JP2014197117A - Speech synthesizer and language dictionary registration method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a speech synthesizer capable of generating a proper synthesized speech signal even for a text including a word having a plurality of different readings or rhythms.SOLUTION: A speech synthesizer 1 has: an input unit 2 which acquires text data; a storage unit 3 in which at least a Chinese character and Kana character notation of a word, and an intermediate notation including the reading of the word and a rhythm symbol showing the rhythm are registered, and which stores a language dictionary which is used for generating the intermediate notation from the text data; a registration range setting unit 25 which expands a registration range to be registered in the language dictionary in such a manner that, when a word included in a correction range in the after-correction intermediate notation in which the intermediate notation is corrected is a predetermined part of speech, words of the same part of speech as the word which continues before/after the word included in the correction range are also included; and a registration unit 26 which registers at least Chinese character and Kana character notation and the intermediate notation of the word in the language dictionary, with a part included in the registration range in the after-correction intermediate notation as one word.

Description

  The present invention relates to a speech synthesizer that synthesizes a speech signal from text data, for example, and a language dictionary registration method used in the speech synthesizer.

  In recent years, speech synthesis technology for automatically synthesizing speech has been developed. Since speech synthesis technology has the advantage that it can create desired speech in a short time, such speech synthesis technology has been adopted in applications that have used pre-recorded speech by professional narrators. There are also things. In particular, in an application in which information provided at a short time interval is updated, such as a guidance broadcast in a commercial facility, a highway radio, a highway telephone, or a weather forecast broadcast, the speech synthesis technology having the above-described advantages is useful.

  In order to generate a speech signal to be synthesized, text data mixed with kanji characters is input to the speech synthesizer via, for example, a keyboard. The speech synthesizer performs language processing such as morphological analysis or dependency analysis on the text data using a language dictionary in which word readings are registered. Then, the speech synthesizer generates morpheme information representing the morpheme-based reading notation and intermediate notation with the prosody symbol representing the prosody such as the accent position, the strength of the accent, or the size of the inflection on the morpheme information by the language processing. . Then, the speech synthesizer generates a synthesized speech signal based on the intermediate notation.

  There are several ways to read kanji, and how to read kanji included in the word differs depending on the word. In addition, since words used in daily life change day by day, it is virtually impossible to register all words in the language dictionary in advance, and the input text data includes words that are not registered in the language dictionary. Sometimes it is. Therefore, the intermediate notation obtained as a result of language processing may not be accurate. In such a case, in order to obtain a correct synthesized speech signal, the user needs to manually correct the intermediate notation. Since such a correction work is complicated for the user, it is preferable that such a correction work does not occur as much as possible. Therefore, when the input text data reaches the end of the input text data, the unknown words extracted so far are collectively transmitted to the user, and the unknown word and the information about the unknown word input by the user are registered in the word dictionary. It has been proposed (see, for example, Patent Document 1). Further, if there is a part that matches the stored replacement condition in the first candidate language analysis result in the language analysis result, the matching part is replaced with replacement information corresponding to the replacement condition, and a new language analysis is performed. A technique for generating a result has been proposed (see, for example, Patent Document 2). In this technique, when the same language analysis result other than the first candidate exists as a language analysis result other than the first candidate, a synthesized speech is generated based on the new language analysis result.

  In the above technique, when a word registered as an unknown word or a word matching the replacement condition is included in the text data, an intermediate notation is generated according to the registered word or replacement condition. However, the word itself may be read differently depending on sentences before and after the word, or may be uttered with different prosody. In such a case, the above technique may not necessarily generate an appropriate intermediate notation. Therefore, a technique has been proposed in which, when editing an intermediate language of a document to be read out, specification of a correction target phrase and correction reflection condition specification are included in the correction instruction (for example, Patent Document 3). reference). In this technique, for example, a related word / phrase is specified as a condition for reflecting the correction.

JP 7-244491 A Japanese Patent Laid-Open No. 10-312377 JP 2006-30326 A

  However, even the technique disclosed in Patent Document 3 may not generate an appropriate intermediate language unless the conditions for reflecting the correction are appropriately specified. It is difficult to determine the conditions for reflecting the reflection in consideration of all cases in advance.

  Accordingly, an object of one aspect of the present specification is to provide a speech synthesizer that can generate an appropriate synthesized speech signal even for text including words having a plurality of different readings or prosody.

  According to one embodiment, there is provided a speech synthesizer that generates a synthesized speech signal based on an intermediate notation that includes a prosodic symbol representing a reading and prosody of text data generated from text data. This speech synthesizer is used to generate an intermediate notation from text data by registering an input unit for acquiring text data and an intermediate notation including at least a kanji kana notation of the word and a reading and prosodic symbol of the word. A storage unit for storing a language dictionary and a word included in the corrected range in the corrected intermediate notation in which the intermediate notation is corrected, if the word included in the corrected range is a word that is continuous before and after the word included in the corrected range The registration range setting unit that expands the registration range to be registered in the language dictionary so as to include up to the word of the same part of speech as the part of speech, and the portion included in the registration range in the corrected intermediate notation as one word, at least the word A registration unit for registering kanji kana notation and intermediate notation in a language dictionary;

The objects and advantages of the invention will be realized and attained by means of the elements and combinations particularly pointed out in the appended claims.
It should be understood that both the foregoing general description and the following detailed description are exemplary and explanatory and are not restrictive of the invention as claimed.

  The speech synthesizer disclosed in the present specification can generate an appropriate synthesized speech signal even for text including words having a plurality of different readings or prosody.

It is a figure which shows an example of the intermediate | middle description produced | generated accidentally with respect to text data. It is a schematic block diagram of the speech synthesizer by one Embodiment. It is a functional block diagram of the process part which the speech synthesizer by one Embodiment has. It is a figure which shows an example of the intermediate notation output by the language processing part, and the intermediate notation corrected by the user. It is a figure showing the morpheme reading notation before a user correction corresponding to the intermediate notation of FIG. 4, and the morpheme reading notation after a user correction. It is a figure which shows another example of the morpheme reading notation before a user corrects, and the morpheme reading notation after a user corrects. It is an operation | movement flowchart of a morpheme information setting process. It is a figure which shows an example of the relationship between each intermediate notation before and after user correction, and the morpheme reading notation after user correction. It is a figure which shows an example of the relationship between each intermediate description before and after user correction, and the correction range set. It is a figure which shows an example of the relationship between the intermediate description before and behind user correction, and the registration range. It is a figure which shows the operation | movement flowchart of a registration range setting process. It is an operation | movement flowchart of a dictionary registration process. It is a figure which shows an example of the word registered into the word registered into the user dictionary and the intermediate notation dictionary. It is an operation | movement flowchart of a dictionary selection process.

Hereinafter, speech synthesis apparatuses according to various embodiments will be described with reference to the drawings.
First, an example of intermediate notation that is erroneously generated for text data input as a speech synthesis target will be described with reference to FIG.

  It is assumed that a kanji kana mixed sentence 100 “Beautiful in the middle of this yokan ...” and a kanji kana mixed sentence 110 “ingenuity and ingenuity” are input as text data. Then, an intermediate notation 101 “Yokansa Ichuwa ...” is generated for the kanji kana mixed sentence 100, and an intermediate notation 111 “so-iku 'fu” is generated for the kanji kana mixed sentence 110. . The symbol “′” in the intermediate notation is a prosodic symbol representing “strong accent”. Since the prosodic symbols themselves are not defined in the standard, symbols other than the prosodic symbols used in this specification may be used as prosodic symbols representing specific prosody.

