JP2009103921A - Abbreviated word determining apparatus, computer program, text analysis apparatus, and speech synthesis apparatus - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an abbreviated-word determining apparatus for determining whether text data is an abbreviated word, a computer program for realizing the abbreviated word determination apparatus by a computer, a text analysis apparatus having the abbreviated-word determining apparatus, and a speech synthesis apparatus having the text analysis apparatus. <P>SOLUTION: A control part 1 divides text data into morphemes, based on the content of the registration in a language dictionary 4a, and assigns accent type to each of the morphemes. The control part 1 determines whether a morpheme that is not registered in the language dictionary 4a (an unknown word) is an abbreviated word, based on the content of the registration in a personal name dictionary 4b or a compound word dictionary 4c. The control part 1 assigns different accent types to the unknown word which is determined as being an abbreviated word, and the unknown word determined as not being an abbreviated word. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to an abbreviation determination device that determines whether text data is an abbreviation, a computer program for realizing the abbreviation determination device by a computer, a text analysis device including the abbreviation determination device, and The present invention relates to a speech synthesizer provided with the text analysis device.

  Text-to-speech synthesis technology that synthesizes speech from text data includes, for example, IVR (Automatic Voice Response) system, in-vehicle information terminal, voice guidance of operation method in mobile phone, etc., reading out e-mail, It is applied to support systems for the speech disabled.

  In the conventional text-to-speech synthesizer, a language dictionary for storing morphemes and accent types of each morpheme is prepared in advance, and the input text data is divided into morphemes based on the registered contents of the language dictionary. Accent type is given to each morpheme. In addition, the conventional text-to-speech synthesizer generates prosody corresponding to each morpheme according to predetermined prosody generation rules based on the divided morphemes and the accent type assigned to each morpheme, and converts the generated prosody into a speech waveform To obtain synthesized speech.

  In such a conventional text-to-speech synthesizer, if text data contains morphemes that are not registered in the language dictionary, there is a possibility that the morphemes may be divided at wrong positions when they are divided. There is a possibility that an incorrect accent type may be given when it is given. Thus, when divided into morphemes at an incorrect position, or when an incorrect accent type is given, it is difficult to generate a correct synthesized speech.

  Further, in the conventional text-to-speech synthesizer, when dividing text data into morphemes, if morphemes that are not registered in the language dictionary are extracted, the morphemes are divided as unknown words, and each morpheme (unknown word) For example, in many cases, an accent type having an accent nucleus is assigned to the third mora from the back. This is because many foreign words such as “Australia” and “Chernobyl” have an accent nucleus in the third mora from the back.

  Conventionally, abbreviations such as Matsuken (registered trademark) (Ken Matsuhira), Kim Taku (Takuya Kimura), Cosplay (costume play), Terrestrial digital (terrestrial digital broadcasting), serial drama (continuous drama), etc. have been frequently used. Since such abbreviations are not registered in the language dictionary, the conventional text-to-speech synthesizer often handles them as unknown words. As described above, for example, an accent type having an accent nucleus is assigned to the third mora from the back. Was. However, such abbreviations are often flat-type (zero-type) accent types, and if an accent type with an accent kernel is given to the third mora from the back, correct synthesized speech cannot be generated. First, there was a possibility of generating synthesized speech that was read like a whisper.

In Patent Document 1, an official name dictionary in which official names for abbreviations are registered and an abbreviation dictionary in which abbreviations predicted from official names are prepared are registered in the abbreviation dictionary in a text document. An apparatus is disclosed that converts an abbreviation into a corresponding formal name when an abbreviation is detected. When such an apparatus is used, even if an abbreviation not registered in the official name dictionary is included in the text document, it can be appropriately handled as an abbreviation without being treated as an unknown word.
JP 2004-326367 A

  By using an abbreviation dictionary as in Patent Document 1 described above, abbreviations registered in the abbreviation dictionary can be divided into correct morphemes when a text document is divided into morphemes, and correct accents Since a type can be given, a correct synthesized speech can be generated. However, since abbreviated words come out every day, it is impossible to register new abbreviated words in the abbreviation dictionary one after another.

  The present invention has been made in view of such circumstances, and an object of the present invention is to omit whether text data is an abbreviation generated by omitting a person's name or not. An object of the present invention is to provide a word determination device, a computer program for realizing the abbreviation determination device by a computer, a text analysis device including the abbreviation determination device, and a speech synthesizer including the text analysis device.

  Another object of the present invention is to realize an abbreviation determination device that can easily determine whether text data is an abbreviation generated by omitting a compound word, and the abbreviation determination device realized by a computer. The present invention provides a computer program, a text analysis device including the abbreviation determination device, and a speech synthesis device including the text analysis device.

  The abbreviation determination apparatus according to the present invention is an abbreviation determination apparatus for determining whether text data is an abbreviation or not. In the abbreviation determination apparatus, a surname and a first name used for a person's name are respectively stored in a person name storage means. A predetermined number of character data at the beginning is extracted, and it is determined whether or not the surname having the extracted character data at the beginning is stored in the personal name storage means. Further, when the abbreviation determination device determines that the last name is stored in the personal name storage unit, the abbreviation determination device extracts and extracts a predetermined number of character data at the beginning from the text data excluding the extracted character data. It is determined whether a name having character data at the head is stored in the personal name storage means. If it is determined that the name is stored in the personal name storage means, the abbreviation determination device determines that the text data is an abbreviation. Therefore, it is possible to easily determine whether or not the text data is an abbreviation generated by extracting a predetermined number of characters from the head of each surname and first name of the person name.

  The abbreviation determination device according to the present invention is a compound word storage means in which a plurality of compound words and constituent words constituting each compound word are associated with each other in the abbreviation determination device for determining whether or not text data is an abbreviation. A predetermined number of character data at the beginning is extracted from the text data, and it is determined whether or not a compound word including a constituent word having the extracted character data at the beginning is stored in the compound word storage means. . Further, when the abbreviation determination device determines that the compound word is stored in the compound word storage unit, the abbreviation determination device extracts a predetermined number of character data at the beginning from the text data excluding the extracted character data, It is determined whether or not the constituent word having the extracted character data at the head is included in the constituent words of the composite word determined to be stored in the composite word storage means. If it is determined that the constituent word is included, the abbreviation determination device determines that the text data is an abbreviation. Therefore, it is possible to easily determine whether or not the text data is an abbreviation generated by extracting a predetermined number of characters from the head of each of the two constituent words constituting the compound word.

  The abbreviation determination device according to the present invention extracts a number of character data corresponding to two syllables from the beginning of the text data, and whether the surname having the extracted character data at the beginning is stored in the personal name storage means, Alternatively, it is determined whether or not a compound word including a constituent word having extracted character data at the head is stored in the compound word storage unit. The abbreviation determination device extracts a number of character data corresponding to two syllables from the head of the text data excluding the extracted character data, and a name having the extracted character data at the head is stored in the personal name storage means. It is determined whether or not it is stored, or whether or not the constituent word having the extracted character data at the head is included in the constituent words of the composite word determined to be stored in the composite word storage means. Therefore, whether the text data is an abbreviation generated by extracting a number of characters corresponding to two syllables from the head of each surname and first name of a person name, or two constituent words constituting a compound word It is possible to easily determine whether or not the abbreviation is generated by extracting a number of characters corresponding to two syllables from the beginning.

  The abbreviation determination device according to the present invention extracts a number of character data corresponding to one syllable from the beginning of the text data, and whether the surname having the extracted character data at the beginning is stored in the personal name storage means, Alternatively, it is determined whether or not a compound word including a constituent word having extracted character data at the head is stored in the compound word storage unit. The abbreviation determination device extracts a number of character data corresponding to two syllables from the head of the text data excluding the extracted character data, and a name having the extracted character data at the head is stored in the personal name storage means. It is determined whether or not it is stored, or whether or not the constituent word having the extracted character data at the head is included in the constituent words of the composite word determined to be stored in the composite word storage means. Therefore, whether or not the text data is an abbreviation generated by extracting the number of characters corresponding to one syllable from the beginning of the surname of the person name and extracting the number of characters corresponding to two syllables from the beginning of the name. Or an abbreviation generated by extracting a number of characters corresponding to one syllable from the beginning of a constituent word constituting a compound word and extracting a number of characters corresponding to two syllables from the beginning of another constituent word. It is possible to easily determine whether or not.

  The abbreviation determination device according to the present invention divides document data including a plurality of text data into text data, extracts a predetermined number of character data at the head from the divided text data, and has the extracted character data at the head. It is determined whether the last name is stored in the personal name storage means, or whether a compound word including a constituent word having the extracted character data at the head is stored in the compound word storage means. Further, the abbreviation determination device extracts a predetermined number of character data at the beginning from the text data excluding the extracted character data, and whether a name having the extracted character data at the beginning is stored in the personal name storage means It is determined whether or not the constituent word having the extracted character data at the head is included in the constituent words of the compound word determined to be stored in the compound word storage means. If it is determined that the name is stored in the personal name storage means, or if it is determined that the constituent word is included, the abbreviation determination device determines that the text data is an abbreviation candidate. The abbreviation determination device stores a plurality of text data and co-occurrence data co-occurring with each text data in association with each other in the co-occurrence data storage means, and the text data determined as a candidate for the abbreviation Is obtained from the co-occurrence data storage means, and it is determined whether or not the acquired co-occurrence data is included in the text data in the document data, The text data determined to be an abbreviation candidate is determined to be an abbreviation. Therefore, the text data includes an abbreviation generated by extracting a predetermined number of characters from the beginning of each surname and first name of a person name, or a predetermined number of characters from the beginning of each of two constituent words constituting a compound word. It is possible to reliably determine whether the abbreviation is generated by extraction based on whether the data co-occurring with the text data is co-occurring data.

  The abbreviation determination method according to the present invention is an abbreviation determination method for determining whether or not text data is an abbreviation. In the abbreviation determination method, a predetermined number of character data at the beginning is extracted from the text data, and the extracted character data at the beginning. It is determined whether or not the last name is stored in the personal name storage means for storing the last name and the first name used for the personal name, and if it is determined that the last name is stored in the personal name storage means, the extracted characters A predetermined number of character data at the beginning is extracted from the text data excluding data, and it is determined whether or not a name having the extracted character data at the beginning is stored in the personal name storage means. If it is determined that the data is stored in the means, it is determined that the text data is an abbreviation.

  The abbreviation determination method according to the present invention is an abbreviation determination method for determining whether or not text data is an abbreviation. In the abbreviation determination method, a predetermined number of character data at the beginning is extracted from the text data, and the extracted character data at the beginning. It is determined whether or not a compound word including a constituent word is stored in a compound word storage unit that stores a plurality of compound words and constituent words constituting each compound word in association with each other, and the compound word is the compound word If it is determined that the word storage means stores, a predetermined number of character data at the beginning is extracted from the text data excluding the extracted character data, and the constituent word having the extracted character data at the beginning is combined It is determined whether or not it is included in a constituent word of a compound word determined to be stored in the word storage means, and when it is determined that the constituent word is included, the text data is an abbreviation judge.

  When the computer program according to the present invention is read by a computer and executed, the abbreviation determination device as described above can be realized by the computer.

