JPH10326273A - Data compressing device and data restoring device and data compressing method and data restoring method and program recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a data compressing device and resorting device suitable for the compression of a Japanese document. SOLUTION: A data compressing device is constituted so that phonetic data being data obtained by replacing a Chinese character idiom in sentence data with information constituted of number of character identification information indicating the number of characters of the reading of the Chinese character idiom and the reading and homonym identification information can be outputted by a phonetics converting part 12 having a homonym dictionary 17 for storing the Chinese character idiom and the reading and homonym identification information by making them correspond to each other, and the phonetic data can be compressed by a non-distortion compressing part 13. At that time, the first byte of a two byte character which is used in the sentence data and one byte information whose content is different from that of one byte character is used as the number of character identification information.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus and a method for compressing and restoring document data containing characters represented by a plurality of bytes, and a program recording medium, for example, for compressing and restoring Japanese document data. Device and method and a program recording medium.

[0002]

2. Description of the Related Art In recent years, with the spread of electronic mail and the like, the amount of digitized documents processed or stored by a personally owned computer has been dramatically increased. For example, 1
Many people process several hundred to one thousand e-mails a day, and it is not unusual that document data of several hundred MB or more is stored in one year.

For this reason, by reducing the amount of data by omitting redundant portions in the data, the data transmission time and the capacity required for storing data are reduced. Various methods for compressing data are known, and there are methods that can compress various data such as character codes, vector information, and images. In such a compression method, encoding called universal encoding is used.

[0004] The outline of some encoding methods classified as universal encoding will be described below. In the following description, following the names used in information theory,
One unit of data is described as “character”, and a plurality of “characters” connected is described as “character string”.

First, an outline of arithmetic coding will be described. Arithmetic coding includes binary arithmetic coding and multi-valued arithmetic coding with three or more values. In multi-valued arithmetic coding, a number line from 0 to less than 1 (hereinafter referred to as [0, 1)). Are sequentially narrowed according to the occurrence probability (appearance frequency) of each character constituting the data to be encoded. Then, when processing for all characters is completed, a numerical value representing one point in the narrowed section is output as a code.

For example, the character to be encoded is a,
b, c, d, and e, and the occurrence probabilities of those characters are 0.2, 0.1, 0.05, 0.15,
If it is 0.5, as shown in FIG. 24, a section having a section width corresponding to the occurrence probability is assigned to each character.

The character string to be encoded is "ab"
25, first, as schematically shown in FIG. 25, the section [0, 1) is divided into sections [0, 1] for the character “a”.
0.2). Next, the section [0, 0.
2) is divided into sections corresponding to the occurrence probabilities of the characters, and the sections [0.04, 0.0] corresponding to the next character “b”
6) is selected as a section for the character string "ab".
Further, the section [0.04, 0.06) is divided into sections corresponding to the occurrence probability of each character, and the next character “e”
The section [0.05, 0.06) corresponding to the character string "ab
Then, a bit string with a decimal point or less when the position of an arbitrary point (for example, the lower limit) in the section is binary-displayed is output as an encoding result.

The arithmetic coding is performed by a method of dividing a section in accordance with the occurrence probability (appearance frequency), and furthermore, by performing a division according to a preset appearance frequency irrespective of the actual appearance frequency of each character. (static) A coding method and a semi-adaptive coding method that divides by the appearance frequency obtained by scanning all character strings first, recalculates the appearance frequency every time a character appears , Adaptive type to reset the section
(adaptive) coding method. Literature on arithmetic coding includes Bell, TC, Cleary, JG, and Witten,
IH (1990) "Text Compression", Prentice Hall.

[0009] Universal coding called spray coding is also known. In spray coding, a process of rearranging a code tree (a code table having a tree structure) is performed such that a shorter code is assigned to a character having a higher appearance frequency every time a character is coded. For details of spray coding, see, for example, Jones, Douglas W. "Ap
replication of Splay Trees Data Compression ", Commo
n. ACM, vol. 31, no. 8, pp. 996-1007, Aug., 1998.

[0010] An encoding method called blend spray encoding is also known. Blend spray coding uses a blending model (blending
This is an encoding that employs a statistical model called “model”.

In the blend spray coding, a code tree is prepared for each context. Here, the context is a character string (“ab”) existing immediately before the encoding target character (“c”), as schematically shown in FIG. blend·
In spray coding (blend model), the number of characters (order) used as a context is controlled by a context tree as shown in FIG. That is, in general, when data having a strong correlation between characters is targeted, a higher compression ratio can be obtained as a higher-order context is used. When the context is used, the compression ratio may be reduced. The technology devised to prevent this is the blend model. In the blend model, if a certain context is likely to appear, the order related to the context is increased, and if it is difficult to appear, the order is kept low. Then, the order of each context is changed to the one adapted to the input data.

[0012]

SUMMARY OF THE INVENTION
Since it was developed in a cultural area that uses alphabets, one byte is treated as one character during compression using each encoding. For this reason, when a sentence including characters represented by 2 bytes, such as Japanese, is compressed using each technique, there is a problem that a higher compression ratio cannot be obtained as compared to an English document.

That is, a combination of two-byte data is significant for a two-byte character, and there is no correlation between the two byte data constituting the two-byte character. For this reason, in the conventional compression method of processing two-byte characters one byte at a time, according to information theory, compression is performed after reducing the information source (dividing two-byte data into one byte). That is, a high compression ratio cannot be obtained.

There is another problem that it is difficult to achieve high compression by using context. In other words, there are thousands of types of kanji that are used on a daily basis, so if a sentence of the same size is compressed by encoding using a context with the same number of bytes, the Japanese document will be More context appears than in English documents. In fact, when counting the number of 4-byte contexts that appear when compressing an 8 kB Japanese document and an 8 kB English document, approximately 3000 types of contexts appeared in the English document, whereas in the Japanese document, ,
Approximately 5000 contexts have emerged. In addition, many Japanese documents to be compressed are relatively small in capacity (about several A4 pages) such as electronic mail. As a result, when compressing Japanese documents, compression of all documents is often completed before sufficient statistical information is collected for each context, which also reduces the compression rate of Japanese documents. It was a factor that was lowering.

An object of the present invention is to provide a data compression apparatus, a data compression method, and a program recording medium suitable for compressing document data of a language having characters represented by a plurality of bytes such as Japanese. is there.

Another object of the present invention is to provide a data decompression device, a data decompression method, and a program recording medium capable of decompressing data compressed by such a data compression device and data compression method.

[0017]

According to a first aspect of the present invention, there is provided a data compression system for compressing original document data including character information in which one character is represented by a plurality of bytes of information. The device creates phonetic document data, which is data obtained by replacing each character information included in the original document data to be compressed with phonetic character information representing a sound generated when a character associated with the character information is pronounced. The phonetic document data creating means and compression means for compressing the phonetic document data created by the phonetic document data creating means are used.

According to the data compression apparatus of the first aspect, the original document data is compressed after being converted into phonogram document data represented by phonogram information having a smaller number of types. Compression with a higher compression ratio can be realized as compared with the case where data is directly compressed.

When the data compressed by the data compression apparatus is restored, the restoration means for restoring the compressed document data, and the phonogram representing the sound when a character is reproduced, restored by the restoration means. A data restoring device is used which includes original document data generating means for generating original document data which is data that is a source of compressed document data by converting character information into corresponding character information.

According to a second aspect of the present invention, a data compression device for compressing original document data including character information in which one character is represented by a plurality of bytes is converted into a plurality of conversion target word information including one or a plurality of character information. On the other hand, phonogram information storage means storing phonogram information which is information representing a sound when the word indicated by the conversion target word information is pronounced, and phonogram information storage means from the original document data. Search / read means for searching the conversion target word information stored in the storage means and reading phonogram information corresponding to the searched conversion target word information from the phonogram character information storage means, and the search / read means. Creating phonetic document data by replacing the conversion target word information in the original document data with the conversion target word replacement information including the phonetic character information read by the search / read means. An intermediate code table creating means for creating an intermediate code table for associating an intermediate code with an information element used in the phonetic document data created by the phonetic document data creating means; Using an intermediate code table created by the code table creating means, by converting each information element constituting the phonetic document data into a corresponding intermediate code, thereby creating intermediate code document data creating intermediate code document data; And compression means for compressing the intermediate code document data created by the intermediate code document data creation means.

That is, in the data compression device according to the second aspect of the present invention, first, the conversion target word in the original document data is replaced with phonetic character information, whereby the character used more than the original document data is replaced. A few types of phonetic document data are created. Thereafter, a new code (intermediate code) is assigned to each information element (character) included in the phonetic document data, and the phonetic document data is converted to intermediate code document data using the intermediate code. Then, the intermediate code data is compressed. For this reason, according to the present data compression apparatus, it is possible to compress document data including characters such as symbols, which are difficult to represent with phonetic character information, at a high compression rate.

When decompressing the data compressed by the data compression apparatus of the second aspect, when the word indicated by the conversion target word information is pronounced with respect to the conversion target word information composed of one or a plurality of character information, Phonetic character information storage means for storing phonetic character information which is information representing the sound of the sound, decompression means for outputting intermediate code document data by decompressing compressed document data, and intermediate code document output by the decompression means A phonetic document data generating means for generating phonetic document data by replacing each intermediate code included in the data with information associated with the intermediate code in an intermediate code table relating to the compressed document data; The conversion target word replacement information included in the phonetic document data generated by the phonetic document data generation means is searched, and the searched conversion target word replacement information is included in the conversion target word replacement information. Original document data generating means for generating original document data that is the basis of compressed document data by substituting the conversion target word information stored in the phonogram information storage means in association with the phonogram information to be converted. Is used.