  In this example, in the intermediate notation 101, the portion corresponding to “middle” is mistakenly written as “situ”. For this reason, as shown in the intermediate notation 102 corrected by the user, “same” is corrected as “mode”. Here, if the word “middle” is registered in the language dictionary, the speech synthesizer automatically extracts the correction range using a technique such as matching based on dynamic programming, for example. The range is “middle” which is a word registered in the language dictionary. Therefore, when the correction made by the user is reflected in the language dictionary, when text data including the word “middle” is input thereafter, the intermediate notation of the word “middle” is always “monaka”. There is a risk. For example, in a sentence such as “in the middle of play ...”, the intermediate notation of the word “middle” should be “sa’ stew ”, not“ mo ”. As described above, in a word having a plurality of readings, the correction of reading by the user may be reflected, which may lead to generation of an erroneous intermediate display. In order to prevent such an error, the user needs to register “in the middle of the sheep” itself as a compound noun in the language dictionary. However, it is difficult for a user who has no specialized knowledge about speech synthesis to make such a determination appropriately. Further, even when the user has specialized knowledge related to speech synthesis, the setting of a word to be registered in the language dictionary may be wrong.

  In the intermediate notation 111, although the reading is correct, the positions of the accent and the break are incorrect. Therefore, as shown in the intermediate notation 112, it is assumed that the user corrects the position of the accent, adds a prosodic symbol “=” representing a delimiter, and corrects it to “So’-i = Kufu”. At this time, if the corrected part is only the prosodic symbol, there is no change in the reading of each word before and after the correction, so that the speech synthesizer may not be able to specify the range to be corrected. In this case, the range to be corrected by the user may be designated as the compound noun “creativeness”. Then, in order to register the compound noun in the language dictionary, the part of speech of “ingenuity” is also designated. Here, the word “creativity” is a common noun, and “devise” is a sabot noun, but since the word “devise” is located at the end of the compound noun, the user can also use the part of speech Should be. However, a user who does not have such specialized knowledge may not know the exact part of speech of the compound noun “creativity” and may end up using the compound noun “creativity” as a common noun. Then, in other text data, even if the compound noun “creative device” is included with the verb “do”, as in “creative device”, “creative device” is not registered as a sabot noun. Therefore, it is not adopted in the connection with the sub-variant “Sei”, which is easy to connect with the sub-noun, and as a result, the common noun “creativity” and the sub-noun “ingenuity” are adopted, and the accent by the user Corrections such as position will not be reflected in the intermediate notation.

  In order to solve the above problems, the speech synthesizer should appropriately determine the part of speech and the word registration range involved in the correction after optimizing the range in which the user has made the correction. Is required. Therefore, the speech synthesizer according to the present embodiment compares the intermediate notation before and after the correction by the user, identifies a matching part and a mismatching part in morpheme units, and based on the matching / mismatching information, Set part-of-speech information for each morpheme included in the intermediate notation. Then, the speech synthesizer sets the entire part of speech including the inconsistent part as the entire correction range by matching the intermediate notation before and after the correction by the user for each part of speech. Then, if the part of speech that is the correction range is a noun, this speech synthesizer sets the correction range and the entire sequence of nouns continuous before and after the correction range as a word registration range that reflects the user's correction. Also, in this speech synthesizer, if the part of speech that is the correction range is a stem of an independence word, the range from the stem to the last inflection ending of the same part of speech is set as a word registration range that reflects the user's correction.

  FIG. 2 is a schematic configuration diagram of a speech synthesizer according to one embodiment. In the present embodiment, the speech synthesizer 1 includes an input unit 2, a storage unit 3, a processing unit 4, and an output unit 5.

The input unit 2 obtains text data that is an original text of synthesized speech and is a kanji-kana mixed text. For this purpose, the input unit 2 includes, for example, a keyboard. The input unit 2 includes a pointing device such as a mouse and a display that displays characters or numerical values to be input, which are instructed by the pointing device. Alternatively, the input unit 2 may have a touch panel display.
Furthermore, the input unit 2 may acquire text data from another device connected to the speech synthesizer 1 via a communication network. In this case, the input unit 2 includes an interface circuit for connecting the speech synthesizer 1 to a communication network.
The input unit 2 passes the input text data to the processing unit 4.

The storage unit 3 includes, for example, at least one of a semiconductor memory circuit, a magnetic storage device, and an optical storage device. The storage unit 3 stores various computer programs used in the processing unit 4 and various data used for speech synthesis processing.
The storage unit 3 stores, for example, a prosodic model and a speech waveform dictionary as data used for speech synthesis processing. Further, the storage unit 3 stores, as data used for language processing, a language dictionary that stores kanji kana notation, intermediate notation, part of speech, and utilization form of the words that are supposed to appear in the text data. To do. Furthermore, the memory | storage part 3 memorize | stores the user dictionary which stored the kanji kana notation of the word, intermediate notation, a part of speech, a utilization form, etc. about the word registered by the user. A user dictionary is also an example of a language dictionary.

The output unit 5 outputs the synthesized voice signal received from the processing unit 4 to the speaker 6. For this purpose, the output unit 5 includes, for example, an audio interface circuit for connecting the speaker 6 to the speech synthesizer 1.
The output unit 5 may output the synthesized speech signal to another device connected to the speech synthesizer 1 via the communication network. In this case, the output unit 5 includes an interface circuit for connecting to the speech synthesizer 1 to the communication network. In addition, when the input part 2 also acquires text data via a communication network, the input part 2 and the output part 5 may be integrated.

The processing unit 4 includes one or a plurality of processors, a memory circuit, and a peripheral circuit. Then, the processing unit 4 creates a synthesized speech signal based on the input text data.
FIG. 3 is a functional block diagram of the processing unit 4. The processing unit 4 includes a language processing unit 10, a speech synthesis unit 11, and a dictionary registration unit 12.
Each of these units included in the processing unit 4 is, for example, a functional module realized by a computer program that operates on a processor included in the processing unit 4. Or these each part which the process part 4 has may be mounted in the speech synthesizer 1 as one integrated circuit which implement | achieves the function of each part.

  The language processing unit 10 generates a morpheme reading notation from the input text data that is a kanji mixed sentence, and specifies each morpheme information included in the text data. Furthermore, the language processing unit 10 specifies intermediate notation and information on each part of speech included in the text data from the input text data. Here, the morpheme reading notation represents the reading notation in units of morphemes, and is expressed in katakana, for example. The intermediate notation is obtained by adding prosodic symbols representing prosody to the morpheme reading notation. The prosodic symbols include, for example, symbols that represent “accent position”, “accent strength”, “pitch pitch”, “inflection magnitude”, “speech speed”, “volume level”, and “separation”. Accordingly, the intermediate notation excluding the prosodic symbols matches the morpheme reading notation. In addition, the information of each part of speech in the intermediate notation excluding the prosodic symbols corresponds to the morpheme information on a one-to-one basis. That is, morpheme reading notation and morpheme information are extracted from the intermediate notation and part-of-speech information.

  The language processing unit 10 reads the language dictionary and the user dictionary stored in the storage unit 3 in order to generate morpheme reading notation and intermediate notation from the input text data. Then, the language processing unit 10 performs morphological analysis and dependency analysis on the text data using the language dictionary and the user dictionary, for example, the order and reading of each word appearing in the text data, and the position of the accent And the position of the break is determined. At that time, when there are words registered in both the language dictionary and the user dictionary in the text data, the language processing unit 10 may preferentially use the words registered in the user dictionary.

