JP2000172289A - Method and record medium for processing natural language, and speech synthesis device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To realize a method for processing a natural language which is convertible to an expression suitable for reading out aloud a hard-to-hear expression irrespective of words or compound words. SOLUTION: A pronunciation pattern extracting part 104 of a text translation part 103 searches whether or not a corresponding pronunciation pattern exists in a pronunciation rule table 111, and extracts a part presumed to be hard to hear when reading it out. Next a text translation processing part 105 converts the extracted part into a text assuming that it is read out, by using a text translation rule table 112. Since a hard-to hear expression is widely searched from the new viewpoint of a combination of sound at the time of reading aloud a document, it is possible to substitute an appropriate expression even for an expression in units of compound words.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a natural language processing method, a natural language processing recording medium, and a speech synthesizer.

[0002]

2. Description of the Related Art With the spread of the Internet and the like, various types of text information can be easily obtained. However, browsing a large amount of text information on a screen may be burdensome for a user. Accordingly, a text-to-speech technology for converting text into speech has been developed, and recently, personal computers and word processors having a text-to-speech function have been released.

[0003] However, since text information is usually described on the premise that it is read, it is often difficult to read the text information intelligibly to the user only by faithfully reproducing it as audio. Techniques to reduce this inconvenience
Various proposals have been made.

For example, there is the following. JP Hei 3
-35296 (conventional example), JP-A-10-171
No. 485 (conventional example), JP-A-7-121543 (conventional example), JP-A-7-325595 (conventional example), and JP-A-10-228471 (conventional example).

In the prior art, a speech synthesizing apparatus is provided with a function of replacing a difficult word or a confusing homonym in an input text with a plain word. The conventional example is a speech synthesis device that analyzes a syntax using a meaning table and replaces words. The conventional example is a syntax analysis device that analyzes a syntax using a conversation sentence dictionary. The conventional example is a speech synthesizer that inserts a pause according to a phrase relation rule. The conventional example is a speech synthesis system in which a word is identified using a word dictionary and a pause is inserted.

[0006]

The above-mentioned prior art is common in that "syntax analysis" is performed in "word units".

[0007] In this case, if processing is performed on a "word-by-word" basis, sufficient measures cannot be taken in the following respects.

That is, a compound word composed of a plurality of words is frequently used in the text of a written word, but it is easy to hear each word constituting the compound word.
When reading continuously, the pronunciation may be difficult to hear. Therefore, word-by-word processing is not sufficient to improve the inaudibility that arises when reading out words in succession. It is conceivable to replace compound words as one unit, but compound words are
Because there are infinite by any combination of words,
In fact, it is impossible to cover the rules in compound words.

The "syntax analysis" is to analyze a sentence structure such as a phrase or a kashikomi after dividing it into words. The above-described processing in units of words (such as replacement of words) is performed in such a manner. It is inseparable from the sentence analysis process,
It is affected by the parsing method and its accuracy.

The present invention has been made based on such studies, and has as its object to provide a novel natural language processing method capable of solving the above-mentioned disadvantages of the conventional natural language processing.

[0011]

SUMMARY OF THE INVENTION According to the present invention, analysis is made not from the structure of a document but from a novel viewpoint of "combination of sounds when reading a document aloud". That is, a combination of pronunciations that is difficult to hear is registered in advance as a pronunciation pattern rule, and when a text is input, first, it is searched for a sound pattern corresponding to the pronunciation pattern rule, and it is estimated that it is difficult to hear. Pick up the part to be processed (first stage processing). Next, a text conversion rule for converting into a style that is supposed to be read out is applied to the picked-up portion (second stage processing).

The contents of the conversion process flexibly change according to the situation. That is, the decision to replace a word with another word, replace a word with a compound word, insert a pause between words, or insert a particle between words, is made individually. The most suitable conversion process is selected.

In the present invention, instead of focusing on the character face of a sentence as in the conventional example, a portion that is likely to be a problem from a new viewpoint of a sound combination pattern is picked up.
Not only when one word is difficult to hear due to a special pronunciation, or when it is difficult to hear because the word has a homonym, but also when a word is read out continuously, due to compound words The relevant part can be extracted including the case where it is difficult to hear.

[0014] Then, by applying a conversion rule to the extracted part and performing an appropriate process for each mode in which it is difficult to hear, the audibility when reading out the text is improved.