  A text analysis apparatus according to the present invention is a text analysis apparatus that analyzes text data, includes any one of the abbreviation determination apparatuses described above, and stores morphemes and accent types in association with each other in a morpheme storage unit. Based on the stored contents of the means, the text data is divided into morphemes, and an accent type is assigned to each of the divided morphemes. The abbreviation determination device determines whether or not a morpheme that is not stored in the morpheme storage means is an abbreviation, and the text analysis device applies the morpheme determined to be an abbreviation by the abbreviation determination device. A predetermined accent type is given. Therefore, in the morphemes that are not stored in the morpheme storage means, different accent types are assigned to the morphemes determined to be abbreviations and other morphemes, so it is possible to assign accent types suitable for abbreviations. It becomes.

  The text analysis apparatus according to the present invention stores abbreviations and accent types in association with each other in the abbreviation storage means, and determines that the abbreviation is determined by the abbreviation determination apparatus based on the stored contents of the abbreviation storage means. Each of the morphemes is given an accent type, and a predetermined accent type is given to a morpheme that is not stored in the abbreviation storage means. Therefore, the corresponding accent type is assigned to the abbreviations stored in advance in the abbreviation storage means, and the predetermined accent type is assigned to the abbreviations not stored in the abbreviation storage means. Therefore, an accent type suitable for an abbreviation can be given.

  The text analysis apparatus according to the present invention is a text analysis apparatus for analyzing text data, and includes the abbreviation determination apparatus having the above-described co-occurrence data storage means, and stores the text data and the accent type in association with each other in the text storage means. ing. The abbreviation determination device divides the document data into text data based on the stored contents of the text storage means, and gives an accent type to each of the divided text data. The abbreviation determination device determines whether text data not stored in the text storage means is an abbreviation, and the text analysis device is determined to be an abbreviation by the abbreviation determination device. A predetermined accent type is given to the text data. Therefore, in the text data not stored in the text storage means, different accent types are assigned to the text data determined to be an abbreviation and the other text data, so an accent type suitable for the abbreviation is assigned. It becomes possible.

  The text analysis apparatus according to the present invention stores abbreviations and accent types in association with each other in the abbreviation storage means, and determines that the abbreviation is determined by the abbreviation determination apparatus based on the stored contents of the abbreviation storage means. An accent type is assigned to each piece of text data, and a predetermined accent type is assigned to text data not stored in the abbreviation storage means. Therefore, the corresponding accent type is assigned to the abbreviations stored in advance in the abbreviation storage means, and the predetermined accent type is assigned to the abbreviations not stored in the abbreviation storage means. Therefore, an accent type suitable for an abbreviation can be given.

  A speech synthesizer according to the present invention is a speech synthesizer that generates synthesized speech from text data. The speech synthesizer includes any of the text analysis devices described above, and is assigned to each morpheme divided by the morpheme dividing unit of the text analysis device. Prosody corresponding to each morpheme is generated based on the accent type, and synthesized speech is generated based on the generated prosody.

  A speech synthesizer according to the present invention is a speech synthesizer that generates synthesized speech from text data. The speech synthesizer includes a text analysis device that includes the abbreviation determination device having the above-described co-occurrence data storage unit. Prosody corresponding to each text data is generated based on the text data divided by the means and the accent type given to each text data by the text analysis device, and synthesized speech is generated based on the generated prosody.

  In the present invention, when the text data is configured by combining a predetermined number of characters at the beginning of the characters used for the surname of the person name and a predetermined number of characters at the beginning of the characters used for the name of the person name, the text The data is determined to be an abbreviation generated by omitting a person's name. In recent years, many abbreviations are a combination of the first two syllables of a family name and the first two syllables of a first name. Therefore, for example, when it is determined that text data formed by combining the first two characters used for the surname of the person name and the first two characters used for the name of the person name is an abbreviation, the text data However, it is possible to easily determine whether the abbreviation is, for example, Matsuken (Ken Matsuhira), Kim Taku (Takuya Kimura), Nakasho (Shoko Nakagawa), or the like.

  In the present invention, when the text data is configured by combining a predetermined number of characters at the head of each of the two constituent words constituting the compound word, the text data is generated by omitting the compound word. Judged to be a word. In recent years, many abbreviations are a combination of the first two syllables of the first word (component word) and the first two syllables of the second word (component word) constituting the formal name in the compound word. Therefore, for example, when it is determined that text data formed by combining the first two characters of the first constituent word and the first two characters of the second constituent word is an abbreviation, the text data is, for example, a cosplay Whether it is an abbreviation such as (costume play), continuous drama (continuous drama), muscle training (muscle training) or the like can be easily determined. Further, when it is determined that text data configured by combining the first character of the first constituent word and the first two characters of the second constituent word is an abbreviation, the text data is, for example, terrestrial digital ( It is possible to easily determine whether the abbreviation is “terrestrial digital broadcasting” or the like.

  In the present invention, when document data is divided into text data, and it is determined that the divided text data is an abbreviation candidate, the co-occurrence data co-occurred with the text data is a document including the text data. It is determined whether or not it is included in the data, and if it is included, this text data is determined to be an abbreviation. Therefore, this text data is determined to be an abbreviation only if the text data determined to be an abbreviation candidate co-occurs with co-occurrence data that is likely to co-occur if it is an original abbreviation. Thus, erroneous determination of abbreviations can be prevented. Specifically, for example, “samba” is registered in the co-occurrence data for “Matsuken”, and text data “Matsuken” in the document data “Matsuken danced samba” is a candidate for abbreviation. If it is determined, since “samba”, which is co-occurrence data of “Matsuken”, is included in the document data, this “Matsuken” can be specified as an abbreviation.

  In the present invention, the text data is divided into morphemes based on the storage contents of the morpheme storage means in which the morpheme and the accent type are stored in association with each other, and an accent type is assigned to each of the divided morphemes. When a morpheme that is not stored in the morpheme storage unit is determined to be an abbreviation by the abbreviation determination device, a predetermined accent type is assigned to the morpheme. In this way, in a morpheme that is not stored in the morpheme storage means, a different accent type is assigned to a morpheme determined to be an abbreviation and another morpheme, and therefore an accent type suitable for the abbreviation is given. Can do. Therefore, when a prosody corresponding to each morpheme is generated based on the accent type appropriately given to each morpheme in this way, and a synthesized speech is generated based on the generated prosody, an appropriate prosody can be generated. In addition, since an appropriate synthesized speech can be generated based on an appropriate prosody, a synthesized speech can be output with a correct accent even for morphemes (unknown words) that are not stored in the morpheme storage means. .

  In the present invention, an accent type is assigned to each morpheme determined to be an abbreviation by the abbreviation determination device based on the stored contents of the abbreviation storage means that stores the abbreviation and the accent type in association with each other. At the same time, a predetermined accent type is given to the morphemes that are not stored in the abbreviation storage means. Thus, since the corresponding accent type is assigned to each abbreviation stored in the abbreviation storage means, a more appropriate accent type can be assigned. Therefore, when a prosody corresponding to each morpheme is generated based on the accent type appropriately given to each morpheme in this way, and a synthesized speech is generated based on the generated prosody, an appropriate prosody can be generated. In addition, an appropriate synthesized speech can be generated based on an appropriate prosody.

  In the present invention, the document data is divided into text data based on the stored contents of the text storage means in which the text data and the accent type are stored in association with each other, and an accent type is assigned to each of the divided text data. When text data not stored in the text storage means is determined to be an abbreviation by the abbreviation determination device, a predetermined accent type is assigned to this text data. In this way, in the text data not stored in the text storage means, different accent types are given to the text data determined to be an abbreviation and the other text data, so an accent type suitable for the abbreviation is selected. Can be granted. Therefore, when a prosody corresponding to each text data is generated based on the accent type appropriately given to each text data in this way, and a synthesized speech is generated based on the generated prosody, an appropriate prosody is generated. Since it is possible to generate an appropriate synthesized speech based on an appropriate prosody, it is possible to output a synthesized speech with a correct accent even for morphemes (unknown words) that are not stored in the text storage means. Can do.

  In the present invention, an accent type is assigned to each text data determined to be an abbreviation by the abbreviation determination device based on the stored contents of the abbreviation storage means that stores the abbreviation and the accent type in association with each other. At the same time, a predetermined accent type is given to the text data not stored in the abbreviation storage means. Thus, since the corresponding accent type is assigned to each abbreviation stored in the abbreviation storage means, a more appropriate accent type can be assigned. Therefore, when a prosody corresponding to each text data is generated based on the accent type appropriately given to each text data in this way, and a synthesized speech is generated based on the generated prosody, an appropriate prosody is generated. In addition, an appropriate synthesized speech can be generated based on an appropriate prosody.

  Hereinafter, an abbreviation determination device, a text analysis device, and a speech synthesis device according to the present invention will be described in detail with reference to the drawings illustrating each embodiment. In the following embodiments, the computer program according to the present invention is read by a known personal computer or the like and executed by the CPU or the like of the personal computer, whereby the abbreviation determination device, text analysis device, and speech synthesis according to the present invention are performed. A configuration for realizing the apparatus will be described. However, the abbreviation determination device, the text analysis device, and the speech synthesis device according to the present invention may be realized by hardware having equivalent functions.

(Embodiment 1)
Hereinafter, a text analysis apparatus according to the present invention provided with an abbreviation determination apparatus according to the present invention will be described in detail with reference to the drawings showing Embodiment 1. FIG. 1 is a block diagram illustrating a configuration example of a text analysis apparatus according to the first embodiment. The text analysis apparatus 10 according to the first embodiment includes a control unit 1, a ROM 2, a RAM 3, an HDD 4, an operation unit 5, a display unit 6, and the like. These hardware units are connected to each other via a bus 1a. Yes.

  The control unit 1 is composed of a CPU (Central Processing Unit) or MPU (Micro Processor Unit) or the like, and reads and executes a control program stored in advance in the ROM 2 or HDD 4 to the RAM 3 as appropriate. Control the behavior. The ROM 2 stores various control programs necessary for operating the text analysis device 10 as an abbreviation determination device and a text analysis device of the present invention. The RAM 3 is configured by SRAM, flash memory, or the like, and temporarily stores various data generated when the control unit 1 executes the control program.

The operation unit 5 is a keyboard, a mouse, and the like, and includes various operation keys necessary for the user to operate the text analysis device 10. When each operation key is operated by the user, the operation unit 5 sends a control signal corresponding to the operated operation key to the control unit 1, and the control unit 1 performs processing corresponding to the control signal acquired from the operation unit 5. Execute.
The display unit 6 is a liquid crystal display, a CRT display, or the like. According to an instruction from the control unit 1, the operating state of the text analysis apparatus 10, information input via the operation unit 5, information to be notified to the user, and the like Is displayed.

  The HDD 4 is a large-capacity storage device. The HDD 4 includes various control programs and text data necessary for operating the text analysis device 10 as an abbreviation determination device and text analysis device of the present invention, as shown in FIG. A language dictionary 4a, a personal name dictionary 4b as shown in FIG. 3, a compound word dictionary 4c as shown in FIG. 4, screen information for notifying the user of various information, and the like are stored in advance.

  The language dictionary 4a, the personal name dictionary 4b, and the compound word dictionary 4c are not only stored in the HDD 4 in advance, but the text analysis device 10 can read data stored in an external memory (not shown). When a driver (not shown) is provided, various dictionaries stored in the external memory may be read out by the driver and stored in the HDD 4. When the text analysis device 10 includes a communication unit (not shown) that can be connected to a network such as the Internet, various dictionaries are downloaded from an external device via the network and stored in the HDD 4. Also good. The text data stored in the HDD 4 may be text data created by the text analysis device 10, and is created by an external device and stored in an external memory (not shown) or a network (not shown). The text data may be read by the text analysis device 10 via the network.