The language processing unit 10 can use, for example, a method using dynamic programming as the morphological analysis. The language processing unit 10 can use a syntax analysis technique such as a prefetch LR parser or an LL method, for example, as dependency analysis. Then, the language processing unit 10 creates a morpheme reading notation and an intermediate notation according to the order of each word, reading, accent position, and break position.
The language processing unit 10 temporarily stores the generated morpheme reading notation and intermediate notation in the storage unit 3.

  The speech synthesizer 11 creates a synthesized speech signal based on the intermediate notation of the input text data.

  The speech synthesizer 11 generates a target prosody for generating a synthesized speech signal based on the intermediate notation. For this purpose, the speech synthesis unit 11 reads a plurality of prosodic models from the storage unit 3. This prosodic model represents a position in which the voice is raised and a position in which the voice is lowered in time order. Then, the speech synthesizer 11 selects a prosodic model that most closely matches the position of the accent indicated by the intermediate notation among a plurality of prosodic models. Then, the speech synthesizer 11 creates a target prosody by setting a position where the voice becomes high or low, a position where the voice becomes low, a voice inflection, a pitch, and the like according to the selected prosodic model and synthesis parameters. . The target prosody includes a phoneme length and a pitch frequency for each phoneme as a unit for determining a speech waveform. Note that the phoneme can be, for example, one vowel or one consonant.

The speech synthesizer 11 creates a synthesized speech signal according to the generated target prosody, for example, by an HMM (Hidden Markov Model) synthesis method, a phoneme connection method, or a corpus-based method.
For example, for each phoneme, the speech synthesizer 11 selects a speech waveform closest to the phoneme length and pitch frequency of the target prosody from a plurality of speech waveforms registered in the speech waveform dictionary by pattern matching, for example. For this purpose, the speech synthesis unit 11 reads a speech waveform dictionary from the storage unit 3. The speech waveform dictionary records a plurality of speech waveforms and an identification number of each speech waveform. The voice waveform is, for example, a waveform signal extracted in units of phonemes from voice signals obtained by recording various voices in which one or more narrators read various texts.
Furthermore, the speech synthesizer 11 connects the selected speech waveform and the waveform pattern of the corresponding phoneme indicated in the target prosody so that the speech waveform selected for each phoneme can be connected along the target prosody. The deviation amount may be calculated as waveform conversion information.
The speech synthesizer 11 creates waveform generation information including the identification number of the speech waveform selected for each phoneme. The waveform generation information may further include waveform conversion information.

The speech synthesizer 11 reads a speech waveform signal corresponding to the speech waveform identification number of each phoneme included in the waveform generation information from the storage unit 3. Then, the speech synthesizer 11 creates a synthesized speech signal by connecting each speech waveform signal continuously. When the waveform conversion information is included in the waveform generation information, the speech synthesizer 11 continuously connects the speech waveform signals by correcting each speech waveform signal according to the waveform conversion information obtained for the corresponding phoneme. By doing so, a synthesized speech signal is created.
The voice synthesizer 11 outputs the synthesized voice signal to the output unit 5.

  When the user corrects the intermediate notation generated by the language processing unit 10, the dictionary registration unit 12 registers the correction content in the user dictionary. For this purpose, the dictionary registration unit 12 includes an editing unit 21, a morpheme information setting unit 22, a part of speech information setting unit 23, a correction range setting unit 24, a registration range setting unit 25, and a registration unit 26.

  When the processing unit 4 receives an operation signal indicating that intermediate notation editing is to be performed from the input unit 2, the processing unit 4 activates the editing unit 21.

For example, the editing unit 21 displays the intermediate notation to be edited on the display of the input unit 2 together with the corresponding text data.
When the editing unit 21 receives an operation signal for correcting a part of the intermediate notation to be edited from the keyboard or the like of the input unit 2, for example, the reading of some words or the position of accents, the editing unit 21 performs intermediate processing according to the operation signal. Correct the notation. The editing unit 21 temporarily stores the corrected intermediate notation in the storage unit 3.

The editing unit 21 displays the prosodic symbols in Japanese on the display of the input unit 2 or displays the morpheme information of each morpheme so that even a user who has no expertise in speech synthesis can correct the intermediate notation. You may let them. For example, when the user selects a prosodic symbol to be corrected through the mouse of the input unit 2 or designates a position where a prosodic symbol is added, the editing unit 21 can be selected using, for example, a pull-down menu. A Japanese notation representing a prosodic symbol is displayed on the display. Then, the editing unit 21 replaces the prosodic symbol at the designated position in the intermediate notation with the prosodic symbol corresponding to the Japanese notation selected via the mouse or the like, or the prosodic symbol corresponding to the selected Japanese notation. Is automatically added to the specified position.
The editing unit 21 may display the corrected intermediate notation on the display so that the user can grasp the contents corrected by the user. Further, the editing unit 21 inputs the corrected intermediate notation to the speech synthesizer 11, thereby causing the user to correct the synthesized speech that is actually generated for the corrected intermediate notation from the speaker 6. The contents may be confirmed.

  FIG. 4 is a diagram illustrating an example of the intermediate notation output by the language processing unit 10 and the intermediate notation corrected by the user. Character strings 401 and 402 represent morpheme reading notation and intermediate notation generated by the language processing unit 10 from input text data, respectively. In this example, the morpheme reading notation 401 includes six morphemes “Yokan” (common noun), “Situ” (common noun), “wa” (particle), “Utsukushi” (adjective (stem)), “ku” "(Adjective (utilization ending)) and", "(symbol (reading marks)). In the intermediate notation 402 corresponding to the morpheme reading notation 401, in addition to the six parts of speech corresponding to the above six morphemes, a phrase representing a delimiter inserted between the particle “wa” and the adjective “Utsukushi” The symbol “=” is included.

  A character string 403 represents intermediate notation after correction by the user. In the intermediate notation 403, the second part-of-speech “sa’ichu” of the intermediate notation 402 before correction is changed to “mo * naka”. Further, the fifth part-of-speech “Utsukushi% '” is changed to “Utsuku'shi%”. The symbols “′”, “*”, “%”, etc. included in the intermediate notations 402 and 403 are prosodic symbols. Here, the prosodic symbol “′” represents “strong accent”, and the prosodic symbol “*” represents “accent weak”. In the intermediate notation 403, since the user does not specify individual parts of speech, when the user corrects the intermediate notation, the range of the intermediate notation corresponding to each part of speech included in the intermediate notation 403 and each part of speech is unknown. It is.

  The morpheme information setting unit 22 sets information of each morpheme included in the intermediate notation after being corrected by the user. Therefore, the morpheme information setting unit 22 reads from the storage unit 3 the intermediate notation before correction by the user and the intermediate notation after correction by the user. Then, the morpheme information setting unit 22 generates a corrected morpheme reading notation corresponding to the intermediate notation corrected by the user by removing the prosodic symbols from the intermediate notation corrected by the user.

  FIG. 5 is a diagram illustrating the morpheme reading notation before the user correction and the morpheme reading notation after the user correction corresponding to the intermediate notation of FIG. The morpheme reading notation 501 before the user modification shown in FIG. 5 is the same as the morpheme reading notation 401 shown in FIG. Further, the user-corrected morpheme reading notation 502 is obtained by removing the prosodic symbols from the user-corrected intermediate notation 403 shown in FIG. Comparing the morpheme reading notation 501 and the morpheme reading notation 502, it can be seen that the morpheme “situ” before correction and the “monaca” in the morpheme reading notation after correction do not match.