In addition, since the problematic portion is narrowed down by the sounding pattern and the rule of narrowing down the problematic portion by the sounding pattern is known, the construction of the text conversion rule is not so complicated.

Further, the process of narrowing down the problem portion by the pronunciation pattern rule can be performed at the stage when the input document is divided into word strings, and has an advantage that it can be executed independently of the syntax analysis process.

[0017]

DESCRIPTION OF THE PREFERRED EMBODIMENTS First Embodiment of the Natural Language Processing Method of the Present Invention
In the first aspect, the input text is compared with a first table in which a predetermined pronunciation pattern is registered as a pronunciation rule, and a portion including the predetermined pronunciation pattern is extracted from the input text. And replacing the word or compound included in the extracted part with another word or another compound based on the conversion rule registered in the second table, or forming the compound. Performing a process of inserting a particle or a pause between words to be executed.

In this aspect, in the first step,
Picking up a problematic point from the viewpoint of sound combination,
In the second step, an appropriate text conversion process (a conversion process into a text that is supposed to be read aloud) is selected according to the cause of the difficulty in hearing due to the comparison with the conversion rule table. This makes it possible to perform text conversion not only in the case where it is difficult to hear in words, but also in the case where it is difficult to hear in terms of compound words.

In a second aspect of the natural language processing method of the present invention, in the first aspect, the predetermined pronunciation pattern is a unit of one pronunciation combining a vowel, a consonant, a consonant, or a repellent. Expressed as a combination of units.

Since a sentence containing a sound pattern that is difficult to hear is specified in units of classification based on phonetic symbols, it is possible to extract a part that is likely to be a problem only by matching with the sound pattern irrespective of the meaning of the sentence. It becomes possible.

According to a third aspect of the natural language processing method of the present invention, in the second aspect, the predetermined sound generation pattern includes a pattern in which sounds of (consonants and repetitive sounds) are continuous.

This clarifies an example of a sound pattern that is difficult to hear.

According to a fourth aspect of the natural language processing method of the present invention, in the second aspect, the predetermined pronunciation pattern is a combination of (consonant / consonant) (consonant / vowel) (consonant / consonant). Patterns.

Another example of a sound pattern that is difficult to hear is clarified.

The fifth aspect of the present invention is directed to the first to fourth aspects.
In the aspect, when the pronunciation pattern that is difficult to hear exists in an adjacent part of two adjacent words that constitute the compound word, and when the two words are both nouns, a particle between the two words is used. Or insert a pose.

In the case of a compound word composed of a plurality of words, even if it is easy to hear each word independently, it may be difficult to hear when read continuously. Even in such a case, it is possible to insert a particle or a pause so that it is easy to hear. Conventional word-based processing based on syntactic analysis cannot deal with the difficulty of hearing pronunciation caused by continuous reading aloud. In this case, it is conceivable to correspond to the dictionary in compound words by rules, but since compound words can be synthesized indefinitely, text processing in compound words is practically impossible. According to the aspect of the present invention, even in the case where the inaudible pronunciation pattern exists in an adjacent portion between two adjacent words, which was impossible in the related art, it can be extracted as a problem part in the first step. In the step, for example, in the case of a connection between a noun and a noun, by applying a rule of inserting a particle or a pause, it is possible to convert into a style that is easy to hear.

According to a sixth aspect of the natural language processing method of the present invention, in the first to fifth aspects, the pronunciation rule registered in the first table or the pronunciation rule registered in the second table is used. The conversion rules described above can be presented to the outside.

Thus, the user can confirm the pronunciation rules or conversion rules.

According to a seventh aspect of the natural language processing method of the present invention, in the first to sixth aspects, the pronunciation rule or the second pronunciation rule registered in the first table is used.
The conversion rules registered in the table can be externally corrected.

Thus, the user can participate in the construction of the table. As a result, pronunciation rules or conversion rules unique to each user can be added, and the burden of table construction is reduced.

In a first aspect of the recording medium for natural language processing according to the present invention, a text input to a natural language processing apparatus is stored in a first table in which a predetermined pronunciation pattern is registered as a pronunciation rule. A first step of collating and extracting a part including the predetermined pronunciation pattern from the input text; and a word included in the extracted part based on a conversion rule registered in a second table. Or, it is assumed that a process of replacing a compound word with another word or another compound word, or a process of inserting a particle or a pause between words constituting the compound word is performed, and the input text is read aloud. A program for executing the second step of converting to text is recorded.