In forming the data compression apparatus of the second aspect, the intermediate code table creating means uses, for each information element used in the phonetic document data, a minimum information element capable of expressing the information element. Means for creating an intermediate code table for allocating an intermediate code having the number of bits.

The phonetic document data creating means may include a predetermined type of character information, replacement information interposed between start position identification information and end position identification information indicating that the character information is a predetermined type of character information. By using means for creating phonetic document data, the intermediate code is used as the intermediate code table creating means, in the information following the start position identification information and the end position identification information in the phonetic document data. A means for creating an uncorrelated intermediate code table is used, and as the intermediate code data creating means, the information following the start position identification information and the end position identification information in the phonetic document data are converted into intermediate codes. No means can be used.

When a data compression apparatus is constituted by using such means, for example, for a predetermined type of character information such as one kanji character, it is converted into phonetic character information and converted into an intermediate code. Can not be performed.

Also, when the type of the information element used in the phonetic document data exceeds the number N of information that can be represented by a predetermined number of bits, the intermediate code table creating means is used for the phonetic document data. "N-1" information elements to which an intermediate code is to be assigned are selected from among the information elements that are present, and the selected "N-1" information elements and the start position identification information have a predetermined number of bits different from each other. Using means for creating an intermediate code table for associating the intermediate codes of, as the intermediate code data creating means, for the information in the phonetic document data, the information to which the intermediate code is associated in the intermediate code table,
The information is converted to a corresponding intermediate code, and for the unallocated information that is information to which no intermediate code is allocated, the unallocated information has an intermediate code associated with the start position identification information at the beginning. In addition, a means for creating intermediate code document data can be used by replacing with unassigned replacement information, which is information including unassigned information and in which the end position is known.

[0027] The phonetic character information storage means is used for identifying a plurality of conversion target words from the phonetic character information and another conversion target word to which the same phonetic character information is associated. The means for storing the synonym identification information and the means for reading the phonetic character information and the homonym identification information corresponding to the searched conversion target word are used as search / readout means. Means for replacing the conversion target word information in the original document data with the conversion target word replacement information including the phonetic character information and the homonym identification information read by the search / read means, Means for creating an intermediate code table for information elements other than homonym identification information can also be used.

According to a third aspect of the present invention, a data compression apparatus for compressing original document data containing character information in which one character is represented by a plurality of bytes is provided for converting target word information comprising one or more character information. The phonetic character information storage means in which phonetic character information that is information representing the sound when the word indicated by the conversion target word information is pronounced, and the phonetic character information storage means from the original document data are stored in the phonetic character information storage means. Searching / reading means for searching the conversion target word information, and reading phonogram information corresponding to the searched conversion target word information from the phonogram information storage means, and an original document searched by the search / read means. Phonetic document data creating means for creating phonetic document data by replacing the conversion target word information in the data with the conversion target word replacement information including the phonetic character information read by the search / read means; this Configured using a compression means for compressing against phonetic document data is sound document data creation unit creates.

According to the data compression apparatus of the third aspect, the original document data is compressed after being converted into phonogram document data represented by phonogram information of a smaller number. Compression with a higher compression ratio can be realized as compared with the case where data is directly compressed.

When decompressing the data compressed by the data compression device of the third aspect, the word indicated by each conversion target word information is pronounced with respect to the conversion target word information composed of one or a plurality of character information. The phonetic character information storage means in which phonetic character information as information representing the sound at the time is stored, the restoring means for outputting phonetic document data by restoring the compressed document data, and the restoring means generate Searching for conversion target word replacement information included in the phonetic document data, and storing the searched conversion target word replacement information in correspondence with the phonetic character information included in the conversion target word replacement information. And a source document data generating means for generating source document data which is the source of the compressed document data by replacing the source document data with the conversion target word information stored in the source document data.

In the data compression device of each mode, any phonetic character information can be used. For example, information representing an alphabet or information representing a vowel and a consonant of a Hangul character can be used. It is also possible to use information representing international phonetic symbols and note characters.

[0032]

Embodiments of the present invention will be specifically described below with reference to the drawings. <First Embodiment> The data compression / decompression device according to the first embodiment is a device for compressing / decompressing a Japanese document, and is realized by operating a computer in accordance with a compression / decompression program.

First, the outline of the data compression / decompression device of the first embodiment will be described with reference to the functional block diagram shown in FIG. As illustrated, the data compression / decompression device of the first embodiment includes a storage unit 11, a phonetic conversion unit 12, and a distortionless compression unit 1.
3, a distortionless restoration unit 14, and a phonetic inverse conversion unit 15.

The storage unit 11 stores plaintext data which is data to be compressed (or decompressed data) and compressed data which is a compression result of the plaintext data. The storage unit 11 stores data using a shift JIS code as plaintext data.

The speech conversion unit 12 and the distortionless compression unit 13
Works when performing data compression. The phonetic conversion unit 12 includes a Japanese analysis dictionary 16 and a homonym dictionary 17. The Japanese analysis dictionary 16 is a dictionary for extracting kanji idioms and one kanji (sound reading, kun reading) from plain text data consisting of non-kanji characters (hiragana, katakana, symbols, etc.) and kanji. The homonym dictionary 17 contains kanji idioms (or kanji)
Is a dictionary in which homonym identification information, which is information for specifying one kanji idiom (or kanji) from a reading (hiragana character string), is stored. That is, the homonym dictionary 17 stores homonym identification information (0, 1,...) Having different contents for a plurality of kanji idioms having the same pronunciation as illustrated in FIG. . In the present embodiment, 1-byte information in which a numerical value is represented by a binary number is used as homonym identification information.

The phonetic conversion unit 12 converts the plaintext data stored in the storage unit 11 using the Japanese analysis dictionary 16 and the homonym dictionary 17, and compresses the converted phonetic data without distortion. To the unit 13. Details will be described later,
The phonetic conversion unit 12 replaces the kanji or kanji idioms contained in the plaintext data with information including character number identification information indicating the number of characters of the reading (hiragana character string), reading, and homonym identification information. To convert the plaintext data into phonetic data. The lossless compression unit 13 performs lossless compression on the phonetic data, and stores the result (compressed data) in the storage unit 11. In the present embodiment, for example, the distortion-free compression unit 13 performs the blend spray encoding in which the highest order of the context is “2”.

The distortionless restoration unit 14 and the phonetic inverse conversion unit 15
Works during data recovery. The distortionless restoration unit 14 has a function of restoring the compressed data output from the distortionless compression unit 13. The distortion-free restoration unit 14 restores the compressed data stored in the storage unit 11 and supplies the restoration result (phonetic data) to the phonetic inverse conversion unit 15. The phonetic inverse conversion unit 15 has a homonym dictionary 18 having the same contents as the homonym dictionary 17. The phonetic inverse conversion unit 15 converts the phonetic data using the dictionary and stores the plaintext data as the conversion result in the storage unit 1.
Stored in 1.

Hereinafter, the operation of the data compression / decompression device of the first embodiment will be specifically described. First, the operation at the time of data compression will be described. FIG. 3 shows the sound conversion unit 12 during data compression.
The following shows the operation procedure. As shown in the figure, at the time of data compression, the phonetic conversion unit 12 first obtains 1-byte data from plaintext data to be compressed in the storage unit 11 (step S101). Then, it is determined whether the data is a one-byte character (control code, alphabet, numeral, half-width katakana) or the first byte of a two-byte character (step S102).

If the acquired data is the first byte of a two-byte character (step S102; two-byte character), the phonetic conversion unit 12 acquires the next one-byte data from the plaintext data (step S103). I do. Next, the phonetic conversion unit 12 determines whether the character represented by the obtained 2-byte data is a Chinese character (step S104). If the character is not a kanji (step S104; non-kanji), that is, if the character is a hiragana, a katakana, a symbol (ruled line segment etc.), the phonetic conversion unit 12
The acquired data is output to the distortionless compression unit 13 (Step S105). Also, when the data obtained in step S101 is a one-byte character (step S102; one byte character), the phonetic conversion unit 12 outputs the obtained data to the distortionless compression unit 14 (step S105). And
If the output data is not “EOF” (step S114; N), the process returns to step S101, and the processing for the remaining data in the plaintext data is executed.

Hereinafter, for convenience of description, both 1-byte data representing a 1-byte character and 2-byte data representing a 2-byte character will be referred to as unit information. If the character represented by the obtained unit information consisting of two bytes of data is a kanji (step S104; kanji), the phonetic conversion unit 12 obtains the next unit information from the plaintext data (step S106). That is, the next one-byte data is obtained from the plaintext data, and if the data is a one-byte character, the data is used as unit information. If the data is the first byte of a two-byte character, the phonetic conversion unit 12 further acquires one-byte data and uses the two-byte data as unit information.

Thereafter, the phonetic conversion unit 12 determines whether the newly acquired unit information represents a kanji, a non-kanji, or a one-byte character (step S107). If the unit information represents a kanji (step S107: kanji), the phonetic conversion unit 12 returns to step S106 and acquires the next unit information from the plaintext data.

If the newly acquired unit information in step S106 represents a non-Kanji character or a one-byte character (step S107; other), the phonetic conversion unit 1
No. 2 determines whether or not the number of kanji acquired so far is "1" (step S108). If the acquired number of kanji is "1" (step S108; Y), it is determined whether or not the kanji is a phonetic kanji (step S1).
09). Note that the phonetic conversion unit 12 performs step S106,
The determination in step 109 is performed by using the unit information (such as sending data) acquired last in the loop of S107.

