keyboard 104 and a pointing device 105 such as a mouse are connected as input devices to the word dictionary editing device 101, and a display 102 and a speaker 103 are connected as output devices for notifying a user of editing work and information of registered words. In addition, it is also possible to integrally configure these devices. The word dictionary editing device 101 can be configured as a dedicated device, but can also be realized by supplying software for executing word dictionary editing to a computer device (PC).
[0015]
Information (operation information) indicating an input operation of a word editor (hereinafter, referred to as “user”) input by the input device is interpreted by the operation information input unit 110 of the word dictionary editing device 101, and the operation information is a word. The data is sent to the data editing unit 108, and the word data editing unit 108 performs notation of each word, editing of reading information, setting of accents, and the like. Then, the edited word data is sent to the word data management unit 109 and stored in the word dictionary 113. The word dictionary 113 is stored in, for example, an area allocated to a hard disk or a semiconductor memory card, and an external speech recognition device or speech synthesis device transmits word data stored in the word dictionary 113 via a predetermined interface. And performs speech recognition and speech synthesis processing.
[0016]
When the word notation is edited by the word data editing unit 108, the word data automatic complementing unit 111 performs an automatic complementing process of automatically adding reading and accent information from the word notation. This automatic interpolation processing can save the user the trouble of inputting a reading or an accent corresponding to the notation, and the setting of the reading or accent by the user is mainly limited to an error in the automatic completion processing. Although the details will be described later, the automatic completion process is a process of searching the language dictionary 114 based on the notation of a word, and setting the pronunciation, accent, and part of speech stored in the language dictionary 114.
[0017]
The information output unit 106 uses the output device to feed back operation information of the user, and sequentially presents the word data managed by the word data management unit 109 to the user. Specifically, when the user inputs a command for confirming a word by voice output, the information output unit 106 performs a voice synthesis process using the selected word information (reading, accent, and part of speech) and the voice synthesis unit data 112. Is performed by the voice synthesizing unit 107, and the generated voice waveform is output to the speaker 103. Since the voice synthesis processing uses a known technique, a detailed description is omitted.
[0018]
Note that, similarly to the word dictionary 113, the language dictionary 114 and the speech synthesis segment data 112 are stored in, for example, an area allocated to a hard disk or a semiconductor memory card.
[0019]
[processing]
FIG. 2 is a flowchart illustrating an example of a word dictionary editing process. Hereinafter, a case will be described in which the word dictionary editing apparatus 101 is implemented in a PC or the like as dialog-based application software having a user interface (UI) screen as shown in FIG.
[0020]
When activated, the word dictionary editing device 101 reads user operation information (S201), and enters the following loop for performing processing corresponding to the read user operation information. In other words, each process is called based on an event generated by a user operation.
[0021]
When "Add new word" is instructed by a button or menu on the UI screen (S202), an empty record (information group) is created for a new word (S211). In the UI screen shown in FIG. 3, one line indicates one piece of word information, and adding a new word corresponds to a process of adding a line. Specifically, the ID of the new word is added to the ID column 402 of the UI screen by a numerical string, for example. Add. It is desirable that IDs are uniquely assigned to registered words.
[0022]
When a specific word displayed on the UI screen is selected and "delete word" is instructed by a button or a menu (S203), the record of the word information is deleted (S212).
[0023]
When a cell of the notation column 403 is selected by the mouse 105 or the like (S204), it is determined that "edit notation" has been instructed, a character string input via the keyboard 104 or the like is displayed in the selected cell, and the character The column is set as word notation information (S213), and the word information (reading, accent, and part of speech) is automatically complemented using the notation information (S214).
[0024]
When a cell of the reading column 404 is selected by the mouse 105 or the like (S205), it is determined that "reading correction process" has been designated, and a character string input via the keyboard 104 or the like is displayed in the selected cell. The character string is set as word reading information (S215).
[0025]
When the cell of the accent row 405 is selected by the mouse 105 or the like (S206), it is determined that “accent correction processing” has been designated, and a character string input via the keyboard 104 or the like is displayed in the selected cell. The character string is set as word accent information (S216). At this time, it is also possible to open another UI screen for setting an accent as shown in FIG. 4 and set an accent using a graphical user interface (GUI). In the example of the GUI shown in FIG. 4, one accent designation button 602 is assigned to each mora (unit of utterance) of word reading information, and an accent is designated according to the state of each button. The relationship between the state of the button and the level of the accent is determined, for example, as follows.
