JPH0272481A - Character string retrieving device by logical expression and control system for such device - Google Patents

Abstract

PURPOSE:To decrease storage capacity and to execute content retrieve by means of a logical expression efficiently and at high speed by representing variety in the expression of a retrieve character string by means of a conception code, and registering it. CONSTITUTION:A first memory means, whose access address is determined by the output of a character collating means 110 to discriminate to which retrieve character string out of the plural strings of a text record corresponds and which accumulates a concept code to be related to each retrieve character string, a mark bit register 130, that temporarily stores which concept code is included in units of a text register, and a second memory means 140, whose access address is determined by the output of the mark bit register 130 and which stores the logical expression of a retrieve condition, are provided. By unifying the character strings of synonyms to the concept code, the storage capacity necessary for the registration of the retrieve condition logical expression is reduced. Thus the storage capacity of the second memory means 140 is reduced, and the whole sentences can be promptly retrieved by the logical expression of the character string data.

Description

[Detailed Description of the Invention] (Field of Industrial Application) The present invention relates to a character string search device and its control method that can efficiently and quickly process content searches using logical formulas for text data such as patents and academic literature. It is something.

(Conventional technology) Text information such as patents and academic literature is available on OA (office,
Automation) continues to increase as equipment technology advances. In order to be reused, they are stored in a file memory whose capacity increases one after another. However, creating a database of file data that allows users to quickly search their contents and extract the necessary information is not progressing smoothly. That is, the work of adding a keyword that matches the content to each record of file data and rearranging the keywords so as to speed up the keyword search requires manual labor and is not easily accomplished. Furthermore, even if the contents match, there is also the problem of variations in the keywords used during the search that do not match the keywords added to the records.

Therefore, various techniques for real-time character string searches have been studied to find ways to quickly search text data in file memory without adding keywords or rearranging data.

Generally, a character string search device has a function of finding out where in text data a search character string given as a keyword is included. In order to realize this function, it is essential to have a string matching means that simultaneously compares and matches each string in text data with a plurality of search strings.

Therefore, the inventors of the present application received in parallel the associative memory unit that stores a plurality of character strings of arbitrary lengths, and the character comparison results of the associative memory unit for each character of the input character string. We have filed a patent application for a symbol string identification device that is configured to enable character string matching using a sequence logic unit that compares the arrangement (order) of characters between strings. Although the names are different, the above devices are the basic elements of a character string search device. (Unexamined Japanese Patent Publication No. 6
0-211539 (Specification of Japanese Patent Application No. 59-068495), Japanese Unexamined Patent Publication No. 61-253536 (Specification of Japanese Patent Application No. 1983)
0-96213), etc.) In such a character string search device, after registering a keyword (search character string) to be searched, when a text string to be searched is input, the text string matches one of the search strings. When a match occurs, it generates a match signal and outputs which part of the text string matches which search string. Therefore, it is possible to perform a full-text search on text string data without keywords. That is, text file information can be created into a database without adding keywords to text string information stored in a file.

However, information retrieval using the full text search method is an index search method in which the file storage location of text string information that matches the search conditions is found by searching only the keyword part of text string data with keywords added. requires much longer search time. If consideration is given to matching the search condition logical expressions, the search processing speed will further decrease. When matching is performed with the help of software, there is a noticeable drop in speed, making it difficult to put the full-text search method into practical use.

(Problem to be Solved by the Invention) Conventional character string search devices only match the keyword (search character string) included in the logical expression of the search condition with the character string of text data. They relied on software (programs) to check whether they matched. Therefore, even if character string matching is achieved at high speed, there is a problem in that it takes a long time to determine whether or not the logical expression of the search condition is met.

Furthermore, in one search condition theoretical expression, the search string contained therein has many synonyms, making it impossible to find a text record that meets the search request. To avoid such a situation, if you register all the search strings for synonyms and express the logical expression of the search condition including them in the product-sum standard form, the number of product terms will be very large. There was a problem in that the amount of hardware that had to be registered was too large.

An object of the present invention is to provide a character string technique that solves the above two problems.