A computer-readable recording medium on which a new natural language processing program is recorded can be obtained.

In a second aspect of the recording medium for natural language processing according to the present invention, a natural language processing apparatus stores an input text in a first table in which a predetermined pronunciation pattern is registered as a pronunciation rule. A first step of collating and extracting a part including the predetermined pronunciation pattern from the input text; and a word included in the extracted part based on a conversion rule registered in a second table. Or the process of replacing compound words with other words or other compound words,
Or performing a process of inserting a particle or a pause between words constituting the compound word, and performing a second step of converting the input text into text on the premise of reading out the text, The first table and the second table are recorded.

A computer-readable recording medium in which a dictionary base (knowledge base) used for a new natural language processing is recorded is obtained.

According to a third aspect of the recording medium for natural language processing of the present invention, in the first or second aspect, the first table or the second table is read / written.
Writable.

This makes it possible to refer to the table / write to the table.

In the first embodiment of the speech synthesizer of the present invention, a language processing for dividing a Japanese input text into word strings, reading each word in the word strings, and adding linguistic information such as part of speech is performed. A pronunciation rule table in which a means and a pronunciation pattern, which is expressed by the combination of vowels, consonants, murmurs, or repellents, and is difficult to hear when voiced, are registered. A text in which a text conversion rule for converting a word string corresponding to the pronunciation pattern registered in the pronunciation rule table and output from the language processing unit into a word string suitable for speech reading is registered. A conversion rule table and a word string output from the language processing means are compared with the pronunciation rule table, and a word or a plurality of words corresponding to the pronunciation pattern is matched. Processing of extracting a compound word composed of the words of the formula (1) and applying the text conversion rule stored in the text conversion rule table to the extracted word or compound word to replace it with a word or compound word that is easy to hear. Or a text converting means for executing a process of inserting a particle or a pause between words constituting the compound word, and synthesizing a speech waveform from linguistic information given to a word string output from the text converting means. A prosody information generating means for generating necessary prosody information such as a pronunciation sequence, a pitch and a power, and an audio waveform synthesizing means for synthesizing an audio waveform in accordance with the prosody information output from the prosody information generating means.

As a result, a speech synthesizer can be obtained which can be converted into a text style that is easy to read out and output as a read-out voice even when the compound words make it difficult to hear.

According to a second aspect of the speech synthesizer of the present invention, in the first aspect, the text conversion rule held in the text conversion rule table is
Text conversion rule management means that edits and manages according to user's search, add, modify, delete, etc. instructions,
Further, a configuration was provided.

As a result, it is possible to add a rule unique to the user to the table, and it is possible to reduce the cost for constructing the table.

(Embodiment 1) Hereinafter, a speech synthesis apparatus (including a natural language processing apparatus) of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of the speech synthesizer according to the first embodiment of the present invention.

As shown, the speech synthesizer includes a text input unit 101, a language processing unit 102, a text conversion unit 103 (including a pronunciation pattern extraction unit 104, a text conversion processing unit 105, and a syntax analysis unit 106). , A prosody information generation unit 107, a voice waveform synthesis unit 108, a voice output unit 109, a display unit 110, a pronunciation rule table 111
, A text conversion rule table 112, a text conversion rule management unit 113, a text conversion rule input / output unit 114
And

The text input unit 101 receives text data to be processed. The linguistic processing unit 102 performs morphological analysis on the input text data and divides the text data into word strings to which linguistic information such as parts of speech has been added.

The pronunciation pattern extraction unit 104 of the text conversion unit 103 scans the “word string reading information” output from the language processing unit 102, and checks (matches) with the pronunciation patterns registered in the pronunciation rule table 111. ) To extract a portion including the sounding pattern (a portion that is determined to be difficult to hear in continuous reading).

The text conversion processing unit 105 of the text conversion unit 103 replaces a word string included in the part extracted by the pronunciation pattern extraction unit 104 with a word string according to a text conversion rule, A text conversion process, such as inserting a particle or a pause, is performed to convert the input text into a text suitable for reading out by voice.