The phonetic conversion unit 12 determines that the acquired kanji is one character and is not a phonetic kanji (step S
109; N), and when the acquired kanji is two or more characters (step S108; N), the acquired kanji or kanji string and its reading (hiragana character string) are used to read from the homonym dictionary 13 Search for homonym identification information (step S110). Next, the character number identification information corresponding to the character number of the reading, the reading, and the searched homonym identification information are output to the distortionless compression unit 14 in this order (step S).
111). Here, the number-of-characters identification information is one-byte data associated with a numerical value.
Codes other than 81-9F and E0-EF (hexadecimal notation) used as the first byte of the F, A1-DF and 2-byte characters are assigned.

For example, if the acquired kanji string is "product", the phonetic conversion unit 12 in step S111, as schematically shown in FIG. 4, reads "seihin" of "product". 1-byte character number identification information representing the number of characters 4 of "", 8-byte data representing "seihin",
1-byte homonym identification information (0x00 when the content of the homonym dictionary 17 is as shown in FIG. 3).

After outputting the homonym identification information, the phonetic conversion unit 12 outputs the unit information finally obtained in the loop of steps S106 and S107 to the distortionless compression unit 13 (step S10).
112). Then, the output unit information is "EOF"
If not (step S114; N), the processing from step S101 is executed again.

If the acquired kanji is one phonetic kanji (step S109; Y), the phonetic conversion unit 12 outputs the two pieces of acquired unit information (a kanji and a non-kanji or 1
(Byte character) is output to the distortionless compression unit 13 (step S11).
3) Then, the process proceeds to step S114. The separate processing is performed for one on-reading kanji for the following reason. Many of the Onyomi Kanji have a large number of homonyms, and some of them have a number of homonyms that cannot be expressed in one byte (for example, "Sanseido, Shinmeiken Kanji Dictionary 3rd Edition 1987" 362 words are listed as homonyms of “ko”.) For this reason, when the same processing is performed for one on-read kanji with another kanji, the number of types of homonym word identification information increases, and as a result, the significance of conversion to reading decreases.

The phonetic conversion unit 12 repeats such processing and outputs "EOF" (step S1).
14; Y), the processing on the plaintext data to be compressed ends.

Although illustration is omitted, in practice, when branching to the “N” side in step S 108, the phonetic conversion unit 12 stores the acquired kanji string in the homonym dictionary 17. It is decomposed into kanji idioms (often kanji) of a size in which homonym identification information is defined, and processing corresponding to steps S110 and S111 is performed on each decomposed kanji idiom. That is, when a kanji string composed of two kanji idioms stored in the homonym dictionary 17 is included in the plaintext data, the phonetic conversion unit 12 outputs the first kanji idiom of the kanji string. After the data before and after the character number identification information and the homonym identification information are added, the data to which the character number identification information and the homonym identification information are added before and after the second kanji phrase are output. Thereafter, characters other than the kanji that existed after the kanji string are output. If the unit information obtained from the plaintext data represents a character other than a kanji, the phonetic conversion unit 12 outputs the unit information as it is. Thus, for example,
The sentence "Semiconductor product sales are strong."
The phonogram conversion unit 12 converts the data into phonogram data as schematically shown in FIG.

In the data compression / decompression device of the first embodiment, such speech data output from the speech conversion unit 12 is compressed by the distortionless compression unit 13. Then, the result of compression by the distortionless compression unit 13 is stored in the storage unit 11 as compressed data, and the data compression is completed. That is, in the present data compression / decompression device, the plain data can be restored to the original state (decompression procedure will be described later), and the data is converted into phonogram data that is data with a smaller number of used characters. . Then, the phonetic data is compressed.
For this reason, in the present data compression / decompression device, the collection of statistical information at the time of compression is performed more efficiently than in the case of directly compressing plaintext data. Compression will be achieved.

Next, the operation of the data compression / decompression device of the first embodiment at the time of data decompression will be described. Distortion-free restoration unit 14
Reads the compressed data designated as the data to be restored from the storage unit 11 and converts the result of the compression data, that is, the phonetic data created based on the plaintext data by the phonetic conversion unit 12 into the phonetic inverse conversion. To the unit 15.

As is apparent from the operation of the phonogram conversion unit 12, the phonogram data includes 2-byte characters, 1-byte characters, character number identification information, and homonym identification information (see FIG. 5). ). Among them, the homonym identification information is one-byte information in which a numerical value is represented by a binary number, and thus cannot be discriminated from a one-byte character or the like. However, before the homonym identification information, there is character number identification information that can be distinguished from the first byte of the one-byte character and the two-byte character. The homonym identification information exists at a position where the character number identification information exists and at a position determined by the content.

For this reason, the phonetic inverse converter 15 restores the plaintext data from the phonetic data according to the flow shown in FIG. The phonetic inverse conversion unit 15 first receives the 1
Byte data is acquired (step S201). Then, the phonetic inverse conversion unit 15 determines whether the data is a one-byte character, the first byte of a two-byte character, or character number identification information (step S202).

If the acquired data represents a one-byte character (step S202; one-byte character), the phonetic inverse conversion unit 15 outputs the data to the storage unit 11 as it is (step S203). Then, when the output data is not “EOF” (Step S210; N)
In step S201, the process returns to step S201 to acquire the next one-byte data from the distortion-free restoration unit 14. If the acquired data is the first byte of a two-byte character (step S202; two-byte character), the phonetic inverse conversion unit 15 further acquires one-byte data, and The data is output to the storage unit 11 (Step S204).
Then, the process returns to step S201.

If the acquired data is character number identification information (step S202; character number identification information), the phonetic inverse conversion unit 15 recognizes the number of characters of the succeeding hiragana character string based on the content of the character number identification information. (Step S20
5). Then, the phonetic inverse conversion unit 15 outputs the data of the number of recognized characters (data of the number of characters × 2 bytes; reading) and the data of the next 1 byte (homophonetic identification information) from the distortionless restoration unit 14. Acquire (Steps S206, S20
7). Next, the phonetic inverse conversion unit 15 uses the acquired reading and the homonym identification information to read from the homonym dictionary 18
Search for kanji or kanji idioms (step S20)
8). Then, the searched kanji or kanji phrase is output to the storage unit 11, and the process returns to step S201.

The phonetic inverse conversion unit 15 repeats such an operation for each data from the distortion-free restoration unit 14, and outputs “EO
When F "is output (Step S210; Y),
The illustrated processing ends.

<Modification> Data compression / compression of the first embodiment
The restoration device is a device in which phonogram data in a form in which kanji or kanji idioms in plaintext data are replaced with data comprising character number identification information and reading and homonym word identification information is created. It goes without saying that the apparatus may be configured so that phonetic data in a form in which homonym identification information and reading are added in this order after the information is created.

Also, instead of preparing a plurality of character number identification information, start position identification information and end position identification information are prepared, and when the kanji reading is output at the time of data compression, the start position is set before and after that. The apparatus may be configured to add the identification information and the end position identification information. At this time, the position of the homonym identification information is set immediately after the start position identification information or immediately after the end position identification information. Further, the apparatus is configured so that data restoration is performed by returning the reading existing between the start position identification information and the end position identification information to kanji.

Further, the data compression / decompression device of the first embodiment is a device for directly outputting a kanji code when one kanji kanji is to be processed. By using the Kanji reading of the kanji as a reading, it is also possible to configure the apparatus such that even one kanji that is originally a phonetic reading kanji is converted into a reading.

The data compression / decompression device of the first embodiment is intended for shift JIS.
It goes without saying that the apparatus may be designed for other code systems such as S and EUC. Further, the technology of the present embodiment is not limited to Japanese and can be applied to any language that requires the definition of multi-byte characters.

<Second Embodiment> The data compression / decompression device of the first embodiment converts kanji idioms contained in plaintext data into
The device converts the phonogram data in a form replaced with data including the reading of the kanji idiom, and then performs compression. On the other hand, the data compression / decompression device according to the second embodiment also replaces data other than kanji idioms with data of another form when creating phonetic data. Further, as the data for replacing the kanji idiom, data including half-width reading of the kanji idiom (information in which the reading is represented by a half-width katakana character string) is used.

FIG. 7 shows a functional block diagram of a data compression / decompression device according to the second embodiment. As is clear from the figure,
The basic configuration of the data compression / decompression device of the second embodiment is the same as the data compression / decompression device of the first embodiment. However, the phonetic conversion unit 12 'provided in the data compression / decompression device of the second embodiment is the same as the homonym dictionary 1
It has only 7 '. The homonym dictionary 17 'is different from the homonym dictionary 17 in that the correspondence between kanji or kanji idioms, half-width reading (information in which the reading is represented by a half-width katakana character string), and homonym identification information is stored. Has become a dictionary. The phonetic inverse conversion unit 15 'is also provided with a homonym dictionary 18' having the same contents as the homonym dictionary 17 '.

When performing data compression, the phonogram conversion unit 12 'creates phonogram data from plaintext data according to the procedure shown in FIG. The phonetic conversion unit 12 'first obtains data for one unit of information from the plaintext data to be compressed (step S301). If the data is a one-byte character (step S302; one byte character)
In step S303, it is determined whether the character represented by the data is a half-width katakana character (step S303). If the character is not a half-width katakana (step S303; N), that is, if it is a control character or a Roman character, the phonetic conversion unit 1
2 'outputs the data as it is to the distortionless compression unit 13 (step S309). And the data is "EO
If it is not F "(step S310; N), the process returns to step S301 to start processing for the next data constituting the plaintext data.

If the obtained one-byte character is a half-width katakana (step S303; Y), the phonetic conversion unit 12 '
Acquires the next unit information (1 byte or 2 byte character) (step S304). If the character represented by the obtained unit information is a half-width katakana (step S
305; Y), the process returns to the step S304. On the other hand, if the character is not a half-width katakana (step S30
5; N), the process proceeds to step S306. In other words, the phonetic conversion unit 12 'executes a loop of steps S304 and S305 to acquire, from the plaintext data, a character string in which half-width katakana is continuous and unit information for the next one character.