  In order to find this inconsistent portion, the morpheme information setting unit 22 performs, for example, a matching process (DP matching) using dynamic programming between the morpheme reading notation after the user correction and the morpheme reading notation before the user correction. Execute. And the morpheme information setting part 22 specifies the part which does not correspond with the part matched in a morpheme unit between the morpheme reading notation after a user correction, and the morpheme reading notation before a user correction. Then, the morpheme information setting unit 22 sets a part that matches the morpheme before the user correction in the morpheme reading notation after the user correction as the morpheme of the same type as the morpheme before the user correction. Also, the morpheme information setting unit 22 corresponds to a non-matching part even when a part that does not match the morpheme before the user correction corresponds to one morpheme before the user correction in the morpheme reading notation after the user correction. The morpheme is the same type as the morpheme before user modification. That is, the morpheme information setting unit 22 has the same type as the corresponding morpheme before the user modification as long as the part corresponding to the morpheme before the user modification is included in the morpheme reading notation after the user modification. Set to the morpheme.

  In FIG. 5, “Yokan”, “Monaca”, “Wa”, “Otsukushi”, “Ku”, “,” in the morpheme reading notation 502 after the user correction are respectively in the morpheme reading notation 501 before the user correction. Correspond to the morphemes “Yokan”, “Situ”, “Wa”, “Otsukushi”, “Ku”, “,”. Therefore, “Yokan”, “Monaka”, “Wa”, “Utsukushi”, “Ku”, “,” in the morpheme reading notation 502 after the user correction are respectively of the same type as the corresponding morpheme before the user correction. Set to morpheme. For example, “Yokan” is a common noun, and “Utsukushi” is the adjective stem. In addition, the morpheme that coincides with “monaca” is not included in the morpheme reading notation 501 before the user correction, but “monaca” has a one-to-one correspondence with the morpheme “situ” by the notation before and after “monaca”. You can see that it corresponds. Therefore, the morpheme information setting unit 22 sets the part of speech of “MONAKA” as a common noun same as the morpheme “Situ”.

  However, in the morpheme reading notation after user correction, a plurality of morphemes in the morpheme before user correction may correspond to a part that does not match the morpheme reading notation before user correction. In this case, the morpheme information setting unit 22 groups a plurality of morphemes corresponding to the non-matching parts into consecutive morphemes of the same type. However, in the present embodiment, when the morpheme is a noun, the morpheme information setting unit 22 ignores the classification of nouns such as ordinary nouns and saun nouns and treats them as the same type of morpheme. For example, when the common noun “Soi” and the Sa-noun “Qufu” are continuously included in the inconsistent portion, the morpheme information setting unit 22 “Soi Khoo” is a combination of the two nouns “Soi” and “Khoo”. "Is a group.

  When there is one morpheme included in the group corresponding to the non-matching part, it corresponds one-to-one before and after the correction, as in the case where only one morpheme corresponds. Therefore, the morpheme information setting unit 22 sets the morpheme of the mismatched part in the morpheme reading notation after the user correction as the morpheme of the same type as the corresponding morpheme group.

On the other hand, when a plurality of morphemes are included in the group corresponding to the non-matching part, if the group of the morpheme is not a noun, the morpheme information setting unit 22 sets the type of the morpheme of the group as it is as a mismatched part after the user correction. Morpheme type. This is due to the following reason.
Usually, when a morpheme group is not a noun, the type of morpheme of the morpheme group is roughly classified into a non-utilization independent word, a utilization independent word, and an ancillary word. Very rarely, there are misreads with the same notation, such as the verbs `` passed '' and `` goed '', but in the morpheme unit, only the stem part is read and corrected The reading is not corrected at the end of the usage. In other words, the inflection endings always match in morpheme units. For this reason, there are virtually no cases where a group includes a plurality of morphemes other than when the morpheme is a noun. Therefore, only a noun need be considered as a group composed of a plurality of morphemes.

As an example, when a morpheme group is composed of a plurality of morphemes, the setting of the morpheme type for the mismatched portion is classified into the following three cases.
(Case 1) When a plurality of morphemes included in the group of morphemes are all common nouns For example, the case where the morphemes included in the group are “sheep” and “middle” corresponds to case 1. In this case, the morpheme information setting unit 22 integrates the plurality of morphemes into one morpheme, that is, forms a compound noun. Then, the morpheme information setting unit 22 sets the type of the morpheme of the mismatched part to “common noun”.
(Case 2) When the last morpheme among a plurality of morphemes included in a group of morphemes is a sub-noun (for example) For example, the case where the morpheme included in a group is “creative” or “devise” corresponds to case 2 To do. In this case, the morpheme information setting unit 22 integrates the plurality of morphemes into one morpheme, that is, forms a compound noun. Then, the morpheme information setting unit 22 sets the type of the morpheme of the inconsistent portion to “sa variable noun”.
(Case 3) When the types of plural morphemes included in the morpheme group are other than the above case In this case, the morpheme information setting unit 22 integrates the plural morphemes into one morpheme, that is, forms a compound noun. . Then, the morpheme information setting unit 22 sets the type of the morpheme of the mismatched part to “proper noun”. In general, there are many reading errors in proper nouns (person names, place names, etc.), and it is preferable to define proper nouns in terms of probability.

  FIG. 6 is a diagram illustrating another example of the morpheme reading notation before correction by the user and the morpheme reading notation after correction by the user. In this example, as a result of language processing on the input text data “IaaS is”, the morpheme reading notation 601 before the user correction is a set of six morphemes. Specifically, the morpheme reading notation 601 includes the morpheme “A” (common noun), “A” (common noun), “A” (common noun), “S” (common noun), “wa” (particle). ), "," (Reading marks). On the other hand, the morpheme reading notation 602 after the user correction is “Yearswa,”. Therefore, the morpheme information setting unit 22 adds the four morphemes “A”, “A”, and “A” in the morpheme reading notation 601 before the user correction to the inconsistent part “Years” in the morpheme reading notation 602 after the user correction. ”And“ S ”correspond to each other. Since these four morphemes are all nouns, they are included in one morpheme group. Furthermore, since the four morpheme noun types are all common nouns, the example of FIG. Therefore, as shown in the morpheme reading notation 603, the morpheme information setting unit 22 integrates the four morphemes into one morpheme, and sets the morpheme type of the corresponding mismatched part in the morpheme reading notation after the user correction. Use common nouns.

FIG. 7 is an operation flowchart of the morpheme information setting process executed by the morpheme information setting unit 22.
The morpheme information setting unit 22 derives a corrected morpheme reading notation from the corrected intermediate notation (step S101). Then, the morpheme information setting unit 22 specifies a matching part and a mismatching part by matching between morpheme reading notations before and after correction (step S102).

The morpheme information setting unit 22 sets the range and type of the corresponding morpheme in the morpheme reading notation before correction to the matching part in the corrected morpheme reading notation (step S103).
On the other hand, the morpheme information setting unit 22 determines whether there is one morpheme corresponding to the mismatched part in the corrected morpheme reading notation (step S104).
When there is one morpheme corresponding to the mismatched part (step S104-Yes), the morpheme information setting unit 22 sets the range and type of the corresponding morpheme in the morpheme reading notation before correction to the mismatched part (step S105). ). On the other hand, when there are a plurality of morphemes corresponding to the mismatched part (step S104-no), the morpheme information setting unit 22 groups the mismatched parts in the morpheme reading notation before correction for each successive morpheme of the same type ( Step S106). The morpheme information setting unit 22 determines whether there is one morpheme included in the group (step S107). When there is one morpheme included in the group (step S107—Yes), the morpheme information setting unit 22 sets the range and type of the corresponding morpheme in the morpheme reading notation before correction in the mismatched part (step S105). . On the other hand, when the number of morphemes included in the group is not one (step S107-no), the morpheme information setting unit 22 determines whether all the morphemes included in the group are common nouns (step S108).