(Means for Solving the Problems) Therefore, the present invention provides a string matching means that stores a plurality of search strings and determines which of the plurality of search strings an input string of a text record matches; An access address is determined by the output of the column matching means, and a first storage means stores a concept code associated with each search character string, a mark bit register temporarily stores which concept code is included in each text record, and the mark bit. A character string search device using a logical formula, wherein an access address is determined by the output of a register, and a second storage means for storing a logical formula of a search condition, wherein a control method of this character string search device is included in the logical formula. A character string initial registration process in which a search character string and its concept code are associated with each other and registered in a character string matching means and a first storage means; ), the output of the first storage means and the mark bit register are used to determine the access address of the second storage means, and write true logic signals in the order of product terms. In the initial expression registration process, when the search string included in a new logical expression does not match the search string registered in the string matching means, it is used as a new search string, and the string is matched by matching it with the concept code. a search string additional registration process for additionally registering the search string in the means and the first storage means, and the output of the string matching means and the first storage means and the mark bit register for the search string that is true and included in each product term of the new logical expression. The method is characterized in that it performs registration control including a logical formula additional registration process in which the access address of the second storage means is determined using the second storage means, and true logical signals are written into another storage area in the order of product terms. Furthermore, in the control method of the above-mentioned string search device, when performing a full-text search after registering a search condition logical expression, the string matching means receives string data sequentially in text record units, and when string matching occurs, A character string matching process that accesses a first storage means and sets its output in a mark bit register; and a string matching process that accesses a second storage means using the output of the mark bit register at the time of text record switching, and performs a check against a logical expression. A search control including a logical expression search process that outputs a result, and a reset process that returns only the mark bit register corresponding to the registered concept code to the reset state after accessing the second storage means, and leaves the rest in the marked state. It is characterized by doing the following.

(Function) In the present invention, a search condition logical formula can be registered, and after registering it, the entire text string data in the text file memory can be searched to find a text record that matches the logical formula without using software.

Further, according to the present invention, the storage capacity required for registration in the storage means can be reduced by combining the synonymous character strings into a concept code for a search condition logical expression that includes a character string with many synonyms. For example, what can be summarized in a conceptual code is as follows. There are synonyms for expressing GaAs, such as GaAs, gallium arsenide, gallium arsenide, and gallium arsenide, but the idea is to unify them into a code such as "0010".

As an example of a search using concept codes, let us consider a case where a search condition logical expression containing five search strings is registered. Assuming that each search string has 5 synonyms, 55
Therefore, there are 225 search condition logical expressions, and the number of product terms in each logical expression is 225. Therefore, in a normal manner, 225×55 (approximately 100 billion) memory cells are required to register one set of search condition logical expressions. Using the conceptual code, a logical formula with 25 product terms can be reduced to 1
The number of memory cells used is reduced to 25 (32). This is an effective method of converting the search condition logical formula into the product-sum standard form and registering it. According to the drawing below,
A more detailed explanation of the present invention will now be given.

(Embodiment) FIG. 1 shows the configuration of a first embodiment of the present invention. The string matching means 110 uses content addressable memory (CAM, abbreviation for content addressable memory) and sequential logic (FSA).

It consists of a finite state automaton (abbreviation for finite state automaton) and an encoder, can store a plurality of search strings, and determines which of the plurality of stored search strings an input string matches.

The access address of the first storage means 120 is determined by the output of the character string matching means 110, and the result of decoding the conceptual code 121 associated with each search string at the X decoder 123 and the result of decoding the search string at the X decoder 122. It is stored in the form of bit information of 1 and 0 corresponding to the presence or absence of the relationship. Therefore, the memory matrix represents an example of storing conceptual codes.

The character string matching means 110 compares the character string A1. B1. B2. C1
, C2, Dl.

B2. Assume that B3 and E are memorized, and the respective concept codes are A, B, B, C/A, C/B, D,
D, D and E/F. Therefore, string A1
' from the first left column of the first storage means 120.
1'' is output, and when the character string B1 or B2 is given, 11111 is output from the second left column, and the character string D1 or B2 is output.
When B3 is given, "1"' is output from the fourth column on the left. This output encoded by encoder 124 represents a conceptual code.

