JPH0555912B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a Japanese information retrieval system for searching data matching search conditions from a database consisting of a large amount of character information in Japanese.

(Prior Art) A conventional Japanese information retrieval system of this type has a data section for storing data in Japanese;
It is equipped with a list that connects and arranges these Japanese words starting from a predetermined basic Japanese language, and when registering data, the Japanese words that make up this data are registered in the list, and when searching for information, it is possible to search based on search conditions. The list is scanned based on the keyword, and data specified by the result is read from the data section.

(Problems to be Solved by the Invention) In such conventional methods, since a hierarchically structured list is used for searching, the search speed deteriorates as the amount of data in the database increases, and keyword searches using words are difficult. is performed, so if it is uncertain whether the string to be used for the search is appropriate, there may be cases where the string is not hit.
When registering data, it is often necessary to register multiple Japanese words in a list along with one data, so there is a problem in that the registration work is cumbersome.

(Means for Solving the Problems) The Japanese information retrieval system of the present invention includes a data section for storing data in Japanese, and a section for specifying numbers of all the data for each predetermined character. A multiple bit table each consisting of a bit string is provided, and when registering the data in the data section, the data number is generated and the data number designation bit in the multiple bit table is assigned to each of all the characters in the data. When turned on, when searching for information using a specified character string as a key, a logical operation is performed based on the search condition on the bits at the same position in the bit string that correspond to all the characters in the string. Accordingly, the data of the specified data number is read out from the data section.

(Example) Next, an example of the present invention will be described with reference to the drawings.

Here, for concrete expression, the multiplicity of the multiple bit table is assumed to be 2, and each is expressed as a primary bit table (or upper index) and a secondary bit table (or lower index). Assume the size is 1000 bits.

FIG. 1 is a block diagram showing one embodiment of the present invention.

In order to avoid complicating the drawing and simplify the explanation, Figure 1 shows two primary tables U ₁ and U ₂ ,
two secondary tables L ₁ and L ₂ and a data section DA
is shown, a primary table and a secondary table are provided for each character. In other words, although Figure 1 corresponds to only two characters, the characters used as search keys are JIS 1.
Level/2nd level kanji (approximately 6400 characters), hiragana,
In the case of katakana and numbers, the same number of sets of first bit tables and second bit tables as the number of these characters are required.

The data section DA has one million entries, and each entry stores one piece of data. Assuming that the data number uniquely assigned to data is N, an entry is designated by coordinates (X, Y, Z) that satisfy equation (1).

N=1000000 (X _-1 ) + 1000 (Y _-1 ) + Z... (1) The primary bit tables U ₁ and U ₂ and the secondary bit tables L ₁ and L ₂ constitute an index section for the data section DA. do. The addresses of the primary bit tables U ₁ and U ₂ are specified by the coordinate value x (1 in this embodiment) which is the number of the upper index, and the bit length thereof is 1000.
Further, the addresses of the secondary bit tables L ₁ and L ₂ correspond to the bits of the upper index, that is, the bit positions of the primary bit tables U ₁ and U ₂ , and are the coordinate values Y (001 to 1000), and its bit length is 1000. Secondary bit table
The bit positions of L ₁ and L ₂ correspond to the coordinate value Z (001 to 1000), which is the bit of the lower index.

As is clear from the above explanation, the data part DA
The data is indexed by a double index of the primary bit tables U ₁ , U ₂ and the secondary bit tables L ₁ , L ₂ .

Next, the operation of this embodiment will be explained separately into data registration and information retrieval.

FIG. 2 is for explaining the operation at the time of data registration in this embodiment, taking data "information storage and retrieval" as an example. The primary bit tables U 1 and U ₂ and the secondary bit tables L _{1 and} L ₂ , which are shown individually for each character in FIG. 1 to facilitate understanding, are not shown in FIGS. 2 and 3. Returned to their actual form and integrated, each
Represented as UX and LX.

Now, when the data "Information storage and retrieval" is input, this system assigns a data number (123002 in this example) and sets the coordinates of the data section (1124002) based on the above formula. Data number 123002
Store with.

Next, extract all the characters included in the data "Information storage and retrieval", and write each character to the number of block address 001 (hereinafter referred to as UX001) of the primary bit table UX as follows.
The 124 bits and the 002nd bit of each character corresponding address in the secondary bit table LX124 are set to logic "1". Although the kana characters "no" and "to" are not shown in FIG. 2, the kana characters are also registered in the same way as the kanji characters.

As is clear from Figure 2, for the characters ``jo'' and ``product'', the first bit table UX0
01 and all other bits of the secondary bit table LX 124 are logic "0", indicating that these characters were previously unregistered.
The first bit table UX is used for the character “information”.
Since the 001st bit of 001 is "1", any hit position (referred to as Z) in the secondary bit table LX001 should be "1", and the data of data number Z is the coordinate ( 1001Z).

