JP2009199434A - Alphabetical character string/japanese pronunciation conversion apparatus and alphabetical character string/japanese pronunciation conversion program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To improve the accuracy of conversion from a character string of an alphabetical notation into Japanese pronunciation. <P>SOLUTION: The alphabetical character string/Japanese pronunciation conversion apparatus includes: a rule database stored with rule data in which an English pronunciation notation is made to correspond to Japanese pronunciation corresponding to the English pronunciation notation based on the English pronunciation notation, English pronunciation notations connected before and after the English pronunciation notation concerned and an alphabetical notation corresponding to the English pronunciation notation; and a Japanese pronunciation conversion part for outputting Japanese pronunciation corresponding to an alphabetical character string from input information in which an English pronunciation notation partial string obtained by decomposing the English pronunciation notation string corresponding to the alphabetical character string into a plurality of components is made to correspond to each of a plurality of alphabetical character partial strings decomposed from the alphabetical character string by using the rule database. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to an alphabet character string Japanese reading conversion device and an alphabet character string Japanese reading conversion program for converting a character string composed of alphabets such as English words into Japanese readings corresponding to the alphabetic character strings.

  As a conventional technique for converting an alphabet character string such as an English word into Japanese reading, for example, Japanese Patent Laid-Open No. 2001-142877 (Patent Document 1) discloses a technique for converting an alphabet character string directly into Japanese reading. ing. This is done by learning the correspondence between the partial character string of the alphabet string and Japanese reading in advance, and obtaining the most probable combination of partial strings when a new alphabet character string is input. To convert.

  In the document “K. Knight and J. Graehl“ Machine Transliteration ”, Computational Linguistics, vol.24, No.4, pp. 599-612 (1998)” (Non-Patent Document 1), A method for converting to Japanese reading is disclosed. This is to convert English phonetic symbols into Japanese readings based on probability information of Japanese readings with respect to English phonetic symbols.

  In addition, as a general technique, there is a technique of preparing Japanese readings for alphabetic character strings in advance and converting the alphabetic character strings into Japanese readings using this database.

K. Knight and J. Graehl "Machine Transliteration", Computational Linguistics, vol.24, No.4, pp. 599-612 (1998) JP 2001-319022 A (paragraph number [0011], FIG. 1)

However, in the prior art in which the correspondence between the partial character string of the alphabet character string and the Japanese reading is learned in advance, it is said that it is not possible to deal with an alphabet character string that does not exist in the learning data, for example, an English word that makes a special reading. There is a problem.
In addition, since English pronunciation notation information is not used, there is a problem that it is not possible to use an existing database regarding English notation for English speech recognition and its pronunciation. Some of these existing databases also have correct pronunciation for English notation that reads special names such as proper nouns, but the method of converting from English notation directly to Japanese does not provide English pronunciation. Since notation information is not used, it is difficult to cope with these special readings.

  Further, in the conventional technique for converting English phonetic symbols into Japanese readings based on the probability information of Japanese readings for English phonetic symbols, the phonetic notation such as the following phonetic symbols (1) representing ambiguous vowels is used. There is a problem that there are pronunciations that are difficult to convert into Japanese readings. For example, as shown in FIG. 1, Japanese vowels corresponding to ambiguous vowels @ (X_SAMPA format) may be / a /, / i /, / e /, / o / (phoneme notation), etc. is there.

  In addition, in the conventional technique for converting an alphabet character string into Japanese reading using data of an alphabet character string and a Japanese reading corresponding to the alphabet character string, a new alphabet character string is added each time a new alphabet character string is added. However, there is a problem that it is necessary to re-establish the database and it is expensive.

  The present invention has been made to solve the above-described problems, and an alphabet character string Japanese reading conversion device and an alphabet character string capable of improving the conversion accuracy of a character string expressed in alphabet to Japanese reading. The purpose is to provide a Japanese reading conversion program.

  According to the present invention, there is provided an alphabetic character string Japanese reading conversion device based on an English pronunciation notation, an English pronunciation notation connected before and after the English pronunciation notation, and an alphabet notation corresponding to the English pronunciation notation. A rule database storing rule data in which notation and Japanese pronunciation corresponding to the English pronunciation notation are associated with each other, and each of the alphabet character substrings obtained by dividing the alphabet character string into a plurality of parts using the rule database Japanese input conversion that outputs Japanese readings corresponding to the above-mentioned alphabet character strings from input information associated with English pronunciation notation sub-sequences obtained by dividing the English phonetic notation sequence corresponding to the above-mentioned alphabet character strings into a plurality of Part.