  FIG. 2 is a schematic diagram showing registered contents of the language dictionary 4a. As shown in FIG. 2, in the language dictionary (morpheme storage means) 4a, word (morpheme) notation, reading, and accent type are registered in association with each other. The part of speech of each morpheme may be registered in the language dictionary 4a.

  FIG. 3 is a schematic diagram showing the registered contents of the personal name dictionary 4b. The personal name dictionary (person name storage means) 4b stores the last name and the first name used for the personal name, and a list of last names as shown in FIG. 3 (a) and a list of names as shown in FIG. 3 (b). And have.

  FIG. 4 is a schematic diagram showing the registered contents of the compound word dictionary 4c. As shown in FIG. 4, the compound word dictionary (compound word storage means) 4c registers compound words, constituent words constituting the compound words, and readings of the constituent words in association with each other. In the first embodiment, each constituent word registered in the constituent word column of the compound word dictionary 4c will be described as an example in which the constituent words are registered in the order of composing the compound words. However, the registration order is not limited to this. . However, it is desirable that the constituent words sequentially registered in the constituent word column and the readings of the constituent words sequentially registered in the reading column are registered in association with each other.

  Hereinafter, in the text analysis apparatus 10 having the above-described configuration, various functions that are realized when the control unit 1 executes a control program stored in the ROM 2 and the HDD 4 will be described. FIG. 5 is a functional block diagram showing a functional configuration example of the text analysis apparatus 10. In the text analysis device 10 according to the first embodiment, the control unit 1 executes a control program stored in the ROM 2 and the HDD 4 to thereby execute a morphological analysis unit 11, an abbreviation determination unit (the abbreviation determination device according to the present invention). ) 12, Each function of the abbreviation accent giving unit 13 and the like is realized. In the following description, an example of processing in which the text analysis device 10 analyzes text data “Matsuken danced samba” will be described.

  The morpheme analysis unit (morpheme dividing means) 11 reads the text data stored in the HDD 4 from the HDD 4 to the RAM 3 and divides the text data read into the RAM 3 into morphemes based on the registered contents of the language dictionary 4a. An accent type is given to each of the morphemes. The morpheme analyzing unit 11 associates each divided morpheme with the accent type assigned to each morpheme and sends it to the abbreviation determination unit 12. Note that the morpheme analysis unit 11 does not know each morpheme and the accent type (unknown word) for the morpheme (unknown word) that could not be given the accent type based on the registered contents of the language dictionary 4a. Is transmitted to the abbreviation determination unit 12 in association with the information indicating.

  Since “Matsuken” is not registered in the language dictionary 4a of the first embodiment, “Matsuken” is treated as an unknown word. Therefore, the morphological analysis unit 11 converts the text data “Matsuken danced samba.” Into “Matsuken (unknown word), Ga (1 mora 0 type), Samba (3 mora 1 type), O (1 mora 0). The phonogram character string is sent to the abbreviation determination unit 12.

  In addition, information indicating that the accent type of each morpheme or an unknown word when each morpheme is an unknown word is given in parentheses in the phonetic character string. Strictly speaking, “dancing” is not a morpheme but is called a phrase or an accent phrase, but it is treated as one morpheme because it is not related to the essence of the present invention.

  The abbreviation determination unit 12 determines whether or not the morpheme determined to be an unknown word by the morpheme analysis unit 11 is an abbreviation based on the registered contents of the personal name dictionary 4b and the compound word dictionary 4c. Here, the abbreviation determination unit 12 determines whether or not “Matsuken” is an abbreviation. The details of the abbreviation determination process by the abbreviation determination unit 12 will be described later.

  When the abbreviation determination unit 12 determines that the morpheme determined to be an unknown word by the morpheme analysis unit 11 is an abbreviation, information indicating that the abbreviation is an abbreviation is given to the morpheme determined to be an abbreviation. Correspondingly, it is sent to the abbreviation accent giving unit 13. Specifically, when it is determined that “Matsuken” is an abbreviation, the abbreviation determination unit 12 determines whether “Matsuken (abbreviation), moth (1 mora 0 type), samba (3 mora 1 type), o ( A phonogram character string of “1 mora 0 type” and ottotta (4 mora 0 type) is sent to the abbreviation accent assigning unit 13.

  On the other hand, if the abbreviation determination unit 12 determines that the morpheme determined to be an unknown word by the morpheme analysis unit 11 is not an abbreviation, the abbreviation determination unit 12 has transmitted the morpheme determined to be not an abbreviation from the morpheme analysis unit 11. The information indicating that it is an unknown word is associated with it as it is and sent to the abbreviation accent assigning unit 13. Specifically, when it is determined that “Matsuken” is not an abbreviation, the abbreviation determination unit 12 reads “Matsuken (unknown word), Ga (1 mora 0 type), Samba (3 mora 1 type), o (1 The phonetic character string “Mora 0 type” and Odotta (4 mora 0 type) ”is sent to the abbreviation accent assigning unit 13.

  The abbreviation accent imparting unit (accent imparting means) 13 imparts a flat type (0 type) accent type (predetermined accent type) to the morpheme determined by the abbreviation determination unit 12 as an abbreviation. Specifically, when “Matsuken” is determined to be an abbreviation by the abbreviation determination unit 12, the abbreviation accent giving unit 13 sets a 4-mora 0 type accent type for the abbreviation “Matsuken”. And “4 mora 0 type” is assigned to the accent type of “Matsuken” in the phonetic character string sent from the abbreviation determination unit 12.

  As a result, the abbreviation accent assigning unit 13 can read “Matsuken (4 Mora 0 type), Ga (1 Mora 0 type), Samba (3 Mora 1 type), O (1 Mora 0 type), Odotta (4 Mora 0 type). ) "Phonetic character string is output. If it is determined by the abbreviation determination unit 12 that “Matsuken” is not an abbreviation, the abbreviation accent assignment unit 13 outputs the phonogram string sent from the abbreviation determination unit 12 as it is.

  The abbreviation determination process by the abbreviation determination unit 12 will be described below. In the first embodiment, the abbreviation determination unit 12 determines whether the morpheme determined to be an unknown word by the morpheme analysis unit 11 is an abbreviation obtained by omitting a person name, or an abbreviation obtained by omitting a compound word. It is determined whether or not. Note that recent abbreviations are a combination of the first two syllables (two letters) of the surname in a person's name and the first two syllables (two letters) of a first name, or the first word (composition) constituting a formal name in a compound word. The first two syllables (two letters) of the word) and the first two syllables (two letters) of the second word (constituent word), the first word (constituent word) constituting the official name in the compound word There are many combinations of the first syllable (one character) of and the first two syllables (two characters) of the second word (component word). Therefore, in the first embodiment, it is determined whether or not the morphemes determined as unknown words by the morpheme analyzer 11 are these abbreviations.

  The abbreviation determination unit 12 of the first exemplary embodiment first extracts a morpheme that is an unknown word by the morpheme analysis unit 11 from the phonetic character string transmitted from the morpheme analysis unit 11, and extracts the extracted unknown word (morpheme ) Is 3 syllables or 4 syllables. If the unknown word is three syllables, for example, an abbreviation combining the first letter of the surname in the name of the person and the first two letters of the name, or the first letter of the first constituent word constituting the compound word And the first two characters of the second component word are likely to be abbreviations. If the unknown word is four syllables, for example, the abbreviation combining the first two letters of the surname in the person's name and the first two letters of the first name, or the first two letters of the first constituent word constituting the compound word And the first two characters of the second component word are likely to be abbreviations.

  Therefore, when the extracted unknown word is three syllables, the abbreviation determination unit (first extraction means) 12 extracts the first one syllable from the unknown word (text data) of the three syllables. Then, the abbreviation determination unit 12 determines whether or not the extracted one syllable matches the first character of any surname registered in the surname list of the personal name dictionary 4b. That is, it is determined whether or not a surname having the first one syllable of the unknown word of three syllables is registered in the surname list. When it is determined that the extracted one syllable matches the first character of any surname registered in the surname list, the abbreviation determination unit (second extraction means) 12 excludes the extracted one syllable. The first two syllables are extracted from three syllable unknown words. That is, the two syllables behind the unknown three-syllable word are extracted.

  The abbreviation determination unit 12 determines whether or not the extracted two syllables match the first two characters of any name registered in the name list of the personal name dictionary 4b. That is, it is determined whether or not a name having the first two syllables behind the unknown word of the three syllables is registered in the name list. When it is determined that the extracted two syllables coincide with the first two characters of any name registered in the name list, the abbreviation determination unit (determination means) 12 abbreviates the unknown word as the abbreviation. It is determined that

  If the first syllable of the unknown word of the three syllables does not match the first character of any last name registered in the surname list, that is, it has the first syllable of the unknown word of the three syllable at the beginning. If the last name is not registered in the list of surnames in the personal name dictionary 4b, or if the last two syllables of the unknown word of the three syllables do not match the first two characters of any name registered in the list of first names, That is, when the name having the first two syllables after the unknown word of the three syllables is not registered in the name list of the personal name dictionary 4b, the abbreviation determination unit 12 performs the same processing based on the compound word dictionary 4c. Do.

  Specifically, the abbreviation determination unit (determination means) 12 extracts the first one syllable from the unknown word (text data) of the three syllables, and the extracted one syllable is in the component word column of the compound word dictionary 4c. It is determined whether or not the first one of the constituent words registered as the first constituent word matches the first character. That is, it is determined whether or not the constituent word having the first one syllable of the unknown words of the three syllables is registered as the first constituent word in the constituent word column of the compound word dictionary 4c. When it is determined that the extracted one syllable matches the first character of one of the constituent words registered as the first constituent word, the abbreviation determination unit 12 determines the second constituent of the compound word including this constituent word. A word is read from the compound word dictionary 4c, and the first two syllables, that is, the two syllables behind the unknown word of the three syllables are extracted from the unknown word of the three syllables excluding the extracted one syllable.

  The abbreviation determination unit 12 determines whether or not the extracted two syllables match the first two characters of the second constituent word read from the compound word dictionary 4c. That is, the constituent word having the first two syllables behind the unknown word of the three syllables is the second constituent word of the compound word formed by the constituent word having the first one syllable of the unknown word of the three syllables. Determine whether or not. When it is determined that the extracted two syllables match the first two characters of the second constituent word, the abbreviation determination unit (determination means) 12 determines that the unknown word is an abbreviation obtained by omitting the compound word.

  When the first syllable of the unknown word of 3 syllables does not match the first character of any of the registered words registered as the first component word, that is, the first syllable of the unknown word of 3 syllables starts Is not registered in the compound word dictionary 4c, or when the two syllables after the unknown word of the three syllables do not match the first two characters of the second constituent word, the abbreviation determination unit 12 This unknown word is determined not to be an abbreviation.

  Similarly, when the morpheme made into an unknown word by the morpheme analysis part 11 is 4 syllables, the abbreviation determination part 12 extracts the head 2 syllable from the unknown word of 4 syllables. Then, the abbreviation determination unit 12 determines whether or not the extracted two syllables match the first two characters of any surname registered in the surname list of the personal name dictionary 4b. That is, it is determined whether or not a surname having the first two syllables of the four syllable unknown words is registered in the surname list. When it is determined that the extracted two syllables match the first two letters of any surname registered in the surname list, the abbreviation determination unit 12 determines whether the extracted four syllable unknown words are excluded. Extract the first two syllables. That is, two syllables after the unknown word of four syllables are extracted.