Further, the syntax analysis unit 106 includes the language processing unit 1
The word string sent from 02 is combined into a phrase string to clarify the connection between sentences. The syntactic analysis unit 106 is not essential for conversion into text suitable for reading aloud, but is provided to provide the prosody information generation unit 107 with information necessary for giving intonation and the like. ing.

The prosody information generation unit 107 generates prosody information such as pronunciation sequence, intonation, pitch and power from linguistic information such as reading of a word string and part of speech converted by the text conversion processing unit 105.

The speech waveform synthesizing section 108 connects speech waveform segments based on the pronunciation sequence and prosody information generated by the prosody information generating section 107 to obtain speech synthesis data.
The synthesized voice is output from the voice output unit 109.

The display unit 110 is a display unit (monitor) for displaying text suitable for reading, obtained by text conversion, as character information, and is used when the user reads the converted document by himself. is there.

In the pronunciation rule table 111, a plurality of voice patterns determined to be difficult to hear are registered. The text conversion rule table 112 includes, for the word string extracted by the pronunciation pattern extraction unit 104,
A text conversion rule for processing such as replacement is registered.

Also, the text conversion rule management unit 113
It functions as a user interface to the text conversion rule table 112, and presents the text conversion rules registered in the text conversion rule table 112 to the user via the text conversion rule input / output unit 114, or adds a user And functions to write the modified rule into the text conversion rule table 112.

Such a speech synthesizer is, for example, shown in FIG.
It is built on a computer system as shown in FIG. This computer system includes a main unit 601.
, A keyboard 602, a display 603, and an input device (mouse) 604. The pronunciation rule table 111 and the text conversion rule table 112 shown in FIG. 1 are stored in a CD-ROM 607 set in the main body 601, in a disk (memory) 606 built in the main body 601, or in another system connected via a line 608. Is stored in the disk 605.

The display unit 110 in the speech synthesizer shown in FIG. 1 is the same as the display 603 in the system shown in FIG.
And the text conversion rule input / output unit 114 in FIG.
This corresponds to the display 603 and the input device 604 in the system of FIG.

In the speech synthesizing apparatus according to the present embodiment, as shown in FIG. 2, roughly two stages of processing are performed in the natural language processing process before speech synthesis.

That is, as a process of the first stage, the pronunciation pattern extraction unit 104 of the text conversion unit 103 converts the word string of the input text (the word string divided by the language processing unit 102) into the pronunciation rule table 111 and Collate,
A part corresponding to a predetermined pronunciation pattern is extracted (step 120).

Next, as a process of the second stage, the text conversion processing unit 105 converts the word or compound word included in the extracted portion into another word according to the conversion rule registered in the text conversion rule table 112. Alternatively, the input text is converted into a text on the assumption that the text is read aloud by performing a process of replacing the word with a compound word or a process of inserting a particle or a pause between words constituting the compound word (step 121).

In the first step, instead of focusing on the connection (the semantic content of the sentence and the structure of the sentence) of the sentence, a portion which is likely to be a problem from a new viewpoint of a sound combination pattern is picked up. Not only when one word is difficult to hear due to a special pronunciation, or when it is difficult to hear because the word has a homonym, but also when a word is read out continuously, due to compound words This makes it possible to extract a wide range of relevant parts, including cases where it is difficult to hear.

Then, in the step of the second stage, a conversion rule is applied to the extracted part, and an appropriate process is performed for each mode in which it is difficult to hear, so that it is easy to hear when reading the input text. Is improved.

Further, since the problem location is narrowed down by the pronunciation pattern and the rule of narrowing down the problem location by the pronunciation pattern is known, the construction of the text conversion rule is not so complicated.

Further, the process of narrowing down the problem portion by the pronunciation pattern rule can be performed at the stage when the input document is divided into word strings, and can be executed independently of the syntax analysis process. For example, by performing "text conversion processing" in parallel with syntax analysis, and utilizing the result of this "text conversion processing" in a form that reinforces "syntax analysis processing", a speech synthesizer capable of outputting speech that is easy to hear is realized. can get.

Hereinafter, specific examples of the contents of the pronunciation rule table 111 and the text conversion rule table 112 and specific examples of text conversion using these rules will be described.

FIG. 3 shows an example of division of the input text into word strings by the language processing unit 102.