  In order to convert an alphabet character string into a Japanese reading corresponding to the alphabet character string, the alphabet character string Japanese reading conversion program according to the present invention connects a computer with an English pronunciation notation and before and after the English pronunciation notation. From a rule database storing rule data in which the English pronunciation notation and Japanese pronunciation corresponding to the English pronunciation notation are associated with each other based on the English pronunciation notation and the alphabet notation corresponding to the English pronunciation notation , The rule data reading means for reading the rule data, and the rule data read by the rule data reading means, each of the alphabet character sub-strings into which the alphabet character string has been divided into a plurality of English phonetic notation subsequence with the corresponding English phonetic notation sequence broken down into multiple From the associated input information is intended to function as a Japanese readings converting means for outputting a read Japanese corresponding to the alphabetic character string.

  According to the present invention, based on the English phonetic notation, the English phonetic notation connected before and after the English phonetic notation, and the alphabet notation corresponding to the English phonetic notation, the English phonetic notation and the English phonetic notation are supported. Using a rule database that stores rule data associated with Japanese readings, the Japanese readings corresponding to the above alphabetic character strings are output, so even if you cannot identify Japanese readings only with English pronunciation notation Since Japanese readings can be specified based on alphabetical notation corresponding to English pronunciation notation, there is an effect that the accuracy of Japanese reading conversion can be improved.

  According to the present invention, based on the English phonetic notation, the English phonetic notation connected before and after the English phonetic notation, and the alphabet notation corresponding to the English phonetic notation, the English phonetic notation and the English phonetic notation are supported. The rule data is read from the rule database in which rule data associated with Japanese readings is stored, and the Japanese readings corresponding to the alphabetic character strings are output using the read rule data. Even if the Japanese pronunciation cannot be specified only by the English pronunciation notation, the Japanese reading can be specified based on the alphabet notation corresponding to the English pronunciation notation, so that the accuracy of the Japanese reading conversion can be improved.

Embodiment 1 FIG.
FIG. 2 is a block diagram showing an alphabet character string Japanese reading conversion apparatus according to Embodiment 1 of the present invention. In the present embodiment, the alphabet character string Japanese reading conversion device in the present embodiment receives an alphabetic character string and an English pronunciation character string corresponding to the alphabet character string as input, and a Japanese character corresponding to the alphabet character string. A case where a reading is output will be described.

  In FIG. 2, the pronunciation association database 1 stores pronunciation association data in which alphabetical expressions and English pronunciation expressions corresponding to the alphabetic expressions are associated. In the present embodiment, there are candidates for alphabet notation and English pronunciation notation that can be taken for the alphabet notation. The alphabet character string / phonetic notation association unit 2 uses the pronunciation correlating database 1 to convert an alphabet character string and an English phonetic notation string corresponding to the alphabet character string into a plurality of alphabet character substrings and English, respectively. While decomposing into phonetic notation subsequences, the decomposed alphabetic character subsequences are associated with English phonetic notation subsequences.

  The rule database 3 corresponds to the English pronunciation notation and the English pronunciation notation based on the English pronunciation notation, the English pronunciation notation connected before and after the English pronunciation notation, and the alphabet notation corresponding to the English pronunciation notation. Stores rule data associated with Japanese readings. In the present embodiment, the rule database 3 includes a pronunciation notation conversion rule database 301 and a correction rule database 302.

  The phonetic notation conversion rule database 301 stores the English phonetic notation and the English phonetic notation based on the English phonetic notation, the English phonetic notation connected before and after the English phonetic notation, and the reference necessity information of the alphabet notation. Stores rule data associated with corresponding Japanese readings. The correction rule database 302 is used, for example, when reference to alphabetical notation is required based on the rule data of the phonetic notation conversion rule database 301, and English pronunciation notation and English concatenated before and after the English pronunciation notation. Based on the phonetic notation and the alphabet notation corresponding to the English phonetic notation, rule data in which the English phonetic notation is associated with the Japanese pronunciation corresponding to the English phonetic notation is stored.