  The abbreviation determination unit 12 determines whether or not the extracted two syllables match the first two characters of any name registered in the name list of the personal name dictionary 4b. That is, it is determined whether or not a name having the first two syllables behind the unknown word of four syllables is registered in the name list. When it is determined that the extracted two syllables match the first two characters of any name registered in the name list, the abbreviation determination unit 12 determines that the unknown word is an abbreviation obtained by omitting a person's name. To do.

  If the first two syllables of an unknown word in four syllables do not match the first two letters of any last name registered in the last name list, or the last two syllables of the unknown word in four syllables are a list of first names If neither of the first two characters registered in the name matches, the abbreviation determination unit 12 performs the same processing based on the compound word dictionary 4c.

  Specifically, the abbreviation determination unit 12 extracts the first two syllables from the four syllable unknown words, and the extracted two syllables are registered as the first component word in the component word column of the compound word dictionary 4c. It is determined whether or not the first two characters of any constituent word match. That is, it is determined whether or not the constituent word having the first two syllables at the head of the four-syllable unknown word is registered as the first constituent word in the constituent word column of the compound word dictionary 4c. When it is determined that the extracted two syllables coincide with the first two characters of any constituent word registered as the first constituent word, the abbreviation determination unit 12 determines the second constituent of the compound word including this constituent word. The word is read from the compound word dictionary 4c, and the first two syllables, that is, the two syllables behind the unknown word of the four syllables are extracted from the three syllable unknown words excluding the extracted two syllables.

  The abbreviation determination unit 12 determines whether or not the extracted two syllables match the first two characters of the second constituent word read from the compound word dictionary 4c. In other words, the constituent word having the first two syllables behind the unknown word of the four syllables is the second constituent word of the compound word formed by the constituent words having the first two syllables of the unknown four syllable word. Determine whether or not. When it is determined that the extracted two syllables match the first two characters of the second constituent word, the abbreviation determination unit (determination means) 12 determines that the unknown word is an abbreviation obtained by omitting the compound word.

  If the first two syllables of an unknown word in four syllables do not match the first two characters of any of the constituent words registered as the first constituent word, or the second syllable after the unknown word in four syllables is the second If the two words do not match the first two characters, the abbreviation determination unit 12 determines that the unknown word is not an abbreviation.

  The abbreviation determination unit 12 performs the above-described processing on all unknown words in the phonogram string sent from the morpheme analysis unit 11, and determines the morpheme (unknown word) determined to be an abbreviation. Is associated with information indicating that it is an abbreviation, and information indicating that it is an unknown word is directly associated with a morpheme determined not to be an abbreviation and sent to the abbreviation accent assigning unit 13.

  Below, the text analysis process by the text analysis apparatus 10 of this Embodiment 1 is explained in full detail based on a flowchart. FIG. 6 is a flowchart showing the procedure of the text analysis process. The following processing is executed by the control unit 1 in accordance with a control program stored in the ROM 2 or the HDD 4 of the text analysis device 10.

  When the user of the text analysis apparatus 10 instructs the execution of text analysis based on one text data by operating the operation unit 5, the control unit 1 reads the text data stored in the HDD 4 into the RAM 3 (S1). . The control unit 1 (morpheme analysis unit 11) divides the text data read into the RAM 3 into morphemes based on the registered contents of the language dictionary 4a, and assigns an accent type to each of the divided morphemes (S2). A phonetic character string in which an accent type is associated with is generated.

  The control unit 1 executes an abbreviation determination process based on the registered contents of the personal name dictionary 4b and the compound word dictionary 4c (S3), and the morpheme (unknown word) that could not be given the accent type in step S2 is an abbreviation. It is determined whether or not. The details of the abbreviation determination process will be described later with reference to FIGS. The control unit 1 (abbreviated word accent assigning unit 13) assigns a flat plate type (0 type) accent type to the morpheme determined to be an abbreviated word in step S3 (S4), and ends the text analysis process. .

  The abbreviation determination process (step S3 in FIG. 6) in the text analysis process described above will be described below. 7 to 11 are flowcharts showing the procedure of the abbreviation determination process. The following processing is executed by the control unit 1 (abbreviated word determination unit 12) according to a control program stored in the ROM 2 or the HDD 4 of the text analysis device 10.

  When the phonetic character string in which the accent type is associated with each morpheme is generated in step S2 in FIG. 6, the control unit 1 extracts an unknown word (morpheme) from the generated phonetic character string (S11). It is determined whether the unknown word is a three syllable (S12). If it is determined that there are three syllables (S12: YES), the control unit 1 extracts the first syllable from this unknown word (S13), and reads one word from the surname list in the personal name dictionary 4b (S14). The control unit 1 determines whether or not the one syllable extracted in step S13 matches the first character of the word (last name) read from the last name list (S15), and if it does not match (S15) : NO), it is determined whether or not all words have been read from the surname list (S16).

  When it is determined that reading of all words from the last name list has not been completed (S16: NO), the control unit 1 returns the process to step S14, and reads one word from the last name list in the personal name dictionary 4b (S14). ), And repeats the determination of whether or not the one syllable extracted in step S13 matches the first character of the word read from the surname list. If it is determined that all words have been read from the surname list (S16: YES), that is, the surname having the first syllable extracted in step S13 is registered in the surname list of the personal name dictionary 4b. If not, the control unit 1 moves the process to step S23. When it is determined that one syllable extracted in step S13 matches the first character of the word read from the surname list (S15: YES), that is, the surname having the first syllable extracted in step S13 at the head is the personal name dictionary 4b. Is registered in the list of surnames, the control unit 1 extracts the two syllables behind the unknown word extracted in step S11 (S17).

  The control unit 1 reads one word from the name list in the personal name dictionary 4b (S18), and whether the two syllables extracted in step S17 match the first two characters of the word (name) read from the name list. Is determined (S19). If it is determined that the extracted two syllables match the first two characters of the word read from the list of names (S19: YES), that is, the name having the two syllables extracted at step S17 at the head is the name of the name dictionary 4b. If registered in the list, the control unit 1 determines that the unknown word extracted in step S11 is an abbreviation (S20), and in step S2 in FIG. It is determined whether or not the above-described processing has been completed for all unknown words in the character string (S21). When it is determined that the process for all unknown words has been completed (S21: YES), the control unit 1 ends the abbreviation determination process, and when it is determined that the process has not been completed (S21: NO), the process proceeds to step S11. And the process described above is repeated for all unknown words in the phonetic character string.

  If it is determined that the two syllables extracted in step S17 do not match the first two characters of the word read from the name list (S19: NO), the control unit 1 has finished reading all the words from the name list. If it is determined whether or not all words have been read (S22: NO), the process returns to step S18, and one word is read from the list of names in the personal name dictionary 4b ( S18), the determination as to whether or not the two syllables extracted in step S17 match the first two characters of the word read from the list of names is repeated. If it is determined that all words have been read from the name list (S22: YES), that is, the name having the first two syllables extracted in step S17 is registered in the name list of the personal name dictionary 4b. If not, the control unit 1 moves the process to step S23.

  The control unit 1 extracts the first syllable from the unknown word extracted in step S11 (S23), and reads one constituent word registered as the first constituent word in the constituent word column of the compound word dictionary 4c ( S24). The control unit 1 determines whether or not the one syllable extracted in step S23 matches the first character of the constituent word read from the compound word dictionary 4c (S25), and determines that it does not match (S25: NO), it is determined whether or not all reading of the first constituent word from the compound word dictionary 4c has been completed (S26).

  If it is determined that the reading of all the first constituent words from the compound word dictionary 4c has not been completed (S26: NO), the control unit 1 returns the process to step S24, and enters the constituent word column of the compound word dictionary 4c. One constituent word registered as the first constituent word is read (S24), and it is determined whether or not one syllable extracted in step S23 matches the first character of the constituent word read from the compound word dictionary 4c. repeat. When it is determined that all the first constituent words have been read from the compound word dictionary 4c (S26: YES), that is, the first component word having one syllable extracted at step S23 is the compound word dictionary 4c. If not registered, the controller 1 determines that the unknown word extracted in step S11 is not an abbreviation (S31), and proceeds to step S21.

  When it is determined that one syllable extracted in step S23 matches the first character of the constituent word read from the compound word dictionary 4c (S25: YES), that is, the first component having one syllable extracted in step S23 at the head. When the word is registered in the compound word dictionary 4c, the control unit 1 extracts the two syllables behind the unknown word extracted in step S11 (S27). The control unit 1 reads out the second constituent word of the compound word including the first constituent word read out in step S24 from the compound word dictionary 4c (S28), and reads out the two syllables extracted in step S27 from the compound word dictionary 4c. It is determined whether or not the first two characters of the second constituent word match (S29).

  When the control unit 1 determines that the extracted two syllables match the first two characters of the second constituent word read from the compound word dictionary 4c (S29: YES), the unknown word extracted in step S11 is an abbreviation. It determines with there (S30), and transfers a process to step S21. If the control unit 1 determines that the extracted two syllables do not match the first two characters of the second constituent word read from the compound word dictionary 4c (S29: NO), the unknown word extracted in step S11 is It determines with it not being an abbreviation (S31), and transfers a process to step S21.

  On the other hand, when it is determined that the unknown word extracted in step S11 is not 3 syllables (S12: NO), the control unit 1 determines whether or not this unknown word is 4 syllables (S32) and is not 4 syllables. If it is determined (S32: NO), the process proceeds to step S21. When it is determined that the unknown word is four syllables (S32: YES), the control unit 1 extracts the first two syllables from the unknown word (S33), and reads one word from the surname list in the personal name dictionary 4b. (S34).

  The control unit 1 determines whether or not the two syllables extracted in step S33 match the first two characters of the word (last name) read from the last name list (S35), and determines that they do not match (S35) : NO), it is determined whether or not all words have been read from the surname list (S36). When it is determined that reading of all words from the last name list has not been completed (S36: NO), the control unit 1 returns the process to step S34, and reads one word from the last name list in the personal name dictionary 4b (S34). ), The determination as to whether or not the two syllables extracted in step S33 match the first two letters of the word read from the surname list is repeated. If it is determined that all words have been read from the surname list (S36: YES), that is, surnames having the first two syllables extracted in step S33 are registered in the surname list of the personal name dictionary 4b. If not, the control unit 1 moves the process to step S42.

  When it is determined that the two syllables extracted in step S33 match the first two letters of the word read from the surname list (S35: YES), that is, the surname having the two syllables extracted in step S33 at the head is the personal name dictionary 4b. Is registered in the list of surnames, the control unit 1 extracts the two syllables behind the unknown word extracted in step S11 (S37).

  The control unit 1 reads one word from the name list of the personal name dictionary 4b (S38), and whether or not the two syllables extracted in step S37 match the first two characters of the word (name) read from the name list. Is determined (S39). When it is determined that the extracted two syllables coincide with the first two characters of the word read from the list of names (S39: YES), that is, the name having the two syllables extracted at step S37 at the head is the name of the name dictionary 4b. When registered in the list, the control unit 1 determines that the unknown word extracted in step S11 is an abbreviation (S40), and proceeds to step S21.

  When it is determined that the two syllables extracted in step S37 do not match the first two letters of the word read from the name list (S39: NO), the control unit 1 has finished reading all the words from the name list. If it is determined whether or not all words have been read (S41: NO), the process returns to step S38, and one word is read from the list of names in the personal name dictionary 4b ( S38) The determination as to whether or not the two syllables extracted in step S37 match the first two characters of the word read from the list of names is repeated. If it is determined that all words have been read from the name list (S41: YES), that is, the name having the first two syllables extracted in step S37 is registered in the name list of the personal name dictionary 4b. If not, the control unit 1 moves the process to step S42.