That is, the input text 201 “Prime Minister Hashimoto attended the athlete award meeting” is “Hashimoto”,
“Prime Minister”, “Ha”, “Athlete”, “Recognition”, “Venue”,
It is divided into a word string 202 of “ni”, “attend”, “shi”, “ta”, “.”.

As shown in FIG. 3, each of the divided words is given three pieces of linguistic information “word notation, part of speech,”. For information that is not read, such as the reading point “,” or the period “.”, The reading information is described as “*”.

FIG. 4A is a diagram showing an example of a pattern of a combination of difficult-to-hear pronunciations registered in the pronunciation rule table 111, and FIG. 4B is a diagram showing vowels, consonants, resonations, and repellent sounds. It is a figure showing contents.

FIG. 4A shows two examples of tone generation patterns. The pronunciation pattern example 301 is a combination pattern of (consonant / consonant) (consonant / consonant), and the pronunciation pattern example 302 is (consonant / consonant) (consonant / vowel).
(Consonant / resonant)

A unit enclosed in parentheses (for example, (consonant / consonant)) represents one reading pattern. “Vowel”, “consonant”, “resonant”, and “pluck” are groups of phonetic symbols when the Japanese pronunciation is represented as a phonetic symbol string.

FIG. 4B shows the grouping of phonetic symbols when the phonetic symbols are represented in Roman alphabet.
That is, a vowel is a group consisting of “a, i, u, e, o” and consonants are “k, s, t, n, h, m, y, r,
w, g, z, d, b, p ", the repetitive sound is a group composed of" ya, yu, yo ", and the sound repellency is a group composed of" nn ".

For example, the pronunciation “(consonant / resonant)” is a pronunciation obtained as a combination of any phonetic symbol included in the “consonant” group and any phonetic symbol included in the “consonant” group. means. For example,
Readings such as “sha” and “gyo” are included as these readings. "Car type (syasy
u) "," zyuzyu ", and compound words such as" democracy (minsyusyugi) "include a combination of (consonants and murmurs). Therefore, these words and compound words are the pronunciations of FIG. The pattern is extracted by the pattern extraction unit 104.

Further, the word "passenger car (kyakusya)" is used in the example 302 of the pronunciation pattern (consonants / consonants),
It corresponds to a combination of (consonant / consonant) and (consonant / consonant), and is an extraction target.

FIG. 5 shows a text conversion rule table 112.
FIG. 7 is a diagram showing an example of a conversion rule (rule for performing processing such as replacing a word string extracted as hard to hear) with another word, which is stored in FIG.

Reference numerals 401, 402, and 403 each indicate a unit of one text conversion rule. As shown in the figure, the rule of each unit is defined in principle by a combination of a word string pattern before conversion shown in the upper part and a word string pattern after conversion shown in the lower part.

In the word string pattern, (notation reading
The unit of “part of speech” represents information of one word.

The text conversion rule 401 is a rule that replaces the word “Prime Minister” with the word “Prime Minister”. The text conversion rule 402 is a rule for replacing the word “transfer (juju)” with the word “transfer (ukewashi)”.

The text conversion rule 403 is a rule for inserting a particle between nouns. In the conversion rule 403, in the pattern of the word string before conversion, “notation” and “reading” are indicated by “?”, Which indicates that it can be collated with arbitrary notation and reading. In other words, this rule is applied to any continuous noun in a word string extracted as including a pronunciation pattern that is difficult to hear.

In the text conversion rule 403,
In the converted word string pattern, “(1)”,
The number “(2)” corresponds to the order from the left of the two “(?? nouns)” appearing in the word string pattern before conversion. In other words, if a compound word having consecutive nouns corresponds to rule 403, the compound word is decomposed into each noun, (1) and (2) are replaced with each word, and a particle is inserted between each noun. For example, "Athlete Award" is the text conversion rule 4
When matching with 03, “(1)” is replaced with “athlete”, “(2)” is replaced with “recognition”, and consequently, a particle “recognition of competitor” is inserted. Will be converted into the sentence. In this case, the particle is inserted, but according to the same structure rule,
Pauses can be inserted between nouns or between specific words.

FIG. 6 shows an example in which text conversion is performed using such rules. The word string 501 shown in the upper part of FIG.
This is a pre-conversion word string obtained as a result of the analysis in step 2.