  The Japanese reading conversion unit 4 uses the rule database 3 to divide an English phonetic expression string corresponding to the alphabet character string into a plurality of alphabet character substrings into which the alphabet character string has been decomposed into a plurality of characters. From the conversion target information (input information) associated with the English pronunciation notation subsequence, the Japanese reading corresponding to the alphabet character string is output. In the present embodiment, the alphabet character partial string and the English pronunciation notation partial string associated with each other by the alphabet character string / phonetic notation association unit 2 are input as conversion target information (input information). In the present embodiment, the Japanese reading conversion unit 4 includes an English pronunciation notation Japanese reading conversion unit 401, an alphabetical notation Japanese reading correction unit 402, and a Japanese reading output unit 403.

  The English phonetic notation Japanese reading conversion unit 401 uses the phonetic notation conversion rule database 301 to generate an English phonetic notation corresponding to the alphabet character string for each of the alphabet character substrings obtained by dividing the alphabet character string into a plurality of parts. From the conversion target information (input information) associated with the English pronunciation notation subsequence in which the sequence is divided into a plurality, the Japanese reading corresponding to the alphabetic character string and the reference necessity information in alphabetic notation are output. Based on the output result of the English pronunciation notation Japanese reading conversion unit 401, the alphabet notation Japanese reading correction unit 402 uses the correction rule database 302 for the English pronunciation notation subsequence that requires alphabetical reference. Correct and output Japanese readings corresponding to the above alphabetic character strings. The Japanese reading output unit 403 adjusts and outputs the Japanese reading format output from the alphabetical Japanese reading correction unit 402.

  In this embodiment, an index indicating the certainty is attached to the Japanese readings output for the English phonetic notation string and output. In addition, in the above-mentioned English pronunciation notation Japanese reading conversion unit 401 and alphabetical notation Japanese reading correction unit 402, if there are a plurality of pronunciations of Japanese readings, a plurality of Japanese readings for one English pronunciation notation sequence are provided. Is output. For example, for a place where pronunciations of Japanese readings are output as multiple candidates (English pronunciation notation subsequence), an index indicating the certainty is given assuming that there are multiple Japanese readings for the input alphabetic character string. Output. If there are no multiple readings in the above configuration, output is performed assuming that one reading is determined.

Next, the operation will be described.
In the present embodiment, a pair of an alphabet character string and an English phonetic notation character string corresponding to the alphabet character string is input. English phonetic notation for alphabetical character strings is obtained from, for example, an alphabetic character string (English word) prepared for English and a database describing its phonetic notation, or from an alphabetic character string (English notation) It is obtained by converting it into English phonetic notation using a program that converts it into English phonetic notation (phoneme notation).
In this way, by combining with a reading database prepared for English and a program for converting English notation into reading, English pronunciation notation corresponding to an alphabet character string can be obtained easily and at low cost.

  The alphabet string / phonetic notation association unit 2 uses the pronunciation association data in which the alphabetic notation stored in the pronunciation association database 1 and the English pronunciation notation corresponding to the alphabetic notation are associated with each other. The input alphabet character string and the English pronunciation notation string corresponding to the alphabet character string are decomposed into a plurality of alphabet character substrings and English pronunciation notation substrings, respectively, and the decomposed alphabet character substring and English pronunciation Corresponds to the notation subsequence and outputs the information.

  FIG. 3 shows an example of pronunciation association data in which the alphabet notation stored in the pronunciation association database 1 is associated with the English pronunciation notation corresponding to the alphabet notation. In FIG. 3, the alphabetic notation and the English pronunciation notation are character strings each consisting of one or more characters. Further, the correspondence between alphabetic notation and English pronunciation notation is not necessarily one-to-one, and may be one-to-many or many-to-many correspondence.

  Further, as a method of association, association is performed by DP matching or the like. Also, a probability is assigned to each association. For example, in the case where the probability of English pronunciation notation B appearing with respect to alphabet notation A is denoted as P (B | A), P (/ e_Hi /, a) = 0.2 , P (/ {_ H /, a) = 0.1 or the like may be selected so as to obtain the maximum probability.

  Next, the Japanese reading conversion unit 4 uses the rule database 3 to convert the alphabet character substring and the English pronunciation notation subsequence associated with each other by the alphabet character string / phonetic notation association unit 2 into the conversion target information ( Japanese input corresponding to the above-mentioned alphabet character string is output as input information).