Button pressed… State with low accent button not pressed… High accent
When the cell of the part-of-speech sequence 406 is selected by the mouse 105 or the like (S207), it is determined that “part-of-speech correction processing” has been designated, and The displayed character string is displayed in the selected cell, and the part of speech name is set as the part of speech information of the word (S217).
[0027]
When a specific word (or a plurality of words) is selected and "voice confirmation" is instructed by a button or a menu (S208), a speech synthesis process and a speech output for causing the user to confirm the reading / accent of the word. Is performed (S218). When a plurality of words are selected, for example, the sounds of the words are output in order from the top of the UI screen. Also, as shown in FIG. 4, a voice confirmation button 603 may be provided in the GUI for accent setting so that the voice can be confirmed at the time of accent setting.
[0028]
When "file writing" is instructed by a button or a menu (S209), all words registered at present are output as a word dictionary 113 as a file (S219). At this time, the format of the word dictionary 113 to be output may be selectable, or may be output as a speech synthesis dictionary for speech synthesis or recognition vocabulary data for speech recognition (word information according to a recognition grammar for speech recognition). May be written to a file. FIG. 5 shows an example of a UI screen for setting the storage destination of the file. In this UI screen, the format of the file to be written can be selected by the data format selection list 702. FIG. 6 is a diagram showing an example of the word dictionary 113 output as a file. The output destination of the file is a hard disk or a semiconductor memory card of the device 101. By providing a predetermined interface, a serial bus such as IEEE1394 or USB (Universal Serial Bus) or a wireless interface such as Bluetooth or IrDA is provided. It is also possible to output to a memory of a speech recognition device or a speech synthesis device via the face.
[0029]
When "end" is instructed by the button or the menu (S210), the process ends. At this time, if there is word information that has not been output to a file, a dialog prompting the user to output a file may be presented.
[0030]
[Complete word information]
FIG. 7 is a flowchart illustrating an example of a word information auto-completion process. This process is executed when a word notation is newly input or edited.
[0031]
First, the language dictionary 114 is searched (dictionary lookup) using the word notation (S301), and it is determined whether a word corresponding to the word notation is stored in the language dictionary 114 (S302). If the corresponding word is stored, the corresponding word information (reading, accent, and part of speech) is extracted from the language dictionary 114 (S306), and the extracted word information is input into the reading or accent corresponding to the input or edited word notation. And part of speech (S307).
[0032]
If no word corresponding to the input word notation is found, the word notation is subdivided and individual notations are searched (S303). The word notation may be divided for each morpheme. However, if the search is not performed even if it is divided for each morpheme, it is divided for each character. Then, the pronunciations corresponding to each of the divided notations are connected in the notation order before the division (S304), the accent is set to a default value (for example, type 0), the part of speech is set to a noun (S305), and the input or edited word notation is used. Are set to the reading, accent, and part of speech (S307).
[0033]
For example, when reading from the notation “YRP Nobi” and automatically complementing accents and parts of speech, the following processing is performed.
[0034]
First, if “YRP Nobi” is looked up in the dictionary from the language dictionary 114 and “YRP Nobi” is not searched, “YRP Nobi” is subjected to morphological analysis. As a result of the analysis, the data is divided into “Y”, “R”, “P”, and “Nobi”. These are connected to one reading, "Wire Pinovi," and the reading is made. The accent is set to type 0 and the part of speech is set to noun.
[0035]
[Key operation]
FIG. 8 is a diagram showing a list of key assignments for key operations that can be performed by the word information complementing apparatus 101. As shown in FIG. 8, the apparatus 101 can issue various command instructions not only from buttons and menus on the UI screen but also from key operations.
[0036]
[Modification]
FIG. 9 is a diagram showing a modification of the UI screen shown in FIG. In this example, a column 902 for specifying a word use destination is added to the UI screen. In each cell of the column 902, "speech synthesis", "speech recognition", and "speech synthesis / recognition" indicating a destination that can be selected from a list box or the like can be input.
[0037]
The usage destination information is information indicating whether the word is used as recognition vocabulary data for speech recognition, as an utterance dictionary for speech synthesis, or both, and the set value is used to set the word dictionary editing device. The processing of 101 can be changed. For example, in step S219, when outputting recognized vocabulary data, only words for which "speech recognition" and "both speech synthesis / recognition" are set are output, and similarly, when outputting a speech dictionary, "speech synthesis" and "voice synthesis" are output. Only words for which both "speech synthesis and recognition" are set can be output.
[0038]
This makes it possible to simultaneously manage the vocabularies for speech recognition and speech synthesis on one UI screen shown in FIG.