In the word string 501 before the conversion, FIG.
(A) pronunciation pattern 301 ((consonant / resonant) (consonant /
The combination of the word string 501a corresponds to the word string 501a.
And "Athlete (Kyogisha) and Commendation (Leopard Show)" in the word string 501b. Therefore, word strings 501a and 501b are extracted as a result of the search using the pronunciation rule table 111 by the pronunciation pattern extraction unit 104 in FIG.

Next, the text conversion processing unit 105 shown in FIG.
Searches the text conversion rule table 112 for text conversion rules to be applied to the extracted word strings 501a and 501b. That is, word string 5
01a corresponds to the text conversion rule 401 shown in FIG. Also, the word string 501b is a text conversion rule 4
This corresponds to 03. Therefore, the text conversion processing unit 105
Performs a process of replacing the word string 501a with another word 502a and inserting particles between consecutive nouns in the word string 501b.

In this way, a text 502 suitable for reading out by voice is obtained. Since the input text is converted into a text that takes into account reading aloud by natural language processing before speech waveform synthesis, clear speech that is easy to hear can be output.

Although the result of the text conversion is reflected in speech reading in principle, the present invention is not necessarily limited to this. The reading is not performed and the display unit 110 (FIG. 7) shown in FIG. Display 603)
May be simply displayed.

For example, at a broadcasting site such as a broadcasting station, a news broadcast is generally performed in such a manner that an article manuscript written in written language is corrected to a sentence expression suitable for reading out in a news broadcast, and an announcer reads out the text. It is a target. However, in the case of a sudden incident, sufficient time for rewriting an article manuscript to a news manuscript cannot often be secured.

In such a case, in order to reduce the burden on the announcer, in the text of the article manuscript written in written language, a portion that is difficult to hear is automatically determined, and an expression example of the spoken word to be replaced is read aloud. It is convenient to have a system as presented before. The present invention can be used for such applications.

In the speech synthesizer of FIG. 1, the pronunciation rule table 111 and the text conversion rule table 112
Is a rewritable table. That is, using the text conversion rule input / output unit 114 and the text conversion rule management unit 113, the pronunciation rule table 111 and the text conversion rule table 112 can be presented to the user, and the user himself can add or modify rules. Can be.

Therefore, it is possible to reduce the labor for constructing each rule table, and to efficiently accumulate rules that meet the needs of the user, thereby constructing a rule table that is easy for the user to use. That is, the know-how of the worker who rewrites the manuscript regarding the conversion from the written language to the spoken language can be efficiently accumulated in the system.

As described above, according to the speech synthesizing apparatus (natural language processing apparatus) of the present embodiment, not only word units but also complex It is possible to determine the difficulty of hearing even when reading consecutive words constituting a word. Then, the text conversion rule table 112
The text conversion is executed using the conversion rules registered in the .symbol, and a compound word that cannot be replaced in word units can be replaced with a more audible expression.

Further, for example, a conversion rule may be applied to the extracted word string to present a candidate for an expression to be converted to a user, and the user may update the candidate by himself. As a result, it is possible to efficiently accumulate the conversion rules into the spoken language corresponding to the actual needs of the user.

[0089]

As described above, according to the present invention, it is possible to replace hard-to-hear expressions included in input text written in written language with easy-to-hear expressions and to provide easy-to-understand speech. is there.
Further, pronunciation rules and conversion rules that match the actual needs of the user can be efficiently accumulated.

[Brief description of the drawings]

FIG. 1 is a block diagram of a speech synthesis device (including a natural language processing device) according to a first embodiment of the present invention;

FIG. 2 is a flowchart showing a procedure of a text conversion process in the speech synthesizer according to the first embodiment;

FIG. 3 is a diagram illustrating an example of division of an input text into word strings.

FIG. 4A is a diagram showing an example of the contents of pronunciation rules. FIG. 4B is a diagram for explaining the classification of pronunciation symbols.

FIG. 5 is a diagram showing an example of the contents of a text conversion rule.

FIG. 6 is a diagram showing an example of text conversion.

FIG. 7 is a diagram showing an example of a computer system for implementing the present invention.