  The control unit 1 extracts the first two syllables from the unknown word extracted in step S11 (S42), and reads one constituent word registered as the first constituent word in the constituent word column of the compound word dictionary 4c ( S43). The control unit 1 determines whether or not the two syllables extracted in step S42 match the first two characters of the constituent words read from the compound word dictionary 4c (S44), and determines that they do not match (S44: NO), it is determined whether or not all of the first constituent words have been read from the compound word dictionary 4c (S45).

  When it is determined that reading of all the first component words from the compound word dictionary 4c has not been completed (S45: NO), the control unit 1 returns the process to step S43, and enters the component word column of the compound word dictionary 4c. One constituent word registered as the first constituent word is read (S43), and it is determined whether or not the two syllables extracted in step S42 match the first two characters of the constituent word read from the compound word dictionary 4c. repeat. When it is determined that all the first constituent words have been read from the compound word dictionary 4c (S45: YES), that is, the first component word having two syllables extracted in step S42 is the compound word dictionary 4c. If not registered, the controller 1 determines that the unknown word extracted in step S11 is not an abbreviation (S50), and proceeds to step S21.

  When it is determined that the two syllables extracted in step S42 match the first two characters of the constituent word read from the compound word dictionary 4c (S44: YES), that is, the first configuration having the two syllables extracted in step S42 at the head. When the word is registered in the compound word dictionary 4c, the control unit 1 extracts the two syllables behind the unknown word extracted in step S11 (S46). The control unit 1 reads the second constituent word of the compound word including the first constituent word read out in step S43 from the compound word dictionary 4c (S47), and the two syllables extracted in step S46 are read out from the compound word dictionary 4c. It is then determined whether or not the first two characters of the second constituent word match (S48).

  When the control unit 1 determines that the extracted two syllables match the first two characters of the second constituent word read from the compound word dictionary 4c (S48: YES), the unknown word extracted in step S11 is an abbreviation. It determines with there (S49), and transfers a process to step S21. If the controller 1 determines that the extracted two syllables do not match the first two characters of the second constituent word read from the compound word dictionary 4c (S48: NO), the unknown word extracted in step S11 is It determines with it not being an abbreviation (S50), and transfers a process to step S21.

  As described above, in the text analysis apparatus 10 of the first embodiment, the abbreviation determination process is performed on the morpheme (unknown word) for which the accent type could not be assigned based on the language dictionary 4a, thereby omitting the person name. It is possible to determine whether the abbreviation is an abbreviation or an abbreviation with the compound word omitted. In addition, a predetermined accent type (flat plate type) is given to unknown words determined to be abbreviations, and unknown words that are not determined to be abbreviations, for example, from the back to the third mora By assigning an accent type with an accent nucleus, different accent types can be assigned to unknown words that are determined to be abbreviations and unknown words that are determined to be not abbreviations. Accent type can be given.

  In Embodiment 1 described above, each unknown word is an abbreviation that combines the first syllable of the last name in the personal name and the first two syllables of the first name, the first two syllables of the last name in the personal name, and the first two syllables of the first name Abbreviations combining the first one syllable of the first component word constituting the compound word and the first two syllables of the second component word, the first of the first component word constituting the compound word In this configuration, it is determined whether the abbreviation is a combination of two syllables and the first two syllables of the second constituent word. However, the number of characters constituting the abbreviation is not limited to these, and can be changed to an arbitrary number of characters by setting from the user via the operation unit 5.

  With the configuration described above, in the text analysis apparatus 10 of the first embodiment, Matsuken (Ken Matsuhira), Kim Tak (Takuya Kimura), Naka Show (Shoko Nakagawa), Cosplay (costume play), series drama (continuous drama), muscle training ( When abbreviations such as muscle training) and terrestrial digital (terrestrial digital broadcasting) are included in the text data, each word can be appropriately determined as an abbreviation.

(Embodiment 2)
Hereinafter, a text analysis device according to the present invention provided with an abbreviation determination device according to the present invention will be described in detail with reference to the drawings showing Embodiment 2. Since the text analysis apparatus according to the second embodiment can be realized by the same configuration as the text analysis apparatus 10 according to the first embodiment, the same reference numerals are given to the same configurations and the description thereof is omitted. .

  FIG. 12 is a functional block diagram illustrating a functional configuration example of the text analysis apparatus 10 according to the second embodiment. In the text analysis apparatus 10 according to the second embodiment, the control unit 1 executes a control program stored in the ROM 2 or the HDD 4, and similarly to the text analysis apparatus 10 according to the first embodiment described above, the morpheme analysis unit 11. Each function of the abbreviation determination unit 12 and the abbreviation accent assignment unit 13 is realized.

  A co-occurrence dictionary 4d as shown in FIG. 13 is stored in the HDD 4 of the text analysis apparatus 10 of the second embodiment. FIG. 13 is a schematic diagram showing the registered contents of the co-occurrence dictionary 4d. As shown in FIG. 13, in the co-occurrence dictionary (co-occurrence data storage means) 4d, words (text data) are read and co-occurrence words (co-occurrence data) co-occurring with each word are registered in association with each other. Has been.

The text analysis process performed by the text analysis apparatus 10 according to the second embodiment having the above-described configuration will be described below. In the following, description will be given by taking as an example a process in which the text analysis apparatus 10 analyzes text data of “Matsuken danced a samba”.
The morpheme analysis unit 11 of the second embodiment reads text data (document data) stored in the HDD 4 from the HDD 4 to the RAM 3 in the same manner as the morpheme analysis unit 11 of the first embodiment described above. ) Based on the registered content of 4a, the text data (document data) read to the RAM 3 is divided into morphemes (text data), and an accent type is given to each of the divided morphemes. Specifically, as described in the first embodiment, the morphological analysis unit 11 “Matsuken (unknown word) • Ga (1 mora 0 type) • Samba (3 mora 1 type) • O (1 mora 0 type) A phonogram string of “Odotta (4 mora 0 type)” is generated and sent to the abbreviation determination unit 12.

  The abbreviation determination unit 12 of the second embodiment first performs the same processing as the abbreviation determination unit 12 of the first embodiment described above, and based on the registered contents of the personal name dictionary 4b or the compound word dictionary 4c, the morpheme analysis unit 11 is determined whether or not the unknown word (morpheme) in the phonogram string sent from 11 is a candidate for an abbreviation obtained by omitting a person name or a compound word. Here, as described in the first embodiment, the unknown word “Matsuken” is determined to be an abbreviation candidate. When the abbreviation determination unit 12 of the second embodiment determines that each unknown word is a candidate for an abbreviation, the formal name for the abbreviation is based on the registered contents of the personal name dictionary 4b or the compound word dictionary 4c. Get it.

  Specifically, the abbreviation determination unit 12 determines that the first one syllable (or two syllables) of the unknown word is the first character (or 2) of any surname registered in the surname list of the personal name dictionary 4b. Character), or when matching the first character (or two characters) of any of the first constituent words registered in the compound word dictionary 4c, the corresponding surname or constituent word is read out. Stored in the RAM 3. The abbreviation determination unit 12 also determines that the two syllables behind the unknown word match the first two characters of any name registered in the name list of the personal name dictionary 4b, or the compound word dictionary 4c. When the first two characters of the second component word registered in the column match, the corresponding name or component word is read out and stored in the RAM 3. Thereby, the abbreviation determination part 12 can acquire the formal name with respect to the unknown word determined to be an abbreviation candidate. Here, the official name “Matsudai Raken” of the abbreviation candidate “Matsuken” is acquired.

  Next, the abbreviation determination unit 12 acquires a co-occurrence word for this abbreviation from the co-occurrence dictionary 4d based on the formal name of the unknown word determined to be an abbreviation candidate. Here, the co-occurrence words “samba” and “general shogun” for the official name “Matsudai Laken” are acquired. When the abbreviation determination unit 12 determines whether or not the co-occurrence word acquired from the co-occurrence dictionary 4d is included in the phonogram string transmitted from the morpheme analysis unit 11, and determines that it is included The unknown word determined to be an abbreviation candidate is determined to be an abbreviation. Then, the abbreviation determination unit 12 displays a table of “Matsuken (abbreviation), Ga (1 mora 0 type), Samba (3 mora 1 type), E (1 mora 0 type), Odotta (4 mora 0 type)”. The phonetic character string is sent to the abbreviation accent assigning unit 13.

  On the other hand, when it is determined that the syllable character string transmitted from the morphological analysis unit 11 does not include the co-occurrence word acquired from the co-occurrence dictionary 4d, the abbreviation determination unit 12 is an abbreviation candidate. It is determined that an unknown word determined to be an abbreviation is not an abbreviation. For example, “Matsuken (unknown word), Ga (1 mora 0 type), Samba (3 mora 1 type), O (1 mora 0 type), Odotta (4 "Mora 0 type" "is sent to the abbreviation accent giving unit 13.

  The abbreviation accent assigning unit 13 is a flat type (0 type) accent for the morphemes determined to be abbreviations by the abbreviation determining unit 12, similarly to the abbreviation accent providing unit 13 of the first embodiment described above. Give type.

  As described above, for unknown words determined to be abbreviation candidates based on the registered contents of the personal name dictionary 4b and the compound word dictionary 4c, the unknown word is included in the text data (document data) including the unknown word. If a co-occurrence word that co-occurs with a word is not included, it is highly likely that this unknown word is not the abbreviation, so it is determined that it is not the abbreviation. Thereby, erroneous determination of abbreviations can be prevented, and a predetermined accent type can be given only to abbreviations.

  Below, the text analysis process by the text analysis apparatus 10 of this Embodiment 2 is explained in full detail based on a flowchart. FIG. 14 is a flowchart showing the procedure of text analysis processing. The following processing is executed by the control unit 1 in accordance with a control program stored in the ROM 2 or the HDD 4 of the text analysis device 10.

  When the user of the text analysis apparatus 10 instructs the execution of text analysis based on one text data by operating the operation unit 5, the control unit 1 reads the text data stored in the HDD 4 into the RAM 3 (S61). . The control unit 1 (morpheme analysis unit 11) divides the text data read into the RAM 3 into morphemes based on the registered contents of the language dictionary 4a, and assigns an accent type to each of the divided morphemes (S62). A phonetic character string in which an accent type is associated with is generated.

  The control unit 1 (abbreviated word determining unit 12) executes the abbreviated word determining process based on the registered contents of the personal name dictionary 4b and the compound word dictionary 4c (S63), and the morpheme that could not give the accent type in step S62 (unknown Word) is a candidate for abbreviation. The abbreviation determination processing of the second embodiment is the same as the processing described in the first embodiment based on FIGS. 7 to 11, but the control unit 1 (abbreviation determination unit 12) of the second embodiment is When it is determined that each unknown word is a candidate for an abbreviation, the official name for the abbreviation is acquired based on the registered contents of the personal name dictionary 4b or the compound word dictionary 4c (S64).

  The control unit 1 acquires a co-occurrence word for the abbreviation from the co-occurrence dictionary 4d based on the official name acquired in step S64 (S65), and includes the co-occurrence dictionary in the phonogram string generated in step S62. It is determined whether or not the co-occurrence word acquired from 4d is included (S66). When it is determined that the co-occurrence word is included (S66: YES), the control unit 1 determines that the unknown word determined to be an abbreviation candidate in step S63 is an abbreviation (S67). Further, the control unit 1 (abbreviated word accent assigning unit 13) assigns a flat type (0 type) accent type to the morpheme determined to be an abbreviated word in step S67 (S68), and performs text analysis processing. finish.