When all the morphemes included in the group are common nouns (step S108-Yes), the morpheme information setting unit 22 sets the entire mismatched part in the corrected morpheme reading notation as one common noun (step S109). On the other hand, if any of the morphemes included in the group is not a common noun (step S108-no), the morpheme information setting unit 22 determines whether the last morpheme included in the group is a sub-noun (step S110). When the last morpheme included in the group is a sa-variant noun (step S110-Yes), the morpheme information setting unit 22 sets the entire inconsistent part in the corrected morpheme reading notation as one sa-noun (step S111). . On the other hand, when the last morpheme included in the group is not a savariant noun (step S110-no), the morpheme information setting unit 22 sets the entire mismatched part in the corrected morpheme reading notation as one proper noun (step S112). ).
After steps S105, S109, S111, or S112, the morpheme information setting unit 22 ends the morpheme information setting process. By this procedure, the morpheme information setting unit 22 can appropriately set the part of speech of each morpheme in the corrected morpheme reading notation.

  The morpheme information setting unit 22 notifies the part-of-speech information setting unit 23 of the morpheme information set for the morpheme reading notation after the user correction.

  The part-of-speech information setting unit 23 sets information on each part-of-speech in the intermediate notation after the user correction based on the information on each morpheme in the morpheme reading notation after the user correction. Due to the relationship between the morpheme reading notation and the intermediate notation, the morpheme reading notation after the user correction and the intermediate notation after the user correction have a one-to-one correspondence except for the prosodic symbols. Therefore, the part-of-speech information setting unit 23 specifies the range of each part-of-speech included in the intermediate notation after the user correction from the portion where the readings of the morpheme reading notation after the user correction match, and sets the type of each part-of-speech after the user correction To the corresponding morpheme type in the morpheme reading notation. For the part of speech including only the prosodic symbol, for example, the part of speech including only the prosodic symbol “=” indicating punctuation, the part of speech information setting unit 23 uses the intermediate notation before the user correction as the part of speech in the intermediate notation after the user correction. May be set to the corresponding part of speech.

FIG. 8 is a diagram illustrating an example of the relationship between the intermediate notation before and after the user correction and the morpheme reading notation after the user correction. The intermediate notation 802 after user correction and the morpheme reading notation 803 after user correction are identical except for the prosodic symbols. Therefore, each of the morphemes “Yokan”, “Monaca”, “Wa”, “Utsukushi”, “Ku” and “,” included in the morpheme reading notation 803 is the part of speech “Yokan”, “Mo *” of the intermediate notation 802, respectively. It can be seen that it corresponds to “Naka”, “Wa”, “Utsuku's%”, “Ku” and “,”. Therefore, the types of parts of speech “Yokan”, “Mo * Naka”, “Wa”, “Utsuku'shi%”, “Ku” and “,” are the types of the corresponding morphemes in the morpheme reading notation 803, respectively. , Common noun, common noun, particle (sub), adjective (stem), adjective (utilization ending) and punctuation. A prosodic symbol “=” having no corresponding part in the morpheme reading notation 803 is set to a symbol (phrase) that is a part of speech of the corresponding prosodic symbol “=” in the intermediate notation 801 before the user correction.
The part of speech information setting unit 23 notifies the correction range setting unit 24 of information on each part of speech in the intermediate notation after the user correction.

  The correction range setting unit 24 sets a range corrected by the user in units of parts of speech from the intermediate notation after the user correction. Therefore, the correction range setting unit 24 performs matching processing, for example, DP matching, on a part-of-speech basis between the intermediate notation after the user correction and the intermediate notation before the user correction, and specifies the part of speech that does not match. Then, the correction range setting unit 24 sets the part of speech included in the intermediate notation after the user correction to the part of speech that does not match the part of speech of the intermediate notation before the user correction. Thus, by setting the correction range in units of parts of speech, the correction range setting unit 24 can appropriately set the correction range to a range intended by the user for correction.

FIG. 9 is a diagram illustrating an example of the relationship between the intermediate notation before and after the user correction and the correction range that is set. In FIG. 9, in the intermediate notation 901 before the user correction, in the intermediate notation 902 after the user correction, the part of speech “sa'ichu” is corrected to the part of speech “mo * naka”. Therefore, as indicated by the arrow 903, the part of speech “MO * NAKA” is set as the correction range.
Further, the part-of-speech “Utsukushi% ′” before correction becomes “Utsukushi%” after correction, and the position of the prosodic symbol representing the accent strength is corrected. Also in this case, not only the prosodic symbols but also the entire part of speech “Utsuku's%” is set as the correction range, as indicated by the arrow 904.
The correction range setting unit 24 notifies the registration range setting unit 25 of the set correction range.

  The registration range setting unit 25 sets a range to be registered in the user dictionary from the correction range set in the intermediate notation after user correction. When the user corrects a part of the continuous nouns, if the correction contents are registered only for the corrected nouns, the corrected nouns can be changed as shown in “middle” described above with reference to FIG. May be converted to an incorrect intermediate representation when used in the context of. In such a case, by registering the noun “middle” in the user dictionary as well as the noun “yogo” prefixed to the noun, and registering it as “middle mushroom”, “middle” When used in reading “Situ”, it can be prevented from being mistakenly “Monaca”. Therefore, in the present embodiment, when the part of speech included in the correction range is a noun, the registration range setting unit 25 sets the registration range so that a noun that is continuous with the noun is also included in the registration range.

  In addition, the accent position of a self-supporting word may change depending on the ending. For example, as shown in FIG. 8, when the correction range is a stem of a use independent word, the prosody is only used when combined with the ending in the actually corrected intermediate notation. It is likely to be applied. Therefore, in the present embodiment, when the part of speech included in the correction range is a stem of a utilized independent word, the registration range setting unit 25 sets the utilized independent word and the ending of the utilized independent word in the registration range. As described above, when the part of speech included in the correction range is a predetermined part of speech, the registration range setting unit 25 sets the registration range so as to include the same part of speech continuous with the part of speech of the correction range. Thereby, the registration range setting unit 25 can appropriately reflect the correction by the user in the user dictionary.

  FIG. 10 is a diagram illustrating an example of the relationship between the intermediate notation before and after the user correction and the registration range. Comparing the intermediate notation 1000 before the user correction and the intermediate notation 1001 after the user correction, the “situation” is corrected as “mo * naka”, and the “development percentage” is modified as “development percentage”. Yes. Therefore, the noun 1002 (“M * Naka”) and the adjective stem 1003 (“Utsuku 'shi%”) are set in the correction range. In this case, for the noun 1002, the registration range 1004 is set so as to include the prefix noun “Yokan”. On the other hand, for the adjective stem 1003, the registration range 1005 is set so as to include “ku”, which is an inflection ending.

  FIG. 11 is an operation flowchart of the registration range setting process. In addition, the registration range setting part 25 performs the following processes for every correction range, when a some correction range is contained in the intermediate description after a user correction.

  First, as an initial process, the registration range setting unit 25 sets the correction range itself as the registration range. Then, the registration range setting unit 25 determines whether or not the part of speech included in the correction range is a noun (step S201). When the part of speech included in the correction range is a noun (step S201—Yes), the registration range setting unit 25 determines whether the prefix part of speech of the registration range is a noun (step S202). When the prefix part of speech is a noun (step S202—Yes), the registration range setting unit 25 extends the head position of the registration range to the head of the prefix part of speech (step S203). Thereafter, the registration range setting unit 25 repeats the processing after step S202.