Thereafter, the phonetic conversion unit 12 'outputs the half-width katakana start position identification information, the half-width katakana character string, and the end position identification information in this order (step S306).
The half-width katakana start position identification information and the end position identification information are 1-byte information whose contents are set so as to be distinguishable from 1-byte characters, 2-byte characters, and character number identification information. (Details will be described later). Data existing between these pieces of position identification information is determined to be data to be output as half-width katakana (to be output as it is).

Next, the phonetic conversion unit 12 'determines whether the character represented by the unit information acquired in the last executed step S304 is a one-byte character or a two-byte character (step S308). If the character is a one-byte character (step S308; one-byte character), the process proceeds to step S309 to output a one-byte character other than the half-width katakana character to the distortionless compression unit 13.

On the other hand, when the last acquired unit information is a 2-byte character (step S308; 2-byte character)
If the character represented by the unit information acquired in step S301 is a two-byte character (step S302; two-byte character), the phonetic conversion unit 12 ', as shown in FIG. Is a hiragana, a kanji, a katakana, or a symbol (ruled line segment or the like) (step S320). More specifically, it is determined whether or not the corresponding half-width katakana character is a character (eg, hiragana, punctuation, etc.). If the corresponding half-width katakana character exists (step S320; hiragana),
The half-width katakana character code corresponding to the two-byte character (character to be processed) is output (step S321), and the process returns to step S301 in FIG.

On the other hand, if the corresponding half-width katakana character does not exist (step S320; katakana, kanji, etc.), the phonetic conversion unit 12 'transmits the plaintext data until a character different from the character to be processed appears. (Steps S322 and S323). That is, if the character to be processed is katakana, the unit information is repeatedly obtained until a character other than katakana (kanji, hiragana, 1-byte character) appears. If the character to be processed is kanji, The acquisition of unit information is repeated until a character other than appears. If the characters to be processed are classified as symbols, the acquisition of unit information is repeated until kanji, katakana, and the like appear.

Then, when unit information representing different types of characters can be obtained (step S323; N), the phonetic conversion unit 12 'determines that the type of the collected characters is a kanji (step S324; N). ), The half-width reading and homonym identification information associated with the acquired kanji string (or kanji) are retrieved from the homonym dictionary 17 '(step S325). Next, the phonetic conversion unit 12 'determines the number-of-characters identification information corresponding to the number of characters (number of bytes) of half-width reading,
The half-width reading and the searched homonym identification information are output to the distortionless compression unit 13 in this order (step S326).
Then, the process returns to step S302 in FIG. 6 to start the processing for the last acquired unit information.

If the character type is katakana (step S324; Y), half-width katakana character string information corresponding to the obtained katakana character string information is created (step S327). Then, the full-width katakana start position identification information, the half-width katakana character string information and the end position identification information are written in this order,
Output to the distortionless compression unit 13 (step S328),
Returning to step S302, processing for the last acquired character is started. The full-width katakana start position identification information is a 1-byte character whose content is set so that it can be distinguished from 1-byte characters, 2-byte characters, character number identification information, half-width katakana start position identification information, and end position identification information. Information. At the time of data restoration, data existing between full-width katakana start position identification information and end position identification information is determined to be data to be converted to full-width katakana and output.

If the character type is a symbol (step S 324; symbol), the phonetic conversion unit 12 ′ outputs the obtained unit information (column) relating to the symbol (column) to the distortionless compression unit 13 as it is. Then, returning to step S302 in FIG.
Processing for the last acquired unit information is started.

The phonetic conversion unit 12 'repeats such processing and outputs "EOF" (step S10).
310; Y), the process for the plaintext data to be compressed ends.

As described above, the phonetic conversion unit 12 '
Replaces hiragana in plaintext data with half-width katakana, and replaces kanji idioms with information consisting of the number of characters identification information and the pronunciation and homonym identification information represented by the half-width katakana character string. Further, the full-width katakana character string is replaced with information consisting of full-width katakana start position identification information, half-width katakana character string equivalent to the full-width katakana character string, and end position identification information. Then, the half-width katakana character string is replaced with information consisting of half-width katakana start position identification information, a half-width katakana character string equivalent to the full-width katakana character string, and end position identification information.

For this reason, sentences that do not include symbols, for example,
As shown schematically in FIG. 10, the sentence "when the data is stored in the memory ..." is converted by the phonetic conversion unit 12 'into one-byte characters and several types of identification information that is one-byte information. It is converted to phonetic data containing only

Since such phonogram data is compressed by the distortionless compression unit 13, the present data compression / decompression device performs compression more efficiently than in the case of directly compressing plaintext data. Will be. Specifically, as schematically shown in FIG. 11, in a conventional device that directly compresses plaintext data, one character is two bytes, so that a certain two-byte character (in the figure, "and" When the first byte of ()) is a target character, only one character existing before the two-byte character is treated as a context. On the other hand, in this device, one character is one byte,
When a certain character (“G”) is the target character,
The two characters that precede that character are treated as context. That is, in the present apparatus, a compression procedure that takes into account the context of a certain order functions as a compression procedure that takes into account the context of twice the order. Therefore, according to the data compression / decompression device of the second embodiment, compression at a high compression ratio can be realized.

Here, the operation of the data compression / decompression device of the second embodiment at the time of data decompression will be briefly described. FIG.
As shown in FIG. 2, when data is restored, the phonetic inverse conversion unit 15 '
First, 1-byte data output from the distortionless restoration unit 14 is obtained (step S401). Next, the phonetic inverse conversion unit 15 'is data representing the character number identification information, the full-width katakana start position identification information, the half-width katakana start position identification information, or the half-width katakana character. It is determined whether the data is data representing a kanji (symbol) (step S402).

If the data is character number identification information (step S402; character number identification information), the phonetic inverse conversion unit 15 'recognizes the character number (byte number) from the character number identification information (step S403). Then, data (half-width reading) for the number of recognized characters is obtained from the distortion-free restoration unit 14 (step S404). Then, the phonetic inverse converter 15 '
Acquires the next one byte of data, that is, homonym identification information (step S405). Next, the phonetic inverse conversion unit 15 'searches the homonym dictionary 17' for information representing a kanji or a kanji string using the obtained half-width reading and homonym identification information (step S406), and searches the searched kanji ( Column) information is output to the storage unit 11 (step S407). Then, returning to step S401,
Start processing for the remaining data.

In step S402, step S4
When the data obtained in step 01 is detected to be full-width katakana start position identification information or full-width katakana start position identification information, the phonetic inverse conversion unit 15 ′ performs distortionless restoration until end position identification information is obtained. The data acquisition from the unit 14 is repeated (step S410). Then, the phonetic inverse conversion unit 1
5 ′, when the end position identification information is acquired, the step S410 ends, and when the current processing is the detection of the full-width katakana start position identification information (step S411; Y), the step S410 is performed. Acquired in S410,
Data related to half-width katakana characters (strings) (excluding end position identification information) is converted to data related to full-width katakana characters (strings) and output to storage unit 11 (step S413).
Then, the process returns to step S401. On the other hand, if the current processing is based on the detection of the half-width katakana start position identification information (step S411; N), the data (excluding the end position identification information) regarding the obtained half-width katakana character (string) is Output to the storage unit 11 as it is (step S
412), and return to step S401.

The data acquired in step S401 is 1
If it is the first byte of a byte character or a double-byte character (step S402; other), the phonetic inverse conversion unit 1
5 'executes the following processing in step S415. For the half-width katakana data, the corresponding hiragana code is output. For the first byte of the two-byte character, another one-byte data is further obtained from the distortionless restoration unit 14, and the two-byte data is output. For other 1-byte characters, the acquired data is output to the storage unit 11 as it is.

The phonetic inverse conversion unit 15 'repeats the above processing and, when outputting "EOF" (step S417; Y), ends the illustrated processing. <Modification> The data compression / decompression device of the second embodiment is also
Modifications similar to those of the data compression / decompression device of the first embodiment are possible. That is, instead of preparing a plurality of character number identification information, start position identification information and end position identification information are prepared. The device may be configured to add the end position identification information to the document, or may be a device that targets a document of another code system.

The data compression / decompression device according to the second embodiment is a device for converting double-byte characters to half-width katakana characters. However, the data compression / decompression device may be configured to convert double-byte characters to Roman characters. You can also. In other words,
The apparatus may be configured so that data representing "SEIZOU" is output instead of data representing.

When the present technology is applied to a language consisting of only two-byte ideographic characters, for example, an international phonetic symbol, information representing a vowel and a consonant of a Hangul character, information representing a note character, etc. A code may be assigned so that each ideographic character data is converted into a code representing the sound of the ideographic character with an international phonetic symbol or the like.

<Third Embodiment> FIG. 13 shows a functional block diagram of a data compression / decompression device according to a third embodiment. First, the outline of the operation of the data compression / decompression device according to the third embodiment will be described with reference to FIG.

As shown, the data compression / decompression device according to the third embodiment includes a storage unit 21 used for storing plaintext data and compressed data. Also, as blocks that function at the time of data compression, there are provided a phonetic conversion unit 22, an intermediate code conversion unit 23, a temporary storage unit 24, and a distortionless compression unit 25. As blocks that function at the time of data restoration, a distortionless restoration unit 26, an intermediate code inverse conversion unit 27, and a phonetic inverse conversion unit 28 are provided.