FIG. 3 is for explaining the operation at the time of information retrieval in this embodiment, taking the character string "information search" as an example. Assume that you are instructed to search for all data containing the character string "information search" as a search condition. The reason why the contents of the primary bit table UX and the second bit table LX in Fig. 3 are different from those in Fig. 2 is that the character string ``information retrieval'' was created from the time of registration of the data ``information storage and retrieval''. This is because other data such as the data "Tokai Television Information Search System" was registered before the information search, and the number of "1" bits increased.

When the character string "information search" is input, the characters "information", "information",
The contents of each address corresponding to "search" and "search" are ANDed with the same bit positions.

Based on the result of this first logical AND operation, the first
The 003rd bit and the 003rd bit of the next bit table UX001
Assuming that the 124 bit becomes "1", the contents of each address corresponding to the characters "information", "information", "search", and "search" are the same for the second bit tables LX003 and LX124, respectively. A logical AND operation is performed between bit positions.

Based on the result of this second logical AND operation, the second
Regarding the next bit table LX003, all bits become "0", and the second bit table LX12
For 4, the 002nd bit and 456th bit are “1”
Suppose that it becomes From this, the data group corresponding to the secondary bit table LX003 includes "emotion",
It is found that there is no data that contains the four characters ``report'', ``search'', and ``search'', which are specified by data numbers 123002 and 123456 of the data group corresponding to the secondary bit table LX124. It can be seen that the data contains these four characters.

Therefore, data number from data part DA
Data 123002 and 123456 are read, but the data number 123002 is "information storage and retrieval" and includes all the constituent characters of "information retrieval" used as a search key, but does not include the word "information retrieval". However, it can be seen that the data with data number 123456 is "Tokai Television Information Search System" and includes the word "information search," so the latter is the data that meets the search conditions.

In this case, the searcher's memory is uncertain and the search key is "information search", but after looking at the search result data displayed on the CRT screen of the information search terminal, the search conditions are "information" and "search". It is possible that we often recall that the data contained . In such a case, you can change the search conditions as described above and start the search again.

In the embodiment described above, the multiplicity of the index is set to 2, but the present invention is not limited to this. For example, a triple index in which formula (2) is used instead of formula (1) is used. An embodiment employing this method can be easily realized.

N=1000000000(X _-1 )+1000000(Y _-1 )+1000
(Z _-1 )+W ...(2) Furthermore, in the embodiment described above, data including all the character strings of the search key is searched as a search condition, but other search conditions,
For example, an embodiment that can support a search condition that searches for data that includes one of a plurality of character strings, such as one of "information" and "search", is also possible. This can be easily realized by changing the contents of logical operations between bit positions.

Note that although the above explanation has been made regarding information retrieval in Japanese, the present invention is even more suitable for information retrieval in Chinese, which employs a kanji structure.

Furthermore, the multiple bit table in the present invention is
It is applicable not only to information retrieval systems but also to general indexes (for checking existence, etc.).

(Effects of the Invention) According to the present invention, unlike the conventional method, words that may be used as search keywords are hierarchically listed in the index section, and the list is scanned with the search keyword to perform the search. Instead of finding data that meets the conditions, the system uses a ``dictionary method'' that turns on the corresponding data number designation bit in the multiple bit table for all characters included in the registered data at the time of data registration. Since the number of characters is finite, even if the amount of registered data increases, the scale of the index section will naturally converge.Furthermore, when searching for information, multiple bit tables can be read out character by character and searches can be made for the same bit position. By adopting a "free word search" that performs logical operations based on conditions, the number of search steps is reduced, so there is no decrease in search speed, and the search keywords are broken down at any time and intermediately Since the search proceeds character by character, it has the effect of eliminating the need for multiple Japanese registrations associated with one piece of data, which is done in the conventional method, in order to deal with the uncertainty of search keywords.

[Brief explanation of drawings]

FIG. 1 shows a diagram for explaining one embodiment of the present invention, and FIGS. 2 and 3 show diagrams for explaining the operation of this embodiment. U ₁ , U ₂ , UX...primary bit table, L ₁ ,
_L2 ...Second bit table, DA...Data section.

Claims (1)

[Scope of Claims] 1. A data section for storing data in Japanese, and a multiple bit table each consisting of a bit string for specifying the number of all the data for each predetermined character, When registering data in the data section, generate the data number, turn on the data number designation bit in the multiple bit table for each of the characters in the data, and use the specified character string as a key. When searching for information, a logical operation is performed based on the search condition on bits at the same position in the bit string that correspond to all the characters in the character string, and the data of the data number specified by the result is stored in the data section. A Japanese information retrieval system characterized by reading information from the .