The operation of the Japanese reading conversion unit 4 will be described below.
First, "Generate Japanese readings from English pronunciation notation" is performed.
The English phonetic transcription Japanese reading conversion unit 401 creates Japanese reading candidates from the English phonetic transcription string and outputs alphabetical reference necessity information. Based on the English phonetic notation and the English phonetic notation connected before and after the English phonetic notation, the English phonetic notation sequence and the Japanese phonetic conversion are as follows. Conversion is performed using rule data associated with. This rule data is stored in the pronunciation notation conversion rule database 301.

FIG. 4 shows an example of rule data stored in the pronunciation notation conversion rule database 301.
In FIG. 4, Japanese phonetic notation is expressed in phoneme notation, and English phonetic notation is expressed in X-SAMPA format. In FIG. 4, “#” described in the preceding and succeeding pronunciation columns represents silence, and the pronunciation notation in which “#” is preceded is described in the beginning and “#” is described in the following. The case represents the tail. “*” Represents an arbitrary notation. For example, when “*” is described in the pronunciation preceding column, it indicates that the same conversion is performed regardless of the preceding pronunciation notation. In addition, if “*” is written, it means that conversion is performed without depending on the notation, and the rules that have “reference” in the notation column change the Japanese reading depending on the alphabet notation. Means. It is also possible to specify a plurality of readings for Japanese readings converted by this rule data.
In the case where “reference” is written in the notation column, the English phonetic notation Japanese reading conversion unit 401 converts the reading into the reading described in the rule data once, and the alphabet notation reference flag (whether or not to refer to the alphabet notation). Information).

An example of Japanese phonetic conversion processing in the English phonetic transcription Japanese reading conversion unit 401 is shown below.
For example, if trouble (/ trV_Hbl /) is converted according to the rules shown in FIG. 4, the pronunciation of Japanese for the pronunciation notation / t / with no preceding pronunciation notation (first) and the following pronunciation notation is / to / /, Japanese pronunciation for / t /, preceding pronunciation notation / t / and subsequent pronunciation notation / V_H / (vowel) / r /, Japanese reading for / r /, pronunciation / V_H / The Japanese pronunciation for the pronunciation notation / b / with the pronunciation notation / l / (consonant) is / bu /, and the Japanese reading for the pronunciation notation / l / at the end (no subsequent vowel) is / ru /. With the above conversion, the Japanese reading of trouble (/ trV_Hbl /) is / toraburu /.

  In the case of animal (/ {_ Hn @ m @ l /), the subsequent consonant is / {_ H / is the Japanese reading for / a /, the vowel is followed by / n / is the Japanese reading is / n /, Since / @ / is notation reference, an alphabet notation reference flag (alphabet notation reference necessity information) is set for correction in the subsequent steps, and / {a, o} / (/ a / Or / o /). The Japanese reading of / m / followed by a vowel is / m /, and the Japanese reading of / l / at the end is / ru /. Therefore, the animal (/ {_ Hn @ m @ l /) Japanese reading result at this step is / an {a, o} m {a, o} ru /, and is written in English pronunciation / @ / A reference flag (alphabetic reference necessity information) is set.

Next, “correction of Japanese readings in alphabetical notation” is performed.
Based on the output result of the above-mentioned English pronunciation notation Japanese reading conversion unit 401, the English pronunciation notation correcting unit 402 requires English reference notation substrings (where the alphabet notation reference flag is set). Is corrected with reference to the alphabet notation of the portion corresponding to the phonetic notation using the correction rule database 302, and is output.

The correction rule database 302 stores rule data that is used when the alphabet reading Japanese reading correction unit 402 corrects Japanese reading with reference to the alphabet notation.
FIG. 5 shows an example of rule data stored in the correction rule database 302.

Hereinafter, an example of correction processing in the alphabetical Japanese reading correction unit 402 will be described.
Taking animal (/ {_ Hn @ m @ l /) as an example, the result of conversion to Japanese reading by the English pronunciation notation Japanese reading conversion unit 401 / an {a, o} m {a, o} ru / However, alphabetic notation reference flags (alphabet notation reference necessity information) are set at two locations of English pronunciation notation / @ /. The first / @ / is precedent pronunciation notation / n / (consonant), subsequent pronunciation notation / m / (consonant), and the corresponding alphabet notation is “i”. Therefore, the Japanese reading is / i / from the rule data of FIG. The second / @ / is preceded pronunciation notation / m / (consonant), subsequent pronunciation notation / l / (consonant), and the corresponding alphabet notation is “a”. Therefore, the Japanese reading is / a / based on the rule of FIG. As a result of the correction as described above, the Japanese reading is / animaru /.