The Japanese analysis dictionary 16 and the homonym dictionary 17 of the phonetic conversion unit 22 are the same as the dictionary of the phonetic conversion unit 12 of the first embodiment. The phonetic conversion unit 22 uses these dictionaries to convert kanji in plaintext data into “reading” in substantially the same procedure as the phonetic conversion unit 12 of the first embodiment, thereby obtaining phonetic data. Generate However, when generating the phonogram data, the phonogram conversion unit 22 also creates a character list which is a list of characters used in the phonogram data and identification information.

The intermediate code conversion unit 23 includes the phonetic conversion unit 22
Uses the phonetic data and the character list generated in the temporary storage unit 24 to generate an intermediate code correspondence table for assigning a new code (intermediate code) to characters and identification information used in the phonetic data. create. Then, the intermediate code conversion unit 23
Converts the phonetic data into intermediate code data, which is data including an intermediate code, using the intermediate code table. In addition,
The intermediate code conversion unit 23 is provided with, as operation modes, a mode in which the number of bits of the intermediate code is not specified (bit length non-designation mode) and a mode in which the number of bits of the intermediate code is specified (bit length specification mode). When operating in the bit length specification mode, information specifying the bit length of the intermediate code is provided to the intermediate code conversion unit 23 prior to data compression.

The distortion-free compression unit 13 compresses the intermediate code correspondence table and the intermediate code data supplied from the intermediate code conversion unit 23 without loss, and stores the compression results in the storage unit 21. To be stored.

The distortion-free restoration section 26 has a function of restoring the data compressed by the distortion-free compression section 25, and restores and outputs an intermediate code correspondence table and intermediate code data based on the compressed data instructed to be restored. I do. The intermediate code inverse conversion unit 27 uses the intermediate code correspondence table given from the distortionless restoration unit 26,
The intermediate code included in the intermediate code data given thereafter is returned to the original information (phonetic data). The phonetic inverse conversion unit 28 uses the homonym dictionary 18 which is a dictionary having the same contents as the homonym dictionary 17 to return the “reading” included in the phoneme data to kanji, thereby obtaining a distortion-free restoration unit. 2
6 generates, in the storage unit 11, plaintext data which is the original information of the processed compressed data.

Hereinafter, the operation of each unit will be specifically described. First, the operation of the phonetic conversion unit 22 will be described with reference to FIG. As shown, when the start of data compression is instructed, the phonetic conversion unit 22 first sets “0” to a variable K (step S500). Although the use of the variable K will be described later, the variable K is a variable in which the maximum value of the homonym identification information output by the phonetic conversion unit 22 is stored.

Next, the phonetic conversion unit 22 obtains data of one character from the plaintext data designated as the compression target (step S501). Next, the phonetic conversion unit 22
The type of the character represented by the data is determined, and if the character is a non-kanji (step S502; non-kanji), the acquired data is output to the temporary storage unit 24 and the data is registered in the character list. If the data has not been registered, the data is registered in a character list (step S503). The output data is "E
OF ”(step S511; N),
Returning to step S501, processing for the data of the next one character in the plaintext data is started.

If the data read in step S501 is data representing a kanji (step S502; kanji), the phonetic conversion unit 22 repeats data acquisition from plaintext data until non-kanji data is acquired (step S502). S
504, 505). When the non-kanji data is acquired (step S505; non-kanji), the phonetic conversion unit 22
Proceeds to step S506.

In step S506, the phonetic conversion unit 2
No. 2 determines whether or not the data obtained before the non-kanji data is data for one phonetic Kanji character. And
If it is data for one on-read kanji (step S5
06; Y), the start position identification information, its on-read kanji data, and the end position identification information are output to the temporary storage unit 24 (step S507). The contents of the start position identification information and the end position identification information are set so that they can be distinguished from the first byte of a one-byte character or a two-byte character.
In this embodiment, the same information is used as start position identification information and end position identification information.

After the execution of step S507, the phonetic conversion unit 2
2 outputs the last acquired data (non-kanji data) to the temporary storage unit 24 (step S508). Next, if the start position identification information output in step S507 is unregistered data that is not registered in the character list,
The data (start position identification information) is registered in a character list. If the data output in step S508 is unregistered data, it is registered in the data character list (step S509). That is, the phonetic conversion unit 22
When the step S509 is first executed, the start position identification information is registered in the character list. If the data existing next to the end position identification information is unregistered data, Also perform registration. When step S509 is executed for the second time or later, since the start position identification information has already been registered, only when the data existing next to the end position identification information is unregistered data, Register data to the character list.

Then, the phonetic conversion section 22 proceeds to step S5.
If the data output in step 08 is not "EOF" (step S511; N), the process returns to step S501.
On the other hand, when the data obtained before the non-kanji data is not data for one phonetic kanji (step S506;
N), the phonetic conversion unit 22 performs the phonetic conversion process (step S5).
Execute 10).

FIG. 15 shows the flow of the phonetic conversion process. As shown in the figure, at the time of the phonetic conversion process, the phonetic conversion unit 22 first converts a kanji string or a kanji represented by data obtained before the non-kanji data into a kana character string (reading) (step S601). . Next, the phonetic conversion unit 22 acquires, from the homonym dictionary 17, homonym identification information associated with the reading and the kanji string (kanji) (step S6).
02). Then, the character number identification information indicating the number of characters of the reading and the phonetic and homonym identification information are output to the temporary storage unit 24 in this order (step S603), and the finally obtained data (non-kanji) is temporarily stored. The data is output to the storage unit 24 (step S604).

Thereafter, the phonetic conversion unit 22 proceeds to step S6.
03, among the data output in S604, if there is unregistered data in the data excluding the homonym identification information, the data is registered in the character list (or those data) (step S605). When the value of the homonym identification information exceeds K, the value is set to K (step S606), and the phonetic conversion process ends (proceeding to step S511 in FIG. 14).

When it is detected that "EOF" has been output (step S511; Y), the phonetic conversion unit 22
Ends the process shown in FIG. In this way, the phonetic conversion unit 22 creates, in the temporary storage unit 24, phonetic data having substantially the same form as the phonetic data output by the phonetic converting unit 12 of the first embodiment. The phonetic conversion unit 22 also creates a character list that is a list of character number identification information, start position identification information, one-byte characters, and two-byte characters included in the phonetic data. The maximum value of the included homonym identification information is stored in the variable K, and the processing shown in the figure is terminated.

The intermediate code conversion unit 23 starts operating when the processing by the phonetic conversion unit 22 is completed. As described above, the intermediate code conversion unit 23 has two operation modes. Here, the operation in the bit length non-designation mode will be described first with reference to FIGS.

When the generation of the phonetic data by the phonetic conversion unit 22 is completed, the intermediate code converting unit 23 firstly sets the minimum value necessary to represent the value of K in binary as shown in FIG. Number of bits M and number of character types n stored in the character list
Is obtained (step S701). Next, the intermediate code conversion unit 23 calculates ^L that satisfies 2 ^L-1 <n ≦ 2 ^L (Step S702). If the calculated value L is smaller than M (step S703; Y), the value of M is set to L (step S704), and the process proceeds to step S705. On the other hand, if L <M is not satisfied (step S70)
3; Y) includes the step S without updating the value of L.
Proceed to 705.

In step S705, the intermediate code conversion unit 23 assigns an L-digit binary number, that is, an L-bit intermediate code, to each of the n types of information (character number identification information, 2-byte characters, etc.) in the character list. By assigning
Create an intermediate code correspondence table. Next, the intermediate code conversion unit 2
3 supplies data representing the contents of the created intermediate code correspondence table to the distortionless compression unit 25 (step S706).

After that, the intermediate code conversion unit 23 acquires data for one unit information from the phonogram data stored in the temporary storage unit 24 (step S711). That is, in this step, the intermediate code conversion unit 23 acquires 1-byte data, and if the data is the first byte of a 2-byte character, further acquires 1-byte data. On the other hand, if the acquired data is one-byte characters, character number identification information or start position identification information,
Step S711 ends without acquiring data. If the data acquired in step S711 is the start position identification information (step S712; start position identification information), the intermediate code conversion unit 23 stores the data in the intermediate code correspondence table in association with the start position identification information. An intermediate code (hereinafter, referred to as an intermediate code for start position identification information)
The data is supplied to the distortionless compression unit 25 (step S713). Next, in the phonetic data, 3-byte data following the start position identification information, that is, data (2
Byte) and the end position identification information (1 byte) are supplied to the distortionless compression unit 25 as they are (step S714).
After that, the intermediate code conversion unit 23 returns to step S711 and starts processing for the next data in the phonetic data.

If the acquired data is character number identification information (step S712; character number identification information), the intermediate code conversion unit 23 recognizes the number of characters to be read from the character number identification information, and responds to the character number identification information. Then, the intermediate code stored in the intermediate code correspondence table is supplied to the distortionless compression unit 25 (step S715). Next, data corresponding to the number of recognized characters is obtained, and an intermediate code corresponding to each data is supplied to the distortionless compression unit 25 (step S71).
6). Then, the next one byte of data, that is, the homonym identification information, is obtained from the phonetic data, and the homonym identification information is converted into L-bit information, and the distortion-free compression unit 2 acquires
5 (step S717). That is, L> 8
, The intermediate code conversion unit 23 adds “L−8” “0” s to the upper bit side of the homonym identification information.
The bit information is supplied to the distortionless compression unit 25. Also, L <
If it is 8, L-bit information obtained by removing “8-L” “0” s from the higher-order bit side in the homonym identification information is supplied to the distortionless compression unit 25. Naturally, L = 8 , The homonym identification information is used as is in the distortionless compression unit 2.
5 After that, the intermediate code conversion unit 23 returns to Step S711.

If the obtained data is a one-byte character or a two-byte character (step S712; other), the intermediate code conversion unit 23 supplies an intermediate code corresponding to the data to the distortionless compression unit 25 ( Step S71
8). If the supplied data is not “EOF” (step S719; N), step S711
To start the process for the next data and return to "EOF"
Is satisfied (step S719; Y), the process ends.