  On the other hand, when it is determined that the co-occurrence word is not included in the phonogram string generated in step S62 (S66: NO), the control unit 1 determines the unknown word determined to be an abbreviation candidate in step S63. If it is not an abbreviation, it is determined (S69), and the text analysis process is terminated.

  As described above, in the text analysis apparatus 10 according to the second embodiment, text data (document data) including an unknown word determined to be an abbreviation candidate based on registered contents of the personal name dictionary 4b and the compound word dictionary 4c. If a co-occurrence word that co-occurs with the abbreviation is not included, it is highly likely that the unknown word is not the abbreviation, so it is determined that it is not the abbreviation. Therefore, if an unknown word determined to be a candidate for an abbreviation is used together with a co-occurrence word that is likely to co-occur if it is the abbreviation, the unknown word is determined to be an abbreviation. If it is not used with a co-occurrence word, it is determined that this unknown word is not an abbreviation.

  Thereby, erroneous determination of an abbreviation can be prevented, and it can be determined that only an abbreviation is a true abbreviation. Therefore, a predetermined accent type is assigned to an unknown word determined to be an abbreviation, and an accent type having an accent nucleus is assigned to the unknown word determined to be not an abbreviation, for example, at the third mora from the back. Thus, different accent types can be assigned to true abbreviations and unknown words that are not abbreviations, so that an appropriate accent type can be assigned to each.

(Embodiment 3)
Hereinafter, a text analysis apparatus according to the present invention provided with an abbreviation determination apparatus according to the present invention will be described in detail with reference to the drawings showing a third embodiment. Note that the text analysis apparatus according to the third embodiment can be realized by the same configuration as the text analysis apparatus 10 according to the first embodiment described above. .

  In the text analysis device 10 according to the first embodiment described above, the control unit 1 (abbreviated word accent assigning unit 13) performs an operation on unknown words determined to be abbreviated words based on the registered contents of the personal name dictionary 4b and the compound word dictionary 4c. It was configured to give a flat accent type.

  In the text analysis apparatus 10 according to the third embodiment, an exception abbreviation dictionary 4e in which an accent type for each abbreviation is registered is stored in the HDD 4 in advance, and the control unit 1 (abbreviation accent assignment unit 13) is a personal name dictionary. Among unknown words (abbreviated words) determined to be abbreviations based on the registered contents of 4b and compound word dictionary 4c, abbreviations registered in exception abbreviation dictionary 4e are registered in exception abbreviation dictionary 4e. The accent type is given, and a flat type accent type is given to abbreviations not registered in the exception abbreviation dictionary 4e.

  FIG. 15 is a functional block diagram illustrating a functional configuration example of the text analysis apparatus 10 according to the third embodiment. In the text analysis device 10 according to the third embodiment, the control unit 1 executes a control program stored in the ROM 2 or the HDD 4, and similarly to the text analysis device 10 according to the first embodiment described above, the morpheme analysis unit 11. Each function of the abbreviation determination unit 12 and the abbreviation accent assignment unit 13 is realized.

  Note that an exception abbreviation dictionary 4e as shown in FIG. 16 is stored in the HDD 4 of the text analysis apparatus 10 of the third embodiment. FIG. 16 is a schematic diagram showing the registered contents of the exception abbreviation dictionary 4e. As shown in FIG. 16, in the exception abbreviation dictionary (abbreviation storage means) 4e, abbreviations and accent types of abbreviations are registered in association with each other.

The text analysis process performed by the text analysis apparatus 10 according to the third embodiment having the above-described configuration will be described below. In the following, description will be given by taking as an example a process in which the text analysis device 10 analyzes text data of “Nakasho has cosplayed”.
Similar to the morpheme analyzer 11 of the first embodiment described above, the morpheme analyzer 11 of the third embodiment reads text data stored in the HDD 4 from the HDD 4 to the RAM 3, and based on the registered contents of the language dictionary 4a. The text data read into the RAM 3 is divided into morphemes, and an accent type is given to each of the divided morphemes. Here, the morphological analysis unit 11 is a phonetic character of “Nakasho (unknown word), Ga (1 mora 0 type), Cosplay (unknown word), Oh (1 mora 0 type), Sita (2 mora 0 type)”. Generate a column.

  The abbreviation determination unit 12 of the third embodiment is generated by the morpheme analysis unit 11 based on the registered contents of the personal name dictionary 4b and the compound word dictionary 4c, similarly to the abbreviation determination unit 12 of the first embodiment. It is determined whether or not the unknown word in the phonetic character string is an abbreviation. The abbreviation determination unit 12 associates information indicating an abbreviation with an unknown word determined to be an abbreviation, and identifies an unknown word for an unknown word determined to be not an abbreviation. A phonetic character string in which the indicated information is associated is generated. Here, the abbreviation determination unit 12 represents the phonetics of “Nakasho (abbreviation), Ga (1 mora 0 type), Cosplay (abbreviation), Oh (1 mora 0 type), Sita (2 mora 0 type)”. Generate a string.

  The abbreviation accent assigning unit 13 according to the third embodiment extracts abbreviations from the phonetic character string generated by the abbreviation determination unit 12, and whether or not the extracted abbreviations are registered in the exception abbreviation dictionary 4e. Determine whether. When the extracted abbreviation is registered in the exception abbreviation dictionary 4e, the abbreviation accent assigning unit 13 reads out the accent type corresponding to the extracted abbreviation from the exception abbreviation dictionary 4e, and for the extracted abbreviation Give. On the other hand, when the extracted abbreviation is not registered in the exception abbreviation dictionary 4e, the abbreviation accent assigning unit 13 assigns a flat (0 type) accent type to the extracted abbreviation.

  Here, since the abbreviation “Nakasho” is registered in the exception abbreviation dictionary 4e, the abbreviation “Nakasho” is given the accent type “4 mora type 2”, and the abbreviation “cosplay” is omitted. Since it is not registered in the word dictionary 4e, a flat accent type is given to the abbreviation “cosplay”. That is, here, the abbreviation accent giving unit 13 is “Nakasho (4 Mora 2 Type), Ga (1 Mora 0 Type), Cosplay (4 Mora 0 Type), Oh (1 Mora 0 Type), Sita (2 Mora "0 type)" is generated.

  The abbreviation accent assigning unit 13 is registered in the exception abbreviation dictionary 4e by executing the above-described processing for all abbreviations in the phonogram string generated by the abbreviation determination unit 12. A pre-registered accent type can be assigned to the abbreviation. Thereby, a more appropriate accent type can be given to an abbreviation that is not a flat accent type, such as the abbreviation “Nakasho”.

  Below, the text analysis process by the text analysis apparatus 10 of this Embodiment 3 is explained in full detail based on a flowchart. FIG. 17 is a flowchart showing the procedure of the text analysis process. The following processing is executed by the control unit 1 in accordance with a control program stored in the ROM 2 or the HDD 4 of the text analysis device 10.

  When the user of the text analysis device 10 instructs the execution of text analysis based on one text data by operating the operation unit 5, the control unit 1 reads the text data stored in the HDD 4 into the RAM 3 (S71). . The control unit 1 (morpheme analysis unit 11) divides the text data read into the RAM 3 into morphemes based on the registered contents of the language dictionary 4a, and assigns an accent type to each of the divided morphemes (S72). A phonetic character string in which an accent type is associated with is generated.

  The control unit 1 (abbreviated word determining unit 12) executes an abbreviated word determining process based on the registered contents of the personal name dictionary 4b and the compound word dictionary 4c (S73), and the morpheme that could not give the accent type in step S72 (unknown Word) is a candidate for abbreviation. The abbreviation determination process of the third embodiment is the same as the process described in the first embodiment based on FIGS.

  The control unit 1 (abbreviated word accent assigning unit 13) extracts abbreviations from the phonetic character string generated in step S73 (S74), and determines whether or not the extracted abbreviations are registered in the exception abbreviation dictionary 4e. Judgment is made (S75). When it is determined that the extracted abbreviation is registered in the exception abbreviation dictionary 4e (S75: YES), the control unit 1 reads the accent type corresponding to the extracted abbreviation word from the exception abbreviation dictionary 4e and extracts it. It is given to the abbreviation (S76).

  When it is determined that the extracted abbreviation is not registered in the exception abbreviation dictionary 4e (S75: NO), the control unit 1 gives a flat (0 type) accent type to the extracted abbreviation ( S77). The control unit 1 determines whether or not the above-described processing has been completed for all abbreviations in the phonogram string generated in step S73 (S78), and determines that it has not ended (S78). (S78: NO), the process returns to step S74, abbreviations are extracted from the phonetic character string (S74), and the determination of whether or not the extracted abbreviations are registered in the exception abbreviation dictionary 4e is repeated. When it is determined that the processing for all abbreviations in the phonetic character string has been completed (S78: YES), the control unit 1 ends the above-described text analysis processing.

  As described above, in the text analysis apparatus 10 of the third embodiment, the accent type for an unknown word determined to be an abbreviation based on the registered contents of the personal name dictionary 4b and the compound word dictionary 4c is stored in the exception abbreviation dictionary 4e. When registered, by assigning this accent type to an abbreviation, an appropriate accent type can be assigned to an abbreviation of an accent type that is not a flat plate type. Even if the accent type for an unknown word determined to be an abbreviation based on the registered contents of the personal name dictionary 4b and the compound word dictionary 4c is not registered in the exception abbreviation dictionary 4e, By giving a flat accent type, an accent type different from an unknown word that is not an abbreviation can be given. Therefore, when a prosody corresponding to each abbreviation is generated based on the accent type appropriately given by each abbreviation, and a synthesized speech is generated based on the generated prosody, an appropriate prosody can be generated, and An appropriate synthesized speech can be generated based on an appropriate prosody.

  The third embodiment described above corresponds to the abbreviations registered in the exception abbreviation dictionary 4e by the control unit 1 (abbreviated word accent assigning unit 13) in the text analysis apparatus 10 of the first embodiment. A configuration in which a flat accent type is assigned to an abbreviation that is assigned an accent type and is not registered in the exception abbreviation dictionary 4e has been described as a modification of the first embodiment. However, a similar modification can be applied to the text analysis apparatus 10 of the second embodiment. That is, when the configuration of the third embodiment is applied to the text analysis device 10 of the second embodiment, each of the abbreviations specified based on the co-occurrence dictionary 4d as well as the personal name dictionary 4b and the compound word dictionary 4c. Appropriate accent types can be given to words.

(Embodiment 4)
Hereinafter, a speech synthesizer according to the present invention will be described in detail with reference to the drawings showing a fourth embodiment. Note that the speech synthesizer of the fourth embodiment includes the configuration of the text analysis device 10 of the first embodiment described above, and the same components are denoted by the same reference numerals and description thereof is omitted. FIG. 18 is a block diagram illustrating a configuration example of a speech synthesizer according to the fourth embodiment. The speech synthesis apparatus 100 according to the fourth embodiment includes a speech output unit 7 in addition to the control unit 1, the ROM 2, the RAM 3, the HDD 4, the operation unit 5, and the display unit 6 illustrated in FIG. Each part of the wear is connected to each other via the bus 1a.