  This vowel / @ / is a pronunciation different from Japanese vowel / a /, / e /, / i /, / o /, / u / being called. Other examples of English words that use this vowel include system (/ sI_Hst @ m /), supply (/ s @ plA_HI /), and the like. When these are normally written in Japanese readings, they become a system (/ sisutemu /) and a supply (/ sapurai /), and the English pronunciation notation / @ / is / e / and / a /, respectively.

  In addition, there is / n / as an example of a consonant that is not specified only by English pronunciation notation but is affected by English notation. For example, panel (/ p {_Hnl /), Stanley (/ st {_Hnli /), final (/ fA_HInl /), vinyl (/ vA_HInl /), and the like. These are usually written in Japanese readings: panel (/ paneru /), stanley (/ sutaNrii /), final (/ fainaru /), vinyl (/ biniiru /), and English pronunciation notation / n / Then, it can be seen that they are / ne /, / N /, / na /, and / ni /, respectively. In these examples, the English pronunciation notation following the English pronunciation notation / n / is / l /, and it can be seen that the Japanese pronunciation is different even if the preceding pronunciation notation is the same. At this time, the alphabet string corresponding to the English pronunciation notation / n / is ne, n, na, ny. Even in such a case, the data is converted to / ne /, / N /, / na /, / ni / using the rule data stored in the correction rule database 302, respectively.

Next, “processing when ambiguity cannot be resolved” is performed.
In the Japanese reading output unit 403, the alphabetic character string input for the portion where the pronunciation of the Japanese reading is output as a plurality of candidates in the English pronunciation notation Japanese reading conversion unit 401 and the alphabet notation Japanese reading correction unit 402. An index (score (probability)) indicating the certainty is assigned and output, assuming that there are multiple Japanese readings for. If there are no multiple readings in the above configuration, output is performed assuming that one reading is determined. Also, an index (score (probability)) indicating the certainty is given to the entire Japanese reading for the alphabet character string. In addition, it is given for a portion where pronunciation of Japanese reading is output as a plurality of candidates. Moreover, it assign | provides about each alphabet character substring or each English pronunciation notation subsequence. Any one of these indices shall be given.

In addition, correction is made for a portion that is output as a candidate for Japanese reading, which is not appropriate as a Japanese reading, and a lengthening process is also performed.
For example, as an example that is output as a plurality of Japanese readings, there is an example such as phonetic symbol string / k {_Hn /, corresponding alphabetic character string can. Examples of such English words include canada (/ k {_Hn @ d @ /), canopy (/ k {_Hn @ pi /), canvas (/ k {_Hnv @ s /), etc. The Japanese readings are Canada (/ kana /), canopy (/ kjanopii /), canvas (/ kjaNbasu /), but canvas may be read as canvas (kaNbasu). , / Kja / is difficult to assign. In such a case, a plurality of candidates are assigned on the assumption that there are both / ka / and / kja / readings in Japanese. For example, for canada (/ k {_Hn @ d @ /), there are two types, Canada (/ kanada /) and canada (/ kjana /).

  As the output method of these two types of multiple reading, in addition to the method of outputting multiple as it is as / kana /, / kjana /, a part where multiple readings exist in a part such as / {ka, kja} nada / is specified. There is a way to output. In addition, when a plurality of readings exist in each database (pronunciation notation conversion rule database 301 and correction rule database 302), the probability of selecting the reading is determined in advance, so that each Japanese reading output unit 403 can output each reading. It is also possible to output such as /{ka:0.8, kja: 0.2} nada / by giving a reading probability.

  Examples of things that are not suitable for Japanese reading include connection relations such as sound repellent (/ N /, n) and prompt sounds (/ Q /, t). For example, there is usually no case where the sound repellent comes to the beginning of the word or when the sound repellent continues. In such a case, the sound repellent is deleted so that the character string is changed to an appropriate character string for Japanese reading. When sounding sounds appear at the end of a word or the like, they are deleted if they are consecutive. As for the prompting sound, it is deleted even when the normal prompting sound is not used, such as when the vowel follows or when the following phoneme is / m /, / n /, / r /, / w /, or to / cu /, etc. Is replaced. In addition, for items that are prone to lengthening sounds such as ray / ei / and ou / ou /, Japanese readings with long sound like / {ei, ee} /, / {ou, oo} / Output as multiple candidates.