As described above, in the non-bit length designation mode, the intermediate code conversion unit 23 converts the one-byte characters included in the phonetic data, the two-byte characters excluding one phonetic kanji, the number-of-characters identification information, and the start-position identification information. Find the minimum number L of bits that can be represented. At this time, if the obtained L is less than the number of bits M required for the homonym identification information, the value of L is changed to M
Update with the value of Then, in the phonetic data, each piece of information excluding one phonetic kanji and homonym identification information is converted into an L-bit intermediate code, and the expression form of the homonym identification information is converted into L bits. One on-read Kanji character is not converted, and is used as an element of the intermediate code data as it is.

For example, when the kanji phrase “product” exists in the plaintext data, as shown schematically in FIG. 18, the phonetic data includes the character number identification information and the same sound as the reading “seihin”. 10 bytes of information including the synonym identification information are included. The intermediate code conversion unit 23 converts this information into information including six L-bit data (five intermediate codes and L-bit homonym identification information) as shown in FIG. .

If the plaintext data contains a phrase including the phonetic kanji character “ma” (reading; kan), the phonetic data includes, as schematically shown in FIG. The information includes four-byte information indicating “the”, start position identification information, two-byte information indicating “between”, and end position identification information. The intermediate code conversion unit 23
This information is converted into information including three L-bit intermediate codes, two-byte information representing "between", and one-byte end position identification information, as shown in FIG.

Next, the operation of the intermediate code converter 23 in the bit length designation mode will be described. In the bit length designation mode, the intermediate code conversion unit 23 converts phonetic data into intermediate code data according to the flowcharts shown in FIGS.

As shown in FIG. 20, the intermediate code conversion unit 2
3, the minimum number of bits M required to represent the value of K, which is the maximum value of the homonym identification information, in a binary number as in the non-bit length designation mode, and stored in the character list. The number of character types n is obtained (step S801). Next, the intermediate code conversion unit 23 determines the magnitude relationship between the designated bit length L 'and M, and if L'<M (step S8).
02; Y), the value of M is set in L '(step S8).
03), and the process proceeds to step S804. On the other hand, if L '<M is not satisfied (step S802; N), the process proceeds to step S804 without updating the value of L'.

In step S804, the intermediate code conversion unit 23 determines whether 2 ^L ≧ n is satisfied (step S804). If 2 ^L ′ ≧ n is satisfied (step S804; Y), that is, if all n types of information stored in the character list can be represented by L ′ bit data, the intermediate code conversion unit 23 An intermediate code correspondence table is created by assigning an L'-bit intermediate code to each of the n types of information (character number identification information, 2-byte characters, etc.) in the character list (step S805).

On the other hand, if 2 ^L ′ ≧ n is not satisfied (step S 804; N), that is, if the L′-bit data cannot represent n kinds of information stored in the character list, The intermediate code converter 23 calculates n
^{_{exc (= n-2 L '}} ) to clear the species characters from the character list (step S806). At this time, the intermediate code conversion unit 23
Deletes characters (symbols) other than Hiragana, thereby leaving the character number identification information, start position identification information, and Hiragana characters in the character list. Then, an intermediate code correspondence table is created by assigning an L 'bit intermediate code to each of the 2 ^L ' types of information in the character list (step S).
807).

After the creation of the intermediate code correspondence table, the intermediate code conversion unit 23 supplies data representing the contents of the created intermediate code correspondence table to the distortionless compression unit 25 (step S808). Then, the processing shown in FIG. 21 is started in order to convert the phonetic data into the intermediate code data.

Steps S811-S81 shown in FIG.
7, the intermediate code conversion unit 23 performs the same processing as in steps S711 to S717 (FIG. 17), respectively. Therefore, the description of the operation in these steps will be omitted.

By the way, in the bit length non-designating mode, an intermediate code is assigned to each 2-byte character and each 1-byte character included in the phonetic data except for one phonetic Kanji character. For this reason, when the data obtained from the phonetic data is a one-byte character or a two-byte character, an intermediate code corresponding to the data exists.

On the other hand, in the bit length designation mode,
For some characters contained in phonetic data,
There is no corresponding intermediate code. Therefore, when the intermediate code conversion unit 23 branches to the “other” side in step S812, the data acquired in step S811 is replaced with data to which no intermediate code is assigned (hereinafter, referred to as excluded character data). ) Is determined (step S820). If the data is not the exclusion character data (step S820; N), the intermediate code corresponding to the data is supplied to the distortionless compression unit 13 (step S820).
821). Next, the intermediate code conversion unit 23 determines whether the output data is “EOF”,
If not F ”(step S825; N), the process returns to step S811 to start processing for the next data.

On the other hand, if the acquired data is the exclusion character data (step S820; Y), the intermediate code conversion unit 23 repeats the acquisition of the data from the phonetic data until data other than the exclusion character data appears (step S820; Y). Step S82
3). Then, the intermediate code for the start position identification information, the data excluding the data acquired last (data other than the exclusion character data), and the end position identification information are supplied to the distortionless compression unit 25 (step S824). Next, the intermediate code conversion unit 2
No. 3 returns to step S812 to process data that is not the last excluded character data.

That is, in the bit length designation mode, if there is a character to which no intermediate code is assigned in the phonetic data, the intermediate code conversion unit 23 sets the intermediate code for start position identification information and the end code before and after the character. The information to which the position identification information is added is supplied to the distortionless compression unit 25. In addition, when a plurality of such characters are continuous, information obtained by adding the start position identification information intermediate code and the end position identification information before and after the character string is supplied to the distortionless compression unit 25.

For example, information representing a character string “αβγ is” exists in the phonetic data, and step S806 is performed.
, It is assumed that information on “α”, “β”, and “γ” has been deleted from the character list. In this case, the intermediate code converter 23, as schematically shown in FIG.
(= L ′) bit start position identification information intermediate code;
The information including the intermediate code corresponding to “αβγ”, the end position identification information, and “ha” is supplied to the distortionless compression unit 25.

According to the data compression / decompression device of the third embodiment, the intermediate code correspondence table and the intermediate code data generated from the phonetic data are converted into the distortionless compression unit 2 by the above procedure.
5, and the compression of the plaintext data is completed.

Finally, the operation of the data compression / decompression device according to the third embodiment when decompressing data will be described. When instructed to decompress certain compressed data, the distortion-free decompression unit 26 decompresses the compressed data relating to the intermediate code correspondence table constituting the compressed data. Next, the distortion-free restoration unit 26 restores the compressed data related to the intermediate code data.

In response, intermediate code inverse conversion section 27
Executes the processing shown in FIG. That is, the intermediate code inverse conversion unit 27 first obtains an intermediate code correspondence table from the distortionless restoration unit 26, and recognizes the number L of bits of the intermediate code (step S901). Next, from the distortionless restoration unit 26,
Acquire L-bit data (intermediate code) (step S
902). Then, information associated with the acquired intermediate code is read from the intermediate code correspondence table (step S
903).

Thereafter, if the read information is the start position identification information (step S904; start position identification information), the intermediate code reverse conversion unit 27 converts the next 1-byte or 2-byte data into a distortion-free restoration unit. 26 (step S905). Then, when the acquired data is not the end position identification information (Step S906; N)
Output the data (excluding the end position identification information) as it is to the phonetic inverse conversion unit 28 (step S907).
Then, the process returns to step S905. On the other hand, when the acquired data is the end position identification information (step S906;
Y), without outputting the data, step S
Returning to step 902, data acquisition in units of L bits is resumed.

If the data read from the intermediate code correspondence table is the character number identification information (step S904; character number identification information), the intermediate code inverse conversion unit 27 recognizes the number of characters in the subsequent reading and processes the read data. It is converted into 1-byte character number identification information and output to the phonetic inverse conversion unit 28 (step S908). Next, the intermediate codes for the number of recognized characters are acquired from the distortion-free restoration unit 26, and data associated with each acquired intermediate code in the intermediate code correspondence table is output (step S909). After that, the intermediate code inverse conversion unit 27 obtains L bits of data, that is, homonym identification information, converts the homonym identification information into 1-byte information, and outputs the same to the phonetic inverse conversion unit 28. (Step S910). Then, the process returns to step S902, and data acquisition in L-bit units is restarted.

If the data read from the intermediate code correspondence table is neither start position identification information nor character number identification information (step S904; other), the intermediate code reverse conversion unit 27
Outputs the data to the phonetic inverse conversion unit 28 (step S911). If the output data is not “EOF” (step S912; N), the process returns to step S902, and if it is “EOF” (step S912).
At 912; Y), the illustrated processing ends.

As described above, when the intermediate code for start position identification information appears, the intermediate code inverse conversion unit 27 performs
It recognizes that the data until the end position identification information appears is data in byte units (one-reading Kanji character data or excluded character data). Then, the data sandwiched between the intermediate code for the start position identification information and the end position identification information is output to the phonetic inverse conversion unit 28 as it is.
At this time, the output of the start position and end position identification information is not performed. For homonym identification information that cannot be distinguished from the intermediate code by itself, the position is specified based on the position of the intermediate code for character number identification information, and the data is returned to 1-byte data.

As a result, phonetic data obtained by removing all of the start position identification information and the end position identification information from the phonetic data output by the phonetic conversion unit 22 is supplied to the phonetic inverse conversion unit 28. . For this reason, the phonetic inverse conversion unit 28
The phonetic data is converted to plaintext data in exactly the same procedure as the phonetic inverse converter 15 of the embodiment.

[0125]

As described above, according to the present invention, it is possible to compress document data relating to a language in which one character is represented by a plurality of bytes at a high compression rate.