  On the other hand, if the prefix part of speech is not a noun (step S202-No), the registration range setting unit 25 determines whether the postfix part of speech of the registration range is a noun (step S204). When the postposition part of speech is a noun (step S204-Yes), the registration range setting unit 25 extends the rear end position of the registration range to the end of the postposition part of speech (step S205). Thereafter, the registration range setting unit 25 repeats the processing after step S204.

  On the other hand, if the postscript part of speech is not a noun (step S204-No), the registration range setting unit 25 sets all the continuous nouns included in the registration range as the registration range (step S206).

If the part of speech included in the correction range is not a noun in step S201 (step S201-No), the registration range setting unit 25 determines whether or not the part of speech included in the correction range is a stem of a use independent word (step S207). ). If the part-of-speech included in the correction range is a stem of a use independent word (step S207-Yes), the registration range setting unit 25 sets the stem of the use independent word and a postfix use ending in the registration range (step S208). ). On the other hand, if the part of speech included in the correction range is not the stem of the independence word (step S207-No), the registration range setting unit 25 sets only the part of speech included in the correction range as the registration range (step S209).
After step S206, S208, or S209, the registration range setting unit 25 ends the registration range setting process.

  The registration range setting unit 25 notifies the registration unit 26 of the set registration range.

  The registration unit 26 registers the intermediate notation and part of speech of the word together with the kanji kana notation of the word in the user dictionary, with the part included in the registration range set in the intermediate notation after the user correction as one word. For example, in the example shown in FIG. 10, an intermediate notation of “Yokanmo * Naka” and a common noun are registered in the user dictionary in association with the word “midst of sheep”. In addition, an intermediate notation “adjective” and an adjective are registered in the user dictionary in association with the word “beautiful”.

FIG. 12 is an operation flowchart of dictionary registration processing according to this embodiment.
The morpheme information setting unit 22 generates a morpheme reading notation after user correction from the intermediate notation after user correction (step S301). The morpheme information setting unit 22 sets the range and type of each morpheme included in the morpheme reading notation after user correction (step S302).

  The part of speech information setting unit 23 specifies the range and type of each part of speech in the intermediate notation after user correction from the range and type of each morpheme after user correction (step S303). Then, the correction range setting unit 24 sets the correction range in parts of speech by matching the intermediate notation before and after the user correction in units of parts of speech (step S304).

The registration range setting unit 25 sets the registration range so that the same part of speech continues before and after when the part of speech included in the correction range is a predetermined part of speech (step S305). The registration unit 26 registers the correction contents in the user dictionary in units of registration ranges (step S306).
Thereafter, the processing unit 4 ends the dictionary registration process.

  As described above, when the user corrects the intermediate notation, the speech synthesizer automatically specifies the range to be registered in the user dictionary and the part of speech included in the range from the corrected intermediate notation. For this reason, this speech synthesizer allows the user to edit the contents of the correction even if the user does not have special knowledge about the intermediate notation, or without setting a range or part of speech to be registered in detail for the point corrected by the user. Appropriately reflected in the dictionary. Therefore, this speech synthesizer can prevent the user from registering inappropriate correction content in the user dictionary or forgetting to register part of the correction content to be registered in the user dictionary. As a result, this speech synthesizer can generate an appropriate synthesized speech signal even for text including words having a plurality of different readings or prosody.

Next, a speech synthesizer according to the second embodiment will be described. The speech synthesizer according to the second embodiment includes, as a dictionary for registering correction contents by a user, a user dictionary that stores kanji kana notation and intermediate notation of words, and one or more words positioned before and after the registered words, It has an intermediate notation dictionary that also stores the relationship. The speech synthesizer executes language processing using the two types of dictionaries together with the language dictionary. In addition, when the intermediate notation is corrected by the user, the speech signal device automatically determines whether the correction content is registered in the user dictionary or the intermediate notation dictionary.
Each of the user dictionary and the intermediate notation dictionary is an example of a language dictionary.

The speech synthesizer according to the second embodiment is different from the speech synthesizer according to the first embodiment in that the storage unit 3 stores an intermediate notation dictionary and the processing and dictionary of the language processor 10 of the processor 4. The processing of the registration unit 26 of the registration unit 12 is different.
Therefore, hereinafter, the intermediate notation dictionary, the language processing unit 10, and the registration unit 26 will be described. For the other components of the speech synthesizer according to the second embodiment, refer to the description of the components that face each other in the first embodiment.

  In the present embodiment, it is assumed that a corrected reading of a single noun and a compound noun, a corrected accent position, and a break position are registered in the user dictionary. On the other hand, in the intermediate notation dictionary, correction contents of intermediate notations other than those to be registered in the user dictionary are registered. For example, prosodic information corrected by the user is registered in the intermediate notation dictionary. This prosodic information includes, for example, the strength of accents, the pitch of pitches, the level of inflection, the speed of speech, the level of volume, and the break position. However, the registration target of the user dictionary and the registration target of the intermediate notation dictionary are not limited to the above example. Since there are no general rules for user dictionaries, registration targets for user dictionaries and intermediate notation dictionaries need only be set according to the specifications of the speech synthesizer. For example, the user dictionary may register not only single nouns and compound nouns but also used independent words or non-used independent words.

  FIG. 13 is a diagram illustrating an example of a word registered in the user dictionary and a word registered in the intermediate notation dictionary. In the user dictionary 1300, the word “Yokanmochu” written in kanji kana is registered in association with the common noun that is the type of part of speech and “Yokanmo 'Naka” that is an intermediate notation including the accented prosodic symbol. ing. On the other hand, in the intermediate notation dictionary 1301, “Yokanmo * Naka”, which is an intermediate notation including a prosodic symbol with weak accent, is registered in association with the word “middle of the sheep” written in Kanji. Furthermore, in the intermediate notation dictionary 1301, the kanji kana notation “this” for the prefixed word and the postfixed word, which are conditions for applying the intermediate notation associated with the word “middle of the sheep”, are applied. The kanji kana notation “Ha Beautiful” is associated and registered. In other words, in the case of “in the middle of the sheep's wing” alone, it is preferable that the accent at “mo” is strong, but for the sentence “in the middle of this ramie”, the accent at “mo” is weakened. The intermediate notation dictionary represents this. Note that the intermediate notation dictionary 1301 may also register an intermediate notation of a prefixed word or a postfixed word.

  Also in this embodiment, the language processing unit 10 first creates an intermediate representation of input text data using a language dictionary and a user dictionary. At this time, if the text data includes notations registered in both the language dictionary and the user dictionary, the language processing unit 10 preferentially uses the user dictionary. Therefore, for example, as shown in FIG. 1, it is assumed that the text data includes a sentence “Beautiful in the midst of this sheep” and “In the middle of the sheep” is registered in the user dictionary. In this case, the language processing unit 10 outputs “... Yokakan 'Naka...” Instead of “.

  Further, the language processing unit 10 refers to the intermediate notation dictionary and refers to the kanji kana notation of the word registered in the intermediate notation dictionary in the text data, the kanji kana notation of the prefixed word, and the kanji of the postfixed word. It is determined whether there is a part that matches the kana notation. If there is a matching part, the language processing unit 10 replaces the intermediate notation of the corresponding part in the intermediate notation with the intermediate notation registered in the intermediate notation dictionary. Thereby, the language processing unit 10 can automatically correct the accent, pitch, intonation, speech speed, volume, or break of the word in the synthesized speech signal according to the context. Therefore, this speech synthesizer can increase the possibility that the synthesized speech desired by the user can be reproduced more appropriately.