As described above, "Generation of Japanese readings from English phonetic notation", "Correction of Japanese readings by alphabetic notation", and "Processing when ambiguity cannot be resolved" are applied to the input alphabet string. Output the corresponding Japanese reading.
In the present embodiment, the operation description of the Japanese reading conversion unit 4 is converted into Japanese readings once by the English pronunciation notation Japanese reading conversion unit 401, and then the notation needs to be referred to. Although it has been described that Japanese reading is corrected by the alphabetical Japanese reading correction unit 402 and is finally output by the Japanese reading output unit 403, the above-described configuration may be operated simultaneously.
Further, the pronunciation notation conversion rule database 301 and the correction rule database 302 may be integrated and used as one rule database.

  In the embodiment described above, all or some of the above functions can be achieved by executing a program as software such as a personal computer or by executing a program as embedded software such as a CPU or firmware. It can be done. Further, it may be realized by an integrated circuit such as an LSI (Large Scale IC), an FPGA (Field Programmable Gate Array), or a logic IC, or may be realized by combining discrete elements. .

  The above software may be stored in advance in storage means such as a ROM, magnetic disk (hard disk, removable disk, etc.), nonvolatile semiconductor memory, etc. Download from storage means on the server computer using wired / wireless communication means such as infrared communication, Bluetooth, mobile phone packet communication, etc., for example, CD-ROM, CD-R, DVD, MO disk, magnetic disk ( (A hard disk, a removable disk, etc.), a non-volatile semiconductor memory, a magnetic tape or other storage medium, or a print medium such as a card printed with a barcode or the like. In this case, the program code of the software read from the storage medium or the like realizes the functions of the above-described embodiment, and these storage medium and the like constitute the present invention.

  Further, the case where each unit is configured on the same computer has been described, but the present invention is not limited to this, and for example, each unit may be configured by being divided into computers or processing devices distributed on a network. Good.

  In addition, the present invention may be applied to a system composed of one or more devices. The server computer distributes a program or the like for realizing the embodiment of the present invention using a communication means such as a network, and executes a program distributed by a plurality of client computers, mobile terminal devices such as mobile phones and PDAs. Can do.

  As described above, according to the present embodiment, the English pronunciation notation is based on the English pronunciation notation, the English pronunciation notation connected before and after the English pronunciation notation, and the alphabet notation corresponding to the English pronunciation notation. And the English pronunciation notation corresponding to the above-mentioned alphabet character string in each of the alphabet character substrings obtained by dividing the alphabet character string into a plurality of parts using the rule data in which the Japanese pronunciation corresponding to the English pronunciation notation is associated. For input information associated with English pronunciation notation subsequences that are divided into a plurality of columns, the English pronunciation notation subsequence of the input information, the English pronunciation notation concatenated before and after the English pronunciation notation subsequence, and the above English By outputting Japanese readings based on the alphabetical notation corresponding to the phonetic notation subsequence, Japanese readings can be identified with only English phonetic notation. Even If you do, because to read Japanese can be identified on the basis of alphabetical notation corresponding to the English pronunciation notation, it is possible to improve the accuracy of the Japanese reading conversion.

  That is, for example, when a Japanese reads out an English word, an vowel used in Japanese (/ a /, / e /, / i /, / O /, / u /, etc.) and insertion of vowels between consecutive consonants (/ nl /) occurs. Since these replacements / insertions are often influenced by the English alphabet notation, the accuracy of the Japanese reading conversion can be improved by referring to the English alphabet notation for the Japanese reading conversion characters.

  Also, in the present embodiment, by outputting a plurality of Japanese readings for one English phonetic notation string, it is possible to create an English character string that can have a plurality of readings and an alphabet character string that cannot be determined only by phonetic notation / alphabet notation. Also, since multiple readings are given, it is possible to output readings closer to Japanese pronunciation.

  In addition, the accuracy of the output Japanese reading can be understood by adding an index indicating the certainty to the Japanese reading output for the English phonetic notation sequence. It is possible to present the certainty of reading.