[Brief description of the drawings]

FIG. 1 is a functional block diagram of a data compression / decompression device according to a first embodiment of the present invention.

FIG. 2 is an explanatory diagram of a homonym dictionary provided in the data compression / decompression device of the first embodiment.

FIG. 3 is a flowchart showing an operation procedure of a speech conversion unit provided in the data compression / decompression device of the first embodiment.

FIG. 4 is an explanatory diagram of an operation of a phonetic conversion unit included in the data compression / decompression device of the first embodiment.

FIG. 5 is an explanatory diagram of an operation of a phonetic conversion unit included in the data compression / decompression device of the first embodiment.

FIG. 6 is a flowchart showing an operation procedure of a phonetic inverse conversion unit provided in the data compression / decompression device of the first embodiment.

FIG. 7 is a functional block diagram of a data compression / decompression device according to a second embodiment of the present invention.

FIG. 8 is a flowchart showing an operation procedure of a phonetic conversion unit provided in the data compression / decompression device of the second embodiment.

FIG. 9 is a flowchart showing an operation procedure of a phonetic conversion unit provided in the data compression / decompression device of the second embodiment.

FIG. 10 is an explanatory diagram of an operation of a phonetic conversion unit included in the data compression / decompression device of the second embodiment.

FIG. 11 is an explanatory diagram showing why data can be compressed at a high compression ratio by the data compression / decompression device of the second embodiment.

FIG. 12 is a flowchart showing an operation procedure of a phonetic inverse conversion unit provided in the data compression / decompression device of the second embodiment.

FIG. 13 is a functional block diagram of a data compression / decompression device according to a third embodiment of the present invention.

FIG. 14 is a flowchart showing an operation procedure of a phonetic conversion unit provided in the data compression / decompression device of the third embodiment.

FIG. 15 is a flowchart of a phonetic conversion process performed by a phonetic conversion unit included in the data compression / decompression device of the third embodiment.

FIG. 16 is a flowchart showing an operation procedure of the intermediate code conversion unit provided in the data compression / decompression device of the third embodiment when the bit length is not specified.

FIG. 17 is a flowchart showing an operation procedure of the intermediate code conversion unit provided in the data compression / decompression device of the third embodiment when the bit length is not specified.

FIG. 18 is an explanatory diagram of an operation of an intermediate code conversion unit included in the data compression / decompression device of the third embodiment.

FIG. 19 is an explanatory diagram of an operation of an intermediate code conversion unit included in the data compression / decompression device of the third embodiment.

FIG. 20 is a flowchart showing an operation procedure when a bit length is specified in an intermediate code conversion unit provided in the data compression / decompression device of the third embodiment.

FIG. 21 is a flowchart showing an operation procedure when a bit length is specified in an intermediate code conversion unit provided in the data compression / decompression device of the third embodiment.

FIG. 22 is an explanatory diagram of an operation of an intermediate code conversion unit included in the data compression / decompression device of the third embodiment.

FIG. 23 is a flowchart showing an operation procedure of an intermediate code inverse conversion unit provided in the data compression / decompression device of the third embodiment.

FIG. 24 is a diagram illustrating the correspondence between characters, occurrence probabilities, and sections in arithmetic coding.

FIG. 25 is an explanatory diagram showing an encoding procedure in arithmetic encoding.

FIG. 26 is an explanatory diagram of a context.

FIG. 27 is a diagram showing an example of a tree structure used for context collection.

[Explanation of symbols]

 11, 21 storage unit 12, 22 phonetic conversion unit 13, 25 distortion-free compression unit 14, 26 distortion-free restoration unit 15, 28 phonetic inverse conversion unit 23 intermediate code conversion unit 24 temporary storage unit 27 intermediate code inverse conversion unit

Claims (35)