  Next, a process for selecting a dictionary that reflects the correction when the user corrects the intermediate notation will be described. In the present embodiment, this dictionary selection is performed by the registration unit 26 of the dictionary registration unit 12.

FIG. 14 is an operation flowchart of dictionary selection processing executed by the registration unit 26. The dictionary selection process is executed in step S306 of the dictionary registration process shown in FIG.
The registration unit 26 determines whether or not the correction content within the set registration range is any one of reading, a break position, and an accent position (step S401). If the correction content is neither reading nor a break position or an accent position (step S401-No), the registration unit 26 selects an intermediate notation dictionary (step S402). The registration unit 26 extracts, from the intermediate notation after the user correction, the kanji kana notation of the part of speech prefixed to the registration range and the kanji kana notation of the part of speech postfixed to the registration range (step S403). In addition, when the postscript part of speech is a particle, the part of speech subsequent to the particle may be extracted. Then, the registration unit 26 registers the intermediate notation and kanji kana notation after the correction of the registration range, the prefix kanji kana notation, and the post-kanji kana notation in the intermediate notation dictionary (step S404). In general, corrections other than reading and accent positions are any of “accent strength”, “pitch pitch”, “inflection magnitude small”, “speaking speed gradual”, “volume loudness”, and “break” in the synthesized speech signal. It is a correction. These corrections are made in consideration of the connection before and after the audibility of the synthesized speech. Therefore, it is preferable that words before and after the registration range are also registered.

  On the other hand, in step S401, when the correction content is any one of reading, a break position, and an accent position (step S401-Yes), the registration unit 26 indicates that the part of speech of the registration range is a single noun or a compound noun. It is determined whether or not it is any (step S405). If the part of speech in the registration range is neither a single noun nor a compound noun (step S405—No), the registration unit 26 executes the processes of steps S402 to S404 and registers the correction content in the intermediate notation dictionary.

On the other hand, when the part of speech of the registration range is either a single noun or a compound noun (step S405-Yes), the registration unit 26 selects a user dictionary (step S406). The registration unit 26 registers the intermediate notation, kanji kana notation, and part-of-speech type after correction of the registration range in the user dictionary (step S407).
After step S404 or S407, the registration unit 26 ends the dictionary selection process.

  According to this embodiment, the speech synthesizer can automatically determine whether or not the user tried to correct the expression of the synthesized speech in a specific context and reflect the result in the dictionary. . Therefore, this speech synthesizer can create a synthesized speech signal having an appropriate prosody according to the context while reducing the correction burden on the user.

  According to the modified example, for a word registered in the intermediate notation dictionary, any one of one or more words preceding the word or one or more words following the word, together with kanji kana notation and intermediate notation of the word Only or may be registered. A plurality of one word may be registered in the intermediate notation dictionary for each combination of one or more different words that precede or follow the word.

  According to another modification, only one language dictionary may be used for language processing. In this case, all corrections made by the user regarding the intermediate notation are reflected in the language dictionary. Then, the language processing unit 10 may perform the language processing by giving priority to the content additionally registered by the result of the user correction in the language dictionary.

  According to another modification, the editing unit 21 of the dictionary registration unit 12 may allow the user himself / herself to specify the correction range and the part-of-speech in the correction range during the intermediate notation. In this case, the dictionary registration unit 12 automatically sets the registration range based on the correction range set by the user. As described above, by allowing the user to specify the details of the correction in detail, it is possible to improve the convenience of the user who has abundant knowledge of speech synthesis.

  Furthermore, a computer program that causes a computer to realize each function of the processing unit of the speech synthesizer according to each of the above embodiments is recorded on a computer-readable medium, for example, a magnetic recording medium, an optical recording medium, or a semiconductor memory. It may be provided in the form.

  All examples and specific terms listed herein are intended for instructional purposes to help the reader understand the concepts contributed by the inventor to the present invention and the promotion of the technology. It should be construed that it is not limited to the construction of any example herein, such specific examples and conditions, with respect to showing the superiority and inferiority of the present invention. Although embodiments of the present invention have been described in detail, it should be understood that various changes, substitutions and modifications can be made thereto without departing from the spirit and scope of the present invention.

The following supplementary notes are further disclosed regarding the embodiment described above and its modifications.
(Appendix 1)
A speech synthesizer that generates a synthesized speech signal based on an intermediate notation that includes prosodic symbols representing prosody and reading of the text data generated from text data,
An input unit for acquiring the text data;
A storage unit that stores at least a kanji kana notation of a word and an intermediate notation including the reading of the word and the prosodic symbol, and stores a language dictionary used to generate the intermediate notation from the text data;
When the word included in the corrected range in the corrected intermediate notation in which the intermediate notation is corrected is a predetermined part of speech, the word including the part of speech that is the same as the part of speech of the word that is continuous before and after the word included in the corrected range is included. A registration range setting unit for extending a registration range to be registered in the language dictionary,
A registration unit for registering at least a kanji kana notation and an intermediate notation of the word in the language dictionary, with a portion included in the registration range in the corrected intermediate notation as one word,
A speech synthesizer.
(Appendix 2)
The predetermined part-of-speech is a noun, and when the word included in the correction range is a noun, the registration range setting unit extends the registration range to include nouns continuous before and after the correction range. The speech synthesizer according to 1.
(Appendix 3)
The predetermined part of speech is a stem of a use independent word, and the registration range setting unit includes up to the ending of a use independent word following the correction range when the word included in the correction range is a stem of the use independent word The speech synthesizer according to appendix 1, wherein the registration range is extended as described above.
(Appendix 4)
The speech synthesizer according to any one of appendices 1 to 3, wherein the registration unit registers the part of speech of a word included in the registration range together with the kanji kana notation and the intermediate notation of the word in the language dictionary.
(Appendix 5)
By removing the prosodic symbols from the modified intermediate notation, a modified morpheme reading notation representing the corrected reading of the text data is generated, and the morpheme reading notation obtained by removing the prosodic symbols from the intermediate notation and the modified A morpheme information setting unit for obtaining a range and a type of each part of speech included in the corrected morpheme reading notation by matching with the post-morpheme reading notation;
For each morpheme included in the corrected morpheme reading notation, a part of speech information setting unit that sets a part of the corrected intermediate notation that matches the reading of the morpheme to the same part of speech as the morpheme,
A correction range setting unit that extracts different parts between the corrected intermediate notation and the intermediate notation, and sets the entire part of speech including the different parts in the correction range;
The speech synthesizer according to any one of appendices 1 to 4, further comprising:
(Appendix 6)
The morpheme information setting unit extracts a portion that matches and does not match between the corrected morpheme reading notation and the morpheme reading notation, and the matching part in the corrected morpheme reading notation includes the morpheme reading notation Set the same part of speech as the corresponding part of speech, and when one part of speech of the morpheme reading notation in the corrected morpheme reading notation corresponds to the part of the morpheme reading notation, The speech synthesizer according to appendix 5, wherein the part of speech corresponding to the reading notation is set.
(Appendix 7)
The morpheme information setting unit, when a plurality of consecutive common nouns of the morpheme reading notation corresponds to the non-matching part in the modified morpheme reading notation, the part of speech of the mismatched part is a common noun, The speech synthesizer according to appendix 6.
(Appendix 8)
The morpheme information setting unit includes a plurality of consecutive nouns in the morpheme reading notation corresponding to the non-matching part in the corrected morpheme reading notation, and the last noun of the plurality of nouns is a saun noun. The speech synthesizer according to appendix 6, wherein in some cases, the part-of-speech of the non-matching part is a sa variable noun.
(Appendix 9)
The morpheme information setting unit includes a plurality of consecutive nouns in the morpheme reading notation that correspond to a non-matching part in the corrected morpheme reading notation, and the last noun of the plurality of nouns is not a saun noun In addition, if any of the plurality of nouns is not a common noun, the speech synthesizer according to appendix 6, wherein the part of speech that does not match is a proper noun.
(Appendix 10)
The storage unit includes, as the language dictionary, a first language dictionary in which an intermediate notation of the word is registered together with at least the kanji kana notation of the word, an intermediate notation of the word and at least the kanji notation of the word Storing a second language dictionary in which kanji notation of one or more words located before or after is registered;
When the prosody other than the accent position is changed in the registration range of the modified intermediate notation, the registration unit includes kanji kana notation of one or more words positioned before or after the registration range, and the registration range. The speech synthesizer according to any one of appendices 1 to 9, wherein kanji kana notation and intermediate notation of an included word are registered in the second language dictionary.
(Appendix 11)
The registration unit, when the part of speech of the word included in the registration range of the corrected intermediate notation is a noun and the prosody other than the accent position is not changed, kanji kana notation of the word included in the registration range The speech synthesizer according to appendix 10, wherein the intermediate notation and the intermediate notation are registered in the first language dictionary.
(Appendix 12)
Among the text data, the intermediate notation is generated by using a word that matches the kanji kana notation registered in the first language dictionary as the intermediate notation of the word, and the text data includes the An intermediate notation of the intermediate notation corresponding to a portion that matches a kanji notation of a word registered in the second language dictionary and also matches a kanji notation of one or more words located before or after the word; The speech synthesizer according to appendix 10 or 11, further comprising a language processing unit for rewriting the intermediate notation of the word registered in the second language dictionary.
(Appendix 13)
A method for registering a language dictionary used to create an intermediate notation including prosodic symbols representing prosody and reading of the text data to generate a synthesized speech signal from text data,
Obtaining the text data,
When the word included in the correction range in the corrected intermediate notation in which the intermediate notation is corrected is a predetermined part of speech, the processor has the same part of speech as the part of speech of the word consecutively before and after the word included in the correction range. Extend the registration range to be registered in the language dictionary to include even words,
The processor stores, in a storage unit, a kanji kana notation of the word and an intermediate notation including the reading of the word and the prosodic symbol, with a portion included in the registered range in the corrected intermediate notation as one word Registered in the language dictionary
Language dictionary registration method including that.