  In addition, with respect to the readings output for one English phonetic notation string, each partial character string of the Japanese readings is output with an index indicating the certainty, which is output in more detail. Since the accuracy of Japanese reading can be understood, for example, the accuracy of Japanese reading can be presented in detail to the user.

  In addition, because Japanese pronunciation is output based on English phonetic notation corresponding to alphabetical character strings, it is possible to use a reading database for English-speaking countries and a reading grant program, and a new database for Japanese has been established. The cost can be reduced because it is not necessary.

  Also, because Japanese pronunciation is output based on English phonetic notation corresponding to alphabetic character strings, English words that produce unique pronunciations such as names of people and places, and characters that are mixed with special symbols such as artist names Columns can be converted into Japanese readings if they exist in the English reading database.

Example of Japanese reading corresponding to ambiguous vowels It is a block diagram which shows the alphabet character string Japanese reading conversion apparatus in Embodiment 1 of this invention. It is explanatory drawing which shows an example of the data for pronunciation matching in Embodiment 1 of this invention. It is explanatory drawing which shows an example of the rule data stored in the pronunciation notation conversion rule database 301 in Embodiment 1 of this invention. It is explanatory drawing which shows an example of the rule data stored in the correction rule database 302 in Embodiment 1 of this invention.

Explanation of symbols

  1 phonetic correspondence database, 2 alphabetic character string / phonetic notation correspondence unit, 3 rule database, 4 Japanese phonetic conversion unit, 301 phonetic phonetic conversion rule database, 302 correction rule database, 401 English phonetic phonetic Japanese phonetic conversion unit , 402 Alphabetic notation Japanese reading correction unit, 403 Japanese reading output unit.

Claims (6)

Based on the English phonetic notation, the English phonetic notation connected before and after the English phonetic notation, and the alphabet notation corresponding to the English phonetic notation, the English phonetic notation and the Japanese pronunciation corresponding to the English phonetic notation are A rule database in which the associated rule data is stored;
Using the rule database, each of the alphabet character substrings obtained by dividing the alphabet character string into a plurality of pieces is associated with an English pronunciation notation substring obtained by dividing the English phonetic expression string corresponding to the alphabet character string into a plurality of pieces. And a Japanese reading conversion unit for outputting a Japanese reading corresponding to the alphabetic character string from the input information. A pronunciation association database storing pronunciation association data in which alphabetic expressions and English pronunciation expressions corresponding to the alphabetic expressions are associated;
Using the pronunciation association database, the alphabet character string and the English pronunciation notation string corresponding to the alphabet character string are decomposed into a plurality of alphabet character substrings and an English pronunciation notation substring, respectively, An alphabet character string / phonetic notation correspondence unit for correlating the alphabet character substring and the English phonetic notation subsequence,
The Japanese reading conversion unit uses the alphabet character substring and the English pronunciation notation partial sequence that are decomposed and correlated by the alphabet character string / phonetic notation association unit as input information. The alphabet character string Japanese reading conversion apparatus of 1.   2. The alphabet character string Japanese reading conversion apparatus according to claim 1, wherein the Japanese reading conversion unit outputs a plurality of Japanese readings for one English phonetic expression string.   2. The alphabetic character string Japanese according to claim 1, wherein the Japanese reading conversion unit assigns an index indicating the certainty to the Japanese reading output for the English phonetic expression string. Word reading conversion device.   The Japanese reading conversion unit outputs a reading output for one English pronunciation notation string with an index indicating the certainty for each partial character string of the Japanese reading. The alphabet character string Japanese reading conversion device according to claim 1. In order to convert an alphabet string into Japanese pronunciation corresponding to the alphabet string, the computer uses English pronunciation notation, English pronunciation notation connected before and after the English pronunciation notation, and alphabet notation corresponding to the above English pronunciation notation A rule data reading means for reading out the rule data from a rule database storing rule data in which the English pronunciation notation and the Japanese reading corresponding to the English pronunciation notation are associated with each other,
Using the rule data read by the rule data reading means, each of the alphabet character substrings into which the alphabetic character string has been decomposed into a plurality of parts is divided into English pronunciation notation strings corresponding to the alphabetic character strings. An alphabet character string Japanese reading conversion program for functioning as a Japanese reading conversion means for outputting Japanese readings corresponding to the above-mentioned alphabet character strings from input information associated with English pronunciation notation subsequences.

Images