When the character string C1 is given, 1'' is output from the first and third left columns, but this occurs when the character string C1 straddles two concept codes A and C.

For example, if the character string A1 is a semiconductor and C1 is an HBT, the concept code A may be MOS, bipolar, or the like. On the other hand, as the concept code C, compound semiconductors, bipolar, etc. can be considered. In other words, the conceptual code of C1 is common to that of A's bipolar. This first storage means 120 also allows the use of three or more concepts. To this end, the first storage means 120 uses a priority encoding function in the encoder 124 that indicates to which concept each character string belongs. When two or more concepts are involved, the priority encoder 124 outputs concept codes 125 in order of priority.

The mark bit register 130 connected to the first storage means 120 comprises a concept code decoder 131 and a number of registers 132 equal to the number of concepts. When the record address is changed, the contents of all registers 132 become zero by a recent pulse applied from the reset terminal 133. After that, when the concept code is given from the first storage means 120,
By decoding it, a mark bit "1" is set in the selected register 132. As a result, "1" mark bits corresponding to all concept codes included in the text record being searched are successively written into the register 132.

The contents of the register 132 when writing is completed determines the access address of the second storage means 140.

The second storage means 140 expresses the logical expression of the search condition in the product-sum standard form, and associates the product term with each address selected via the X decoder 141 according to the output of the mark bit register 130. The logical formula is stored by writing '1'' in the address corresponding to the product term included in the logical formula. is selected by the X decoder 142.The logical formula specified by the selection code 145 and the first storage means 1
The matching result 144 with the concept code outputted by 20 is outputted from this second storage means 140.

The memory matrix 160 of the second storage means 140 shows examples of storing three logical formulas. From mark bit register 130, three conceptual codes (A,
B, C) are given, and the remaining concept codes are assumed to be unregistered and fixed to 1. Therefore, the leftmost row 1 of the memory matrix 160 in the figure
llN ABC, ABm, ABC, ABC,
Eight B.C.

ABC, AErC, ABN, 9. , allocating information. On the other hand, the first row of the memo matrix is a logical formula (ABC)
, the second line shows the storage state when storing the logical formula (ABC Junin BG), and the third line shows the storage state when storing the logical formula (λUC). Assuming that the stored value is "kite" at the j-th bit of the i-address, the following logical formula will be stored in the j-th bit line.

Wl, ABC+W2. ABC+W3. A100C+...W
, are logical 1 or 0 signals determined by writing the desired output for each text record to the address selected by the output of the second storage means 120 for that text record.

When the circuit configuration as described above is realized, the registration of the search condition logical expression therein and the control of the search are performed as follows. First, the search string included in the logical formula and its concept code are paired and the string matching means 110 and the first
Initial registration is made in the storage means 120. This is called the string registration process. Of course, before the initial registration, the string matching means 1
It is assumed that the storage contents of 10 and first storage means 120 have been cleared. In particular, it is assumed that all read outputs of the first storage means 120 are "0". After clearing, when registering the search character string in the character string matching means 110, the registered address is stored in the X decoder 122 of the first storage means 120.
The conceptual code is registered by writing "1" to the bit line selected by the Y decoder 123 for the conceptual code.

Second, "1" is initially registered in the address selected by the output of the mark bit register 130 for the logical product term between conceptual codes included in the logical formula expressed in the product-sum standard form. However, it is assumed that the stored contents of the memory matrix of the second storage means 140 are cleared to logic signal 0 before the initial registration. That is, before registration, Ox A-B-C+ Ox k-B:C+ Ox A-
B-C+..., but the logical formula (ABC+ AB
To register C), IXABC+0XABC+IXA
-B-C..., and write the logic signal 1 given from the input terminal 143 to addresses 1 and 3.

To initially register a large number of different logical expressions, a selection code of the logical expression is given to the Y decoder of the second storage means 140, and the logic signal 1 applied from the input terminal 143 from the upper row to the lower row of the memory matrix is applied. You can write them in order.

After the logical expression of the search condition has been registered, the text data file memory is searched for the full text of the character string data. Search control is performed by first specifying the logical expression of the search condition using the selection code 145, and then sequentially reading out the character