The audio output unit 7 includes an audio amplification circuit, a speaker, and the like, and outputs audio based on audio information (audio waveform) stored in the HDD 4, for example.
In the HDD 4, text data, a language dictionary 4a as shown in FIG. 2, a personal name dictionary 4b as shown in FIG. 3, a compound word dictionary 4c as shown in FIG. 4, and various kinds of information are notified to the user. In addition to the screen information, various control programs necessary for operating the speech synthesizer 100 as the speech synthesizer of the present invention, the prosody generation rule dictionary 4f, the waveform dictionary 4g, and the like are stored in advance.

  Although details of the prosody generation rule dictionary 4f and the waveform dictionary 4g are not shown, rules for generating prosody data based on the reading and accent type of each morpheme are registered in the prosody generation rule dictionary 4f. In the waveform dictionary 4g, a waveform group of each voice is registered corresponding to a sentence composed of a plurality of phonemes (phoneme strings).

  Hereinafter, in the speech synthesizer 100 having the above-described configuration, various functions realized by the control unit 1 executing a control program stored in the ROM 2 and the HDD 4 will be described. FIG. 19 is a functional block diagram illustrating a functional configuration example of the speech synthesizer 100 according to the fourth embodiment. In the speech synthesizer 100 according to the fourth embodiment, the control unit 1 executes the control program stored in the ROM 2 and the HDD 4 to thereby execute the text analysis device 10, the prosody generation unit 20, and the waveform generation described above according to the first embodiment. Each function of the unit 30 and the like is realized.

  The prosody generation unit 20 generates prosody data corresponding to the phonetic character string generated by the text analysis device 10 according to the registered contents of the prosody generation rule dictionary 4f. Specifically, the prosody generation unit 20 generates prosody data corresponding to the reading and accent type of each morpheme in the phonogram string sent from the text analysis device 10.

  The waveform generation unit 30 converts the prosody data generated by the prosody generation unit 20 into a speech waveform based on the registered contents of the waveform dictionary 4g to generate a synthesized speech. Specifically, the waveform generation unit 30 extracts a speech waveform corresponding to each morpheme in the prosody data sent from the prosody generation unit 20 from the waveform dictionary 4g, and synthesizes based on the extracted speech waveform and prosody data. Generate audio. The synthesized voice generated by the waveform generation unit 30 is temporarily stored in the RAM 3 or the HDD 4 and then sent to the voice output unit 7 at a predetermined timing according to control by the control unit 1. Is done.

  With the above-described configuration, the speech synthesizer 100 according to the fourth embodiment generates text phonetic character strings by analyzing text data using the text analysis device 10 and generates synthesized speech corresponding to the generated phonetic character strings. Can do. Therefore, as in the text analysis apparatus 10 of the first embodiment described above, an appropriate accent type is assigned to each morpheme based on the language dictionary 4a, and an abbreviation is used for unknown words that are not registered in the language dictionary 4a. If an appropriate accent type is given to each morpheme by giving an accent type (flat plate type) suitable for abbreviations, an appropriate composition based on such an accent type is given. Voice can be generated.

  Below, the synthetic | combination speech production | generation process by the speech synthesizer 100 of this Embodiment 4 is explained in full detail based on a flowchart. FIG. 20 is a flowchart showing a procedure of synthetic speech generation processing. Note that the following processing is executed by the control unit 1 in accordance with a control program stored in the ROM 2 or the HDD 4 of the speech synthesizer 100.

  When the user of the speech synthesizer 100 instructs the execution of the synthetic speech generation process based on one text data by operating the operation unit 5, the control unit 1 reads the text data stored in the HDD 4 into the RAM 3. (S81). The control unit 1 (morpheme analysis unit 11) divides the text data read into the RAM 3 into morphemes based on the registered contents of the language dictionary 4a, and assigns an accent type to each of the divided morphemes (S82). A phonetic character string in which an accent type is associated with is generated.

  The control unit 1 (abbreviated word determining unit 12) executes an abbreviated word determining process based on the registered contents of the personal name dictionary 4b and the compound word dictionary 4c (S83), and the morpheme that could not be given the accent type in step S82 (unknown Word) is an abbreviation. Note that the abbreviation determination process of the fourth embodiment is the same as the process described in the first embodiment with reference to FIGS.

  The control unit 1 (abbreviated word accent assigning unit 13) assigns a flat plate type (0 type) accent type to the morpheme determined to be an abbreviated word in step S83 (S84). The control unit 1 (prosody generation unit 20) generates prosody data corresponding to the obtained phonetic character string based on the registered contents of the prosody generation rule dictionary 4f (S85). The control unit 1 (waveform generation unit 30) generates a speech waveform from the generated prosodic data based on the registered contents of the waveform dictionary 4g (S86), and ends the synthetic speech (speech waveform) generation process.

  As described above, in the speech synthesizer 100 according to the fourth embodiment, as described in the first embodiment, whether or not a morpheme (unknown word) for which an accent type could not be assigned based on the language dictionary 4a is an abbreviation. By assigning an appropriate accent type to each unknown word depending on whether it is an abbreviation, it is possible to generate an appropriate prosody based on the appropriately assigned accent type, An appropriate synthesized speech can be generated based on an appropriate prosody. Therefore, it is possible to generate synthesized speech with correct accent and intonation even for unknown words that are not registered in the language dictionary 4a.

  In the fourth embodiment described above, the present invention has been described by taking the speech synthesizer 100 including the text analysis device 10 of the first embodiment as an example, but the speech synthesizer of the present invention is the text of the second and third embodiments described above. It can also be set as the structure provided with the analysis apparatus 10. FIG. In the case of the configuration including the text analysis device 10 according to the third embodiment described above, the speech synthesizer 100 is an unknown word determined to be an abbreviation by the abbreviation determination process and is registered in the exception abbreviation dictionary 4e. Since an abbreviation that has been registered can be given a pre-registered accent type, a more appropriate accent type can be assigned to an abbreviation that is not a flat accent type. A more appropriate synthesized speech can be generated based on the type.

  As described above, the abbreviation determination device according to the present invention is an abbreviation in which a morpheme (unknown word) that is not registered in the language dictionary 4a is an abbreviation in which a person name is omitted or a compound word is omitted. In such a case, a flat accent type is given to such abbreviations, and an accent type having an accent nucleus is assigned to the third unknown mora, for example, as is conventionally done for other unknown words. Thus, different accent types can be assigned to abbreviations and unknown words that are not abbreviations.

  Note that abbreviations omitting names of persons and abbreviations omitting compound words often have a flat accent type. Therefore, by adding a flat accent type to such abbreviations, Even a document including abbreviated abbreviations can generate synthesized speech with appropriate accents, and more natural speech can be output based on such synthesized speech. The present invention is not configured to register new abbreviations that appear every day in the dictionary, but is configured to determine whether or not the abbreviation is based on the registered contents of the personal name dictionary 4b and the compound word dictionary 4c. Therefore, it is possible to appropriately determine the abbreviations in the document without assigning the abbreviations to the dictionary and to assign an appropriate accent type to the abbreviations.

(Appendix 1)
In an abbreviation determination device that determines whether text data is an abbreviation,
Personal name storage means for storing the last name and first name used for the personal name;
First extraction means for extracting a predetermined number of character data at the beginning from the text data;
Means for determining whether or not a last name having the character data extracted by the first extraction means at the head is stored in the personal name storage means;
A second extraction means for extracting a predetermined number of character data at the beginning from the text data excluding the character data extracted by the first extraction means when it is determined that the last name is stored in the personal name storage means; ,
Means for determining whether or not a name having character data extracted by the second extraction means at the head is stored in the personal name storage means;
An abbreviation determination device comprising: determination means for determining that the text data is an abbreviation when it is determined that the name is stored in the personal name storage means.

(Appendix 2)
In an abbreviation determination device that determines whether text data is an abbreviation,
A compound word storage means for storing a plurality of compound words and constituent words constituting each compound word in association with each other;
First extraction means for extracting a predetermined number of character data at the beginning from the text data;
Determining means for determining whether or not a compound word including a constituent word having the character data extracted by the first extracting means is stored in the compound word storage means;
A second extraction for extracting a predetermined number of character data from the text data excluding the character data extracted by the first extraction means when it is determined that the compound word is stored in the compound word storage means; Means,
It is determined whether or not the constituent word having the character data extracted by the second extraction means at the head is included in the constituent words of the composite word determined by the determination means to be stored in the compound word storage means. Means,
An abbreviation determination device comprising: a determination unit that determines that the text data is an abbreviation when it is determined that the constituent word is included.

(Appendix 3)
The first extraction means is configured to extract a number of character data corresponding to two syllables from the beginning of the text data,
The second extraction means is configured to extract a number of character data corresponding to two syllables from the beginning of the text data excluding the character data extracted by the first extraction means. The abbreviation determination device according to appendix 1 or 2.

(Appendix 4)
The first extraction means is configured to extract a number of character data corresponding to one syllable from the beginning of the text data,
The second extraction means is configured to extract a number of character data corresponding to two syllables from the beginning of the text data excluding the character data extracted by the first extraction means. The abbreviation determination device according to appendix 1 or 2.

(Appendix 5)
A dividing unit for dividing document data including a plurality of text data into text data;
The first extracting means is configured to extract a predetermined number of character data at the beginning from each of the divided text data,
The second extraction means is configured to extract a predetermined number of character data at the beginning from each of the divided text data excluding the character data extracted by the first extraction means,
The determination means is configured to determine whether each of the divided text data is an abbreviation candidate,
Co-occurrence data storage means for storing a plurality of text data and co-occurrence data co-occurring with each text data in association with each other;
Means for acquiring from the co-occurrence data storage means co-occurrence data corresponding to the text data determined by the determination means to be abbreviation candidates;
Means for determining whether the text data in the document data includes co-occurrence data acquired from the co-occurrence data storage means;
Appendices 1 to 4, further comprising: means for determining, when it is determined that the co-occurrence data is included, the text data determined by the determining means as an abbreviation candidate as an abbreviation. An abbreviation determination device according to any one of the above.

(Appendix 6)
In the abbreviation determination method for determining whether text data is an abbreviation,
A first extraction step of extracting a predetermined number of character data at the beginning from the text data;
A step of determining whether or not the surname having the character data extracted in the first extraction step at the beginning is stored in a personal name storage means for storing the surname and the first name used for the personal name;
A second extraction step of extracting a predetermined number of character data from the text data excluding the character data extracted in the first extraction step when it is determined that the last name is stored in the personal name storage means; ,
Determining whether the name having the character data extracted in the second extraction step at the head is stored in the personal name storage means;
And a step of determining that the text data is an abbreviation when it is determined that the name is stored in the personal name storage means.

(Appendix 7)
In the abbreviation determination method for determining whether text data is an abbreviation,
A first extraction step of extracting a predetermined number of character data at the beginning from the text data;
A compound word including a constituent word having the character data extracted in the first extraction step at the head is stored in a compound word storage unit that stores a plurality of compound words and the constituent words constituting each compound word in association with each other. A determination step for determining whether or not,
A second extraction for extracting a predetermined number of character data from the text data excluding the character data extracted in the first extraction step when it is determined that the compound word is stored in the compound word storage means; Steps,
It is determined whether or not the constituent word having the character data extracted in the second extraction step at the head is included in the constituent words of the compound word determined to be stored in the compound word storage means in the determination step. And steps to
And a step of determining that the text data is an abbreviation when it is determined that the constituent word is included.