1 speech synthesis device 2 input unit 3 storage unit 4 processing unit 5 output unit 6 speaker 10 language processing unit 11 speech synthesis unit 12 dictionary registration unit 21 editing unit 22 morpheme information setting unit 23 part of speech information setting unit 24 correction range setting unit 25 registration Range setting part 26 Registration part

Claims (10)

A speech synthesizer that generates a synthesized speech signal based on an intermediate notation that includes prosodic symbols representing prosody and reading of the text data generated from text data,
An input unit for acquiring the text data;
A storage unit that stores at least a kanji kana notation of a word and an intermediate notation including the reading of the word and the prosodic symbol, and stores a language dictionary used to generate the intermediate notation from the text data;
When the word included in the corrected range in the corrected intermediate notation in which the intermediate notation is corrected is a predetermined part of speech, the word including the part of speech that is the same as the part of speech of the word that is continuous before and after the word included in the corrected range is included. A registration range setting unit for extending a registration range to be registered in the language dictionary,
A registration unit for registering at least a kanji kana notation and an intermediate notation of the word in the language dictionary, with a portion included in the registration range in the corrected intermediate notation as one word,
A speech synthesizer.   The predetermined part-of-speech is a noun, and when the word included in the correction range is a noun, the registration range setting unit extends the registration range to include nouns continuous before and after the correction range. Item 2. The speech synthesizer according to Item 1.   The predetermined part of speech is a stem of a use independent word, and the registration range setting unit includes up to the ending of a use independent word following the correction range when the word included in the correction range is a stem of the use independent word The speech synthesis apparatus according to claim 1, wherein the registration range is extended as follows. By removing the prosodic symbols from the modified intermediate notation, a modified morpheme reading notation representing the corrected reading of the text data is generated, and the morpheme reading notation obtained by removing the prosodic symbols from the intermediate notation and the modified A morpheme information setting unit for obtaining a range and a type of each part of speech included in the corrected morpheme reading notation by matching with the post-morpheme reading notation;
For each morpheme included in the corrected morpheme reading notation, a part of speech information setting unit that sets a part of the corrected intermediate notation that matches the reading of the morpheme to the same part of speech as the morpheme,
A correction range setting unit that extracts different parts between the corrected intermediate notation and the intermediate notation, and sets the entire part of speech including the different parts in the correction range;
The speech synthesizer according to claim 1, further comprising:   The morpheme information setting unit extracts a portion that matches and does not match between the corrected morpheme reading notation and the morpheme reading notation, and the matching part in the corrected morpheme reading notation includes the morpheme reading notation Set the same part of speech as the corresponding part of speech, and when one part of speech of the morpheme reading notation in the corrected morpheme reading notation corresponds to the part of the morpheme reading notation, The speech synthesizer according to claim 4, wherein the part of speech of a corresponding part of the reading notation is set.   The morpheme information setting unit includes a plurality of consecutive nouns in the morpheme reading notation corresponding to the non-matching part in the corrected morpheme reading notation, and the last noun of the plurality of nouns is a saun noun. The speech synthesizer according to claim 5, wherein, in some cases, the part of speech of the non-matching part is a sa-changing noun. The storage unit includes, as the language dictionary, a first language dictionary in which an intermediate notation of the word is registered together with at least the kanji kana notation of the word, an intermediate notation of the word and at least the kanji notation of the word Storing a second language dictionary in which kanji notation of one or more words located before or after is registered;
When the prosody other than the accent position is changed in the registration range of the modified intermediate notation, the registration unit includes kanji kana notation of one or more words positioned before or after the registration range, and the registration range. The speech synthesizer according to claim 1, wherein kanji kana notation and intermediate notation of an included word are registered in the second language dictionary.   The registration unit, when the part of speech of the word included in the registration range of the corrected intermediate notation is a noun and the prosody other than the accent position is not changed, kanji kana notation of the word included in the registration range The speech synthesizer according to claim 7, wherein the intermediate notation and the intermediate notation are registered in the first language dictionary.   Among the text data, the intermediate notation is generated by using a word that matches the kanji kana notation registered in the first language dictionary as the intermediate notation of the word, and the text data includes the An intermediate notation of the intermediate notation corresponding to a portion that matches a kanji notation of a word registered in the second language dictionary and also matches a kanji notation of one or more words located before or after the word; The speech synthesizer according to claim 7, further comprising a language processing unit that rewrites the intermediate representation of the word registered in the second language dictionary. A method for registering a language dictionary used to create an intermediate notation including prosodic symbols representing prosody and reading of the text data to generate a synthesized speech signal from text data,
Obtaining the text data,
When the word included in the correction range in the corrected intermediate notation in which the intermediate notation is corrected is a predetermined part of speech, the processor has the same part of speech as the part of speech of the word consecutively before and after the word included in the correction range. Extend the registration range to be registered in the language dictionary to include even words,
The processor stores, in a storage unit, a kanji kana notation of the word and an intermediate notation including the reading of the word and the prosodic symbol, with a portion included in the registered range in the corrected intermediate notation as one word Registered in the language dictionary
Language dictionary registration method including that.

Images