[Claims] 1. A data compression apparatus for compressing original document data including character information in which one character is represented by a plurality of bytes of information, wherein each character information included in the original document data to be compressed is converted into the character information. Phonetic document data creating means for creating phonetic document data which is data replaced with phonetic character information representing the sound when the associated character is pronounced, and phonetic document created by the phonetic document data creating means A data compression device, comprising: compression means for compressing data. 2. A data compression apparatus for compressing original document data including character information in which one character is represented by a plurality of bytes, comprising: a plurality of conversion target word information including one or a plurality of character information; Phonogram information storage means for storing phonogram information which is information representing a sound when the word indicated by the information is pronounced, and phonogram information storage means for storing the phonogram information storage means from the original document data Search / read means for searching conversion target word information and reading phonogram information corresponding to the searched conversion target word information from the phonogram information storage means; and the original document searched by the search / read means Phonetic document data for creating phonetic document data by replacing the conversion target word information in the data with the conversion target word replacement information including the phonetic character information read by the search / read means. Creating means, and an intermediate code table creating means for creating an intermediate code table for associating an intermediate code with information elements used in the phonetic document data created by the phonetic document data creating means, Intermediate code document data for creating intermediate code document data by converting each information element constituting the phonetic document data into a corresponding intermediate code using the intermediate code table created by the intermediate code table creating means. A data compression apparatus comprising: a creation unit; and a compression unit that compresses the intermediate code document data created by the intermediate code document data creation unit. 3. An intermediate code table creating means for allocating, to each information element used in the phonetic document data, an intermediate code having a minimum number of bits capable of expressing the information element. 3. The data compression device according to claim 2, wherein an intermediate code table is created. 4. The phonetic document data creating means replaces a predetermined type of character information with start position identification information and end position identification information indicating that the character information is a predetermined type of character information. By replacing with information, phonetic document data is created, The intermediate code table creating means, in the phonetic document data,
The information subsequent to the start position identification information and the end position identification information, create an intermediate code table that is not associated with an intermediate code, the intermediate code data creating means, in phonetic document data,
4. The data compression apparatus according to claim 2, wherein the information following the start position identification information and the end position identification information are not converted into an intermediate code. 5. When the type of information element used in the phonetic document data exceeds the number N of information that can be represented by a predetermined number of bits, the intermediate code table creating means generates the intermediate code table. From among the used information elements, "N-1" information elements to which an intermediate code is assigned are selected, and the selected "N-1" information elements and the start position identification information have different contents. Creating an intermediate code table for associating the intermediate code of the predetermined number of bits, wherein the intermediate code data creating means is for the information in the phonetic document data to which the intermediate code is associated in the intermediate code table. Converts that information into the corresponding intermediate code,
Regarding the unallocated information that is the information to which the intermediate code is not allocated, the unallocated information is the information having the intermediate code at the head associated with the start position identification information, and including the unallocated information. 3. An intermediate-coded document data is created by replacing with unassigned replacement information which is information in a form in which an end position is known.
A data compression device as described. 6. The phonetic character information storage means for discriminating between the plurality of conversion target words and other conversion target words to which the same phonetic character information is associated with the plurality of conversion target words. The search / readout unit reads out phonetic character information and homonymous identification information corresponding to the searched conversion target word, and the phonetic document data creating unit includes: The conversion target word information in the original document data is replaced with conversion target word replacement information including phonetic character information and homonym identification information read by the search / read means, and the intermediate code creating means includes 6. The data compression apparatus according to claim 2, wherein an intermediate code table is created for information elements other than homonym identification information. 7. A data compression apparatus for compressing original document data including character information in which one character is represented by a plurality of bytes, wherein a word indicated by conversion target word information corresponds to conversion target word information including one or more character information. A phonetic character information storage unit storing information on a phonetic character which is information representing a sound generated when the sound is pronounced, and conversion target word information stored in the phonetic character information storage unit from the original document data. Searching and reading means for reading from the phonetic character information storage means the phonetic character information corresponding to the searched conversion target word information, and converting the original document data searched by the searching and reading means. Phonetic document data creating means for creating phonetic document data by replacing target word information with conversion target word replacement information including phonetic character information read by the search / read means, A data compression apparatus comprising compression means for compressing the phonetic document data created by the phonetic document data creating means. 8. The phonetic character information storage means identifies the phonetic character information and another conversion target word to which the same phonetic character information is associated with the plurality of conversion target word information. The search / readout means reads phonetic character information and homonym identification information corresponding to the searched conversion target word information, and the phonetic document data creating means The conversion target word information in the original document data, the conversion target word replacement information including the phonetic character information and the homonym identification information read by the search / read means, and the homonym identification information is 8. The data compression device according to claim 2, wherein replacement is performed with conversion target word replacement information included in a specific position. 9. The phonetic document data creating means adds numerical identification information indicating the length of the phonetic character information read by the search / read means to the head of the conversion target word replacement information. 9. The data compression device according to claim 2, wherein: 10. The phonetic document data creating means converts the target word information into phonetic character information start position identification information indicating the start and end of the target word information and the phonetic character information end. 10. The data compression apparatus according to claim 2, wherein phonetic document data is created by substituting conversion target word substitution information sandwiched between position identification information. 11. A decompression means for decompressing compressed document data, and converting the phonetic character information representing a sound when a character is reproduced, decompressed by the decompression means, into corresponding character information, A data restoration device comprising: original document data generation means for generating original document data which is data that is the source of document data. 12. Phonetic character information in which phonetic character information that is information representing a sound when the word indicated by the conversion target word information is pronounced with respect to the conversion target word information including one or a plurality of character information is stored. Storage means, decompression means for outputting intermediate code document data by decompressing the compressed document data, and each intermediate code included in the intermediate code document data output by the decompression means, in an intermediate code table for the compressed document data. A phonetic document data generating means for generating phonetic document data by substituting the information associated with the intermediate code, and a conversion included in the phonetic document data generated by the phonetic document data generating means The target word replacement information is searched, and the searched conversion target word replacement information is associated with the phonetic character information included in the conversion target word replacement information. A data restoration apparatus comprising: original document data generating means for generating original document data that is a source of the compressed document data by replacing the compressed document data with conversion target word information stored in a storage means. 13. The phonetic document data generation means uses the intermediate code table for information sandwiched between an intermediate code corresponding to start position identification information and end position identification information output by the restoration means. 13. The data restoration apparatus according to claim 12, wherein the replacement is not performed and the information is used as it is as an element of the phonetic document data. 14. The intermediate code table is a code table in which "N-1" information elements and start position identification information are associated with intermediate codes of a predetermined number of bits having different contents from each other. Document data generation means, for the unassigned replacement information starting with the intermediate code associated with the start position identification information included in the intermediate code document data output by the restoration means, from the unassigned replacement information, the intermediate code and the 14. The data restoration apparatus according to claim 13, wherein information is output from which information added to identify the end position is removed. 15. The phonetic character information storage means, for each of the plurality of conversion target word information, identifies a phonetic character information and another conversion target word to which the same phonetic character information is associated. The original document data generating means searches for the conversion target word replacement information contained in the phonetic document data generated by the phonetic document data generating means, and performs the search. The conversion target word replacement information is replaced with the conversion target word information stored in the phonogram information storage unit in association with the phonogram information and the homonym identification information included in the conversion target word replacement information. 15. The data restoration apparatus according to claim 12, wherein the data restoration is performed. 16. A phonogram in which phonogram information, which is information representing a sound when a word indicated by each conversion target word information is pronounced, is stored for conversion target word information including one or a plurality of character information. Character information storage means, restoration means for outputting phonetic document data by restoring the compressed document data, and search for and search for conversion target word substitution information contained in the phonetic document data generated by the restoring means. By replacing the conversion target word replacement information with the conversion target word information stored in the phonogram information storage unit in association with the phonogram information included in the conversion target word replacement information, the compression is performed. A data restoration apparatus comprising: original document data generation means for generating original document data that is the source of document data. 17. The phonetic character information storage means, for each of the plurality of conversion target word information, distinguishes between phonetic character information and another conversion target word having the same phonetic character information. The original document data generating means searches for conversion target word replacement information contained in the phonetic document data generated by the restoration means, and performs the searched conversion target word replacement. By replacing the information with the conversion target word information stored in the phonogram information storage means in association with the phonogram information and the homonym identification information included in the conversion target word replacement information, 15. The method according to claim 12, wherein the original document data which is the basis of the compressed document data is generated.
7. The data restoration device according to any one of 6. 18. The method according to claim 12, wherein said original document data creating means searches for information starting with numerical identification information indicating the length of phonetic character information when searching for said conversion target word replacement information. The data restoration device according to claim 17. 19. The original document data creation means searches for information sandwiched between phonetic information start position identification information and phonetic information end position information when searching for the conversion target word replacement information. The data restoration device according to claim 12, wherein 20. A data compression method for compressing original document data including character information in which one character is represented by a plurality of bytes of information, wherein each character information included in the original document data is converted based on the original document data. A phonetic document data creating step of creating phonetic document data which is data replaced with phonetic character information representing a sound generated when a character associated with the character information is pronounced; and A compression step of compressing the phonetic document data created in the processing. 21. A data compression method for compressing original document data including character information in which one character is represented by a plurality of bytes, comprising: converting original document data to be compressed to conversion target word information including one or more character information; On the other hand, while searching for the conversion target word information stored in the dictionary in which the phonogram information that is the sound when the word indicated by the conversion target word information is pronounced is stored, and the searched conversion target word information A reading / reading step of reading phonetic character information corresponding to the word from the dictionary; and a conversion target word information in the original document data searched by the searching / reading step, a table read by the searching / reading step. A phonetic document data creating step of creating phonetic document data by replacing with the conversion target word replacement information including phonetic information, An intermediate code table creating step for creating an intermediate code table for associating an intermediate code with each information element used in the phonetic document data created by the creating step, and the intermediate code table creating step An intermediate code document data creating step of creating intermediate code document data by converting each information element constituting the phonetic document data into a corresponding intermediate code using the intermediate code table thus obtained; A compression step of compressing the intermediate code document data created by the data creation step. 22. The intermediate code table creating step, for each information element used in the phonetic document data,
22. The data compression method according to claim 21, further comprising a step of creating an intermediate code table for assigning an intermediate code having a minimum number of bits capable of expressing these information elements. 23. The intermediate code table creation step, if the type of information element used in the phonetic document data exceeds the number N of information that can be represented by a predetermined number of bits, From among the used information elements, "N-1" information elements to which an intermediate code is assigned are selected, and the selected "N-1" information elements and the start position identification information have different contents. It is a step of creating an intermediate code table for associating the intermediate code of the predetermined number of bits, the intermediate code data creating step, in the phonetic document data, the intermediate code is associated with the intermediate code table For the information that has been assigned, the information is converted into a corresponding intermediate code, and for the unassigned information that is information to which no intermediate code has been assigned, the intermediate information associated with the start position identification information is used. Is a step of creating intermediate-coded document data by replacing with unassigned replacement information, which is information having unsigned information at the head and including unassigned information, and in a form in which the end position is known. 23. A method according to claim 22.
Data compression method described. 24. A data compression method for compressing original document data including character information in which one character is represented by a plurality of bytes, wherein a word indicated by the conversion target word information corresponds to a conversion target word information including one or a plurality of character information. Using a dictionary in which phonogram information, which is information representing a sound when pronounced, is stored, from the original document data to be compressed, searching for conversion target word information stored in the dictionary, A search / read step of reading phonetic character information corresponding to the searched conversion target word information from the dictionary; and a search / read step of converting conversion target word information in the original document data searched by the search / read step. A phonetic document data creating step of creating phonetic document data by replacing with the conversion target word replacement information including phonetic character information read by A compression step of compressing the phonetic document data created by the phonetic document data creating step. 25. The searching / reading step for identifying, for each of the plurality of conversion target words, phonetic character information and another conversion target word having the same phonetic character information. Reading the phonetic character information and the homonym identification information corresponding to the conversion target word from the dictionary in which the homonym identification information is stored, wherein the phonetic document data creation step includes: 22. The step of replacing the conversion target word information with the conversion target word replacement information including the phonetic character information and the homonym identification information read in the search / read step. Item 25. The data compression method according to any one of Items 24. 26. The phonetic document data creating step,
26. The method according to claim 21, wherein numerical identification information indicating a length of the phonetic character information read in the search / read step is added to a head of the conversion target word replacement information. Data compression method described. 27. A decompression step of decompressing compressed document data, and converting the phonetic character information representing a sound when a character is reproduced, which is decompressed in the decompression step, into corresponding character information, An original document data generating step of generating original document data which is data that is a source of the document data. 28. A decompression step of outputting intermediate code document data by decompressing compressed document data, and converting each intermediate code included in the intermediate code document data output by the processing of the decompression step to the compressed document data. A phonetic document data generating step of generating phonetic document data by substituting information associated with the intermediate code table, and converting target word information comprising one or more character information; The conversion target included in the phonetic document data generated by the processing of the phonetic document data generating step using a dictionary in which phonetic character information that is information representing a sound when the word indicated by is generated is stored. Search for word replacement information, and convert the searched conversion target word replacement information to the conversion target word information corresponding to the phonetic character information included in the conversion target word replacement information. In by replacing the data restoring method which is characterized in that it comprises a source document data generation step of generating the original document data is the source of the compressed document data. 29. The intermediate code table is a code table in which “N−1” information elements and start position identification information are associated with intermediate codes of a predetermined number of bits having different contents from each other. The document data generating step includes, for the unallocated replacement information that is information that starts with an intermediate code associated with the start position identification information included in the intermediate code document data output by the processing of the restoration step, 29. The data restoration method according to claim 28, further comprising outputting information obtained by removing information added to identify the intermediate code and its end position from the intermediate code. 30. A restoring step of outputting phonetic document data by restoring compressed document data, and when a word indicated by the conversion target word information is pronounced with respect to the conversion target word information including one or a plurality of character information. Using a dictionary in which phonogram information, which is information representing the sound of the utterance, is stored, the conversion target word replacement information included in the phonogram document data generated by the processing of the restoration step is searched for. Original document data generation that replaces word replacement information with conversion target word information corresponding to the phonogram information included in the conversion target word replacement information, thereby generating original document data from which the compressed document data is based. And a data restoring method. 31. The original document data generating step identifies, for each of the plurality of conversion target word information, phonogram information and another conversion target word having the same phonogram information. Conversion target word replacement information including phonetic character information and homonym identification information in the phonetic document data generated by the processing of the restoring step using a dictionary in which homonym identification information and homonym identification information are stored. 31. The data restoration method according to claim 28, wherein is replaced with the corresponding conversion target word information. 32. The original document data creating step searches for the conversion target word replacement information by searching for information sandwiched between start position identification information and end position information. 32. The data restoration method according to claim 31. 33. The original document data creating step searches for the conversion target word replacement information by searching for information sandwiched between phonetic information start position identification information and phonetic information end position identification information. 32. The data restoration method according to claim 28, wherein: 34. A phonogram which is data obtained by replacing each character information included in original document data to be compressed with phonogram information representing a sound when a character corresponding to the character information is pronounced. A program recording medium in which a phonetic document data creating means for creating document data and a program for functioning as a compressing means for compressing the phonetic document data created by the phonetic document data creating means are recorded. 35. A computer, comprising: restoring means for restoring compressed document data; and converting phonetic character information representing a sound when a character is reproduced, restored by the restoring means, into corresponding character information. And a program recording medium on which a program for functioning as original document data generating means for generating original document data which is data on which the compressed document data is based is recorded.