(Appendix 8)
In a computer program for causing a computer to determine whether text data is an abbreviation,
In a computer equipped with personal name storage means for storing the last name and first name used for the personal name,
A first extraction step of extracting a predetermined number of character data at the beginning from the text data;
Determining whether a surname having the character data extracted in the first extraction step at the head is stored in the personal name storage means;
A second extraction step of extracting a predetermined number of character data from the text data excluding the character data extracted in the first extraction step when it is determined that the last name is stored in the personal name storage means; ,
Determining whether the name having the character data extracted in the second extraction step at the head is stored in the personal name storage means;
And a step of determining that the text data is an abbreviation when it is determined that the name is stored in the personal name storage means.

(Appendix 9)
In a computer program for causing a computer to determine whether text data is an abbreviation,
In a computer provided with a compound word storage means for storing a plurality of compound words and constituent words constituting each compound word in association with each other,
A first extraction step of extracting a predetermined number of character data at the beginning from the text data;
A determination step of determining whether or not a compound word including a constituent word having the character data extracted in the first extraction step at the head is stored in the compound word storage unit;
A second extraction for extracting a predetermined number of character data from the text data excluding the character data extracted in the first extraction step when it is determined that the compound word is stored in the compound word storage means; Steps,
It is determined whether or not the constituent word having the character data extracted in the second extraction step at the head is included in the constituent words of the compound word determined to be stored in the compound word storage means in the determination step. And steps to
A computer program for executing the step of determining that the text data is an abbreviation when it is determined that the constituent word is included.

(Appendix 10)
In a text analysis device that analyzes text data,
The abbreviation determination device according to any one of appendices 1 to 4,
Morpheme storage means for storing morphemes and accent types in association with each other;
Morpheme dividing means for dividing text data into morphemes based on the stored contents of the morpheme storage means;
Means for giving an accent type to each of the morphemes divided by the morpheme dividing means based on the storage contents of the morpheme storage means,
The abbreviation determination device is configured to determine whether or not a morpheme that is not stored in the morpheme storage unit is an abbreviation,
A text analysis apparatus comprising: an accent imparting unit that imparts a predetermined accent type to a morpheme determined to be an abbreviation by the abbreviation determination apparatus.

(Appendix 11)
Abbreviation storage means for storing abbreviations and accent types in association with each other;
The accent giving means is
Means for assigning an accent type to each of the morphemes determined to be abbreviations by the abbreviation determination device based on the stored contents of the abbreviation storage means;
The text analysis apparatus according to appendix 10, further comprising means for giving a predetermined accent type to a morpheme not stored in the abbreviation storage means.

(Appendix 12)
In a text analysis device that analyzes text data,
An abbreviation determination device according to appendix 5,
Text storage means for storing text data and an accent type in association with each other,
The dividing means of the abbreviation determination device is configured to divide document data into text data based on the stored contents of the text storage means,
Means for giving an accent type to each of the text data divided by the dividing means based on the stored contents of the text storing means;
The abbreviation determination device is configured to determine whether text data not stored in the text storage means is an abbreviation,
A text analysis apparatus comprising means for giving a predetermined accent type to text data determined to be an abbreviation by the abbreviation determination apparatus.

(Appendix 13)
Abbreviation storage means for storing abbreviations and accent types in association with each other;
The accent giving means is
Means for assigning an accent type to each of the text data determined to be abbreviations by the abbreviation determination device based on the stored contents of the abbreviation storage means;
The text analysis apparatus according to claim 12, further comprising means for giving a predetermined accent type to text data not stored in the abbreviation storage means.

(Appendix 14)
In a speech synthesizer that generates synthesized speech from text data,
The text analysis device according to appendix 10 or 11,
Prosody generation means for generating prosody corresponding to each morpheme based on the morpheme divided by the morpheme dividing means of the text analysis device and the accent type assigned to each morpheme;
A speech synthesizer comprising: waveform generation means for generating synthesized speech based on the prosody generated by the prosody generation means.

(Appendix 15)
In a speech synthesizer that generates synthesized speech from text data,
The text analysis device according to appendix 12 or 13,
Prosody generation means for generating prosody corresponding to each text data based on the text data divided by the dividing means of the abbreviation determination device and the accent type given to each text data by the text analysis device;
A speech synthesizer comprising: waveform generation means for generating synthesized speech based on the prosody generated by the prosody generation means.

It is a block diagram which shows the structural example of the text analysis apparatus which concerns on Embodiment 1. FIG. It is a schematic diagram which shows the registration content of a language dictionary. It is a schematic diagram which shows the registration content of a personal name dictionary. It is a schematic diagram which shows the registration content of a compound word dictionary. It is a functional block diagram which shows the function structural example of a text analysis apparatus. It is a flowchart which shows the procedure of a text analysis process. It is a flowchart which shows the procedure of an abbreviation determination process. It is a flowchart which shows the procedure of an abbreviation determination process. It is a flowchart which shows the procedure of an abbreviation determination process. It is a flowchart which shows the procedure of an abbreviation determination process. It is a flowchart which shows the procedure of an abbreviation determination process. It is a functional block diagram which shows the function structural example of the text analysis apparatus of Embodiment 2. FIG. It is a schematic diagram which shows the registration content of a co-occurrence dictionary. It is a flowchart which shows the procedure of a text analysis process. It is a functional block diagram which shows the function structural example of the text analysis apparatus of Embodiment 3. FIG. It is a schematic diagram which shows the registration content of an exception abbreviation dictionary. It is a flowchart which shows the procedure of a text analysis process. It is a block diagram which shows the structural example of the speech synthesizer which concerns on Embodiment 4. FIG. 10 is a functional block diagram illustrating a functional configuration example of a speech synthesis device according to a fourth embodiment. It is a flowchart which shows the procedure of the production | generation process of synthetic | combination speech.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Text analyzer 1 Control part 11 Morphological analysis part (morpheme division means)
12 abbreviation determination unit (first extraction means, second extraction means, determination means, determination means)
13 Abbreviated accent giving part (accent giving means)
4a Language dictionary (morpheme storage means)
4b Personal name dictionary (person name storage means)
4c Compound word dictionary (compound word storage means)
4d Co-occurrence dictionary (co-occurrence data storage means)
4e Exception abbreviation dictionary (abbreviation storage means)
20 Prosody generation part (prosody generation means)
30 Waveform generator (waveform generator)

Claims (8)

In an abbreviation determination device that determines whether text data is an abbreviation,
Personal name storage means for storing the last name and first name used for the personal name;
First extraction means for extracting a predetermined number of character data at the beginning from the text data;
Means for determining whether or not a last name having the character data extracted by the first extraction means at the head is stored in the personal name storage means;
A second extraction means for extracting a predetermined number of character data at the beginning from the text data excluding the character data extracted by the first extraction means when it is determined that the last name is stored in the personal name storage means; ,
Means for determining whether or not a name having character data extracted by the second extraction means at the head is stored in the personal name storage means;
An abbreviation determination device comprising: determination means for determining that the text data is an abbreviation when it is determined that the name is stored in the personal name storage means. In an abbreviation determination device that determines whether text data is an abbreviation,
A compound word storage means for storing a plurality of compound words and constituent words constituting each compound word in association with each other;
First extraction means for extracting a predetermined number of character data at the beginning from the text data;
Determining means for determining whether or not a compound word including a constituent word having the character data extracted by the first extracting means is stored in the compound word storage means;
A second extraction for extracting a predetermined number of character data from the text data excluding the character data extracted by the first extraction means when it is determined that the compound word is stored in the compound word storage means; Means,
It is determined whether or not the constituent word having the character data extracted by the second extraction means at the head is included in the constituent words of the composite word determined by the determination means to be stored in the compound word storage means. Means,
An abbreviation determination device comprising: a determination unit that determines that the text data is an abbreviation when it is determined that the constituent word is included. A dividing unit for dividing document data including a plurality of text data into text data;
The first extracting means is configured to extract a predetermined number of character data at the beginning from each of the divided text data,
The second extraction means is configured to extract a predetermined number of character data at the beginning from each of the divided text data excluding the character data extracted by the first extraction means,
The determination means is configured to determine whether each of the divided text data is an abbreviation candidate,
Co-occurrence data storage means for storing a plurality of text data and co-occurrence data co-occurring with each text data in association with each other;
Means for acquiring from the co-occurrence data storage means co-occurrence data corresponding to the text data determined by the determination means to be abbreviation candidates;
Means for determining whether the text data in the document data includes co-occurrence data acquired from the co-occurrence data storage means;
2. The method according to claim 1, further comprising: means for determining that the text data determined by the determining means as an abbreviation candidate is an abbreviation when it is determined that the co-occurrence data is included. 2. The abbreviation determination device according to 2. In a computer program for causing a computer to determine whether text data is an abbreviation,
In a computer equipped with personal name storage means for storing the last name and first name used for the personal name,
A first extraction step of extracting a predetermined number of character data at the beginning from the text data;
Determining whether a surname having the character data extracted in the first extraction step at the head is stored in the personal name storage means;
A second extraction step of extracting a predetermined number of character data from the text data excluding the character data extracted in the first extraction step when it is determined that the last name is stored in the personal name storage means; ,
Determining whether the name having the character data extracted in the second extraction step at the head is stored in the personal name storage means;
And a step of determining that the text data is an abbreviation when it is determined that the name is stored in the personal name storage means. In a computer program for causing a computer to determine whether text data is an abbreviation,
In a computer provided with a compound word storage means for storing a plurality of compound words and constituent words constituting each compound word in association with each other,
A first extraction step of extracting a predetermined number of character data at the beginning from the text data;
A determination step of determining whether or not a compound word including a constituent word having the character data extracted in the first extraction step at the head is stored in the compound word storage unit;
A second extraction for extracting a predetermined number of character data from the text data excluding the character data extracted in the first extraction step when it is determined that the compound word is stored in the compound word storage means; Steps,
It is determined whether or not the constituent word having the character data extracted in the second extraction step at the head is included in the constituent words of the compound word determined to be stored in the compound word storage means in the determination step. And steps to
A computer program for executing the step of determining that the text data is an abbreviation when it is determined that the constituent word is included. In a text analysis device that analyzes text data,
An abbreviation determination device according to claim 1 or 2,
Morpheme storage means for storing morphemes and accent types in association with each other;
Morpheme dividing means for dividing text data into morphemes based on the stored contents of the morpheme storage means;
Means for giving an accent type to each of the morphemes divided by the morpheme dividing means based on the storage contents of the morpheme storage means,
The abbreviation determination device is configured to determine whether or not a morpheme that is not stored in the morpheme storage unit is an abbreviation,
A text analysis apparatus comprising: an accent imparting unit that imparts a predetermined accent type to a morpheme determined to be an abbreviation by the abbreviation determination apparatus. In a text analysis device that analyzes text data,
An abbreviation determination device according to claim 3,
Text storage means for storing text data and an accent type in association with each other,
The dividing means of the abbreviation determination device is configured to divide document data into text data based on the stored contents of the text storage means,
Means for giving an accent type to each of the text data divided by the dividing means based on the stored contents of the text storing means;
The abbreviation determination device is configured to determine whether text data not stored in the text storage means is an abbreviation,
A text analysis apparatus comprising means for giving a predetermined accent type to text data determined to be an abbreviation by the abbreviation determination apparatus. In a speech synthesizer that generates synthesized speech from text data,
A text analysis device according to claim 6;
Prosody generation means for generating prosody corresponding to each morpheme based on the morpheme divided by the morpheme dividing means of the text analysis device and the accent type assigned to each morpheme;
A speech synthesizer comprising: waveform generation means for generating synthesized speech based on the prosody generated by the prosody generation means.

Images