[Explanation of symbols]

 Reference Signs List 101 text input unit 102 language processing unit 103 text conversion unit 104 pronunciation pattern extraction unit 105 text conversion processing unit 106 syntax analysis unit 107 prosody information generation unit 108 voice waveform synthesis unit 109 voice output unit 111 pronunciation rule table 112 text conversion rule table 113 Text conversion rule management unit 114 Text conversion rule input / output unit

Claims (12)

[Claims] 1. A method for comparing an input text with a first table in which a predetermined pronunciation pattern is registered as a pronunciation rule, and extracting a part including the predetermined pronunciation pattern from the input text. 1) replacing the word or compound included in the extracted part with another word or another compound based on the conversion rule registered in the second table; or A second step of executing a process of inserting a particle or a pause between constituent words. 2. The natural language processing according to claim 1, wherein the predetermined sounding pattern is expressed by a unit of one sounding combination of a vowel, a consonant, a consonant sound, or a spelling sound. Method. 3. The natural language processing method according to claim 2, wherein the predetermined pronunciation pattern includes a pattern in which pronunciations composed of (consonants / consonants) are continuous. 4. The natural language processing method according to claim 2, wherein the predetermined pronunciation pattern includes a pattern composed of a combination of (consonant / consonant) (consonant / vowel) (consonant / consonant). 5. When the inaudible pronunciation pattern exists in an adjacent part of two adjacent words constituting the compound word, and the two words are both nouns,
5. The natural language processing method according to claim 1, wherein a particle or a pause is inserted between the two words. 6. The method according to claim 1, wherein the pronunciation rules registered in the first table or the conversion rules registered in the second table can be externally presented. Item 6. The natural language processing method according to any one of Items 5. 7. The apparatus according to claim 1, wherein the pronunciation rules registered in the first table or the conversion rules registered in the second table can be externally corrected. The natural language processing method according to claim 1. 8. The natural language processing apparatus checks an input text against a first table in which a predetermined pronunciation pattern is registered as a pronunciation rule, and converts the predetermined pronunciation pattern from the input text. A first step of extracting a part that includes the part, and a process of replacing a word or a compound word included in the extracted part with another word or another compound word based on a conversion rule registered in a second table Or a second step of executing a process of inserting a particle or a pause between words constituting the compound word, and converting the input text into text on the assumption that the text is read aloud. And a computer-readable recording medium for natural language processing. 9. The natural language processing apparatus checks an input text against a first table in which a predetermined pronunciation pattern is registered as a pronunciation rule, and converts the predetermined pronunciation pattern from the input text. A first step of extracting a part that includes the part, and a process of replacing a word or a compound word included in the extracted part with another word or another compound word based on a conversion rule registered in a second table Or performing a process of inserting a particle or a pause between words constituting the compound word, and performing a second step of converting the input text into text on the assumption that the text is read aloud, A computer-readable recording medium for natural language processing, wherein the first table and the second table are recorded. 10. The recording medium for natural language processing according to claim 8, wherein the first table or the second table is readable / writable. 11. A linguistic processing means for dividing a Japanese input text into word strings, reading each word in the word strings, and adding linguistic information such as part of speech, and combining vowels, consonants, murmurs, or vowel sounds. A pronunciation rule table in which a pronunciation pattern that is expressed by a combination of the pronunciation units and is difficult to hear when voiced is registered, and the pronunciation pattern registered in the pronunciation rule table And a text conversion rule table in which a text conversion rule for converting a word string output from the language processing unit into a word string suitable for reading aloud is registered; and a text conversion rule table output from the language processing unit. The word sequence is compared with the pronunciation rule table, and a word or a compound word including a plurality of words corresponding to the pronunciation pattern is extracted and extracted. Applying the text conversion rules stored in the text conversion rule table to the words or compound words, and replacing the words or compound words with words or compounds that are easy to hear, or particles between the words constituting the compound words Or text conversion means for executing a process of inserting a pause, and prosody information such as a pronunciation sequence, pitch, and power necessary for synthesizing a speech waveform from linguistic information given to a word string output from the text conversion means. And a speech waveform synthesizing unit that synthesizes a speech waveform in accordance with the prosody information output by the prosody information generating unit. 12. A text conversion rule management means for editing and managing the text conversion rules stored in the text conversion rule table in accordance with a search, addition, correction, deletion, etc. instruction from a user. The speech synthesizer according to claim 11, wherein