[0039]
FIG. 10 is a diagram showing a modification of the UI screen shown in FIG. In this example, instead of instructing the height of the accent with a button, the slider bar 1002 can be used to give a finer instruction. In addition, a curve (accent image) 1004 connecting the sliders of the slider bar 1002 makes it possible to graphically display the accent set for the word.
[0040]
In the above description, it has been described that the word information is composed of notation, reading, accent, and part of speech, but other information can be included. For example, an appearance probability, a degree of importance, a name of a process to be executed when speech recognition is performed, a waveform dictionary name for speech synthesis, and parameters of speed and pitch for speech synthesis can be included.
[0041]
Of course, the word dictionary editing apparatus 101 can be applied to languages other than Japanese.
[0042]
[Other embodiments]
The present invention can be applied to a system including a plurality of devices (for example, a host computer, an interface device, a reader, a printer, etc.), but can be applied to a device including one device (for example, a copier, a facsimile machine, etc.). May be applied.
[0043]
Further, an object of the present invention is to supply a storage medium (or a recording medium) in which a program code of software for realizing the functions of the above-described embodiments is recorded to a system or an apparatus, and a computer (or a CPU or a CPU) of the system or the apparatus. Needless to say, the present invention can also be achieved by an MPU) reading and executing a program code stored in a storage medium. In this case, the program code itself read from the storage medium realizes the function of the above-described embodiment, and the storage medium storing the program code constitutes the present invention. When the computer executes the readout program code, not only the functions of the above-described embodiments are realized, but also an operating system (OS) running on the computer based on the instruction of the program code. It goes without saying that a part or all of the actual processing is performed and the functions of the above-described embodiments are realized by the processing.
[0044]
Further, after the program code read from the storage medium is written into a memory provided in a function expansion card inserted into the computer or a function expansion unit connected to the computer, the function is executed based on the instruction of the program code. It goes without saying that the CPU included in the expansion card or the function expansion unit performs part or all of the actual processing, and the processing realizes the functions of the above-described embodiments.
[0045]
When the present invention is applied to the storage medium, the storage medium stores program codes corresponding to the flowcharts described above.
[0046]
【The invention's effect】
As described above, according to the present invention,
[0047]
In the present word dictionary editing device, when a word notation is input, its reading, accent, part of speech, and the like are automatically complemented. Only when the complemented word information is wrong, it is sufficient to edit the contents thereof, so that it is possible to greatly reduce the trouble of the user.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a configuration example of a device for editing a word dictionary;
FIG. 2 is a flowchart showing an example of a word dictionary editing process;
FIG. 3 is a diagram showing an example of a user interface screen.
FIG. 4 is a diagram showing an example of a user interface screen for setting an accent.
FIG. 5 is an example of a user interface screen for setting a save destination of a file;
FIG. 6 is a diagram showing an example of a word dictionary output as a file.
FIG. 7 is a flowchart showing an example of an automatic completion process of word information;
FIG. 8 is a diagram showing a list of key assignments for key operations that can be performed by the word information complementing device;
FIG. 9 is a view showing a modification of the user interface screen shown in FIG. 3;
FIG. 10 is a diagram showing a modification of the user interface screen shown in FIG.

Claims (11)

Input editing means for inputting or editing word notation information;
A complementing means for complementing word information indicating the reading, accent, and part of speech of the word based on the input or edited written information;
Correction means for correcting the word information complemented by the interpolation means, if necessary,
An output unit that outputs the created word information. Further, there is a setting means for setting a use form of the word information,
The information processing apparatus according to claim 1, wherein the output unit outputs the word information according to the use mode. The information processing apparatus according to claim 1, wherein the output unit outputs the word information as information according to a recognition grammar for voice recognition. The information processing apparatus according to claim 1, wherein the output unit outputs the word information as an utterance dictionary for speech synthesis. The information processing apparatus according to any one of claims 1 to 4, further comprising a confirmation unit configured to confirm reading of the selected word information by voice. Furthermore, it has accent setting means for setting accent information for each mora of the word,
The information processing apparatus according to claim 1, wherein the correction unit reflects the accent information set by the accent setting unit in the word information. 7. The information processing apparatus according to claim 6, wherein the accent information is set by a slider bar displayed in a graphic form. Enter or edit the word ’s notation information,
Based on the input or edited written information, complement the word information indicating the reading, accent and part of speech of the word,
Correct the complemented word information as necessary,
An information processing method characterized by outputting the created word information. Further, a use mode of the word information is set,
The information processing method according to claim 8, wherein the word information is output according to the use mode. A program for controlling an information processing apparatus to execute the information processing according to claim 8. A recording medium on which the program according to claim 10 is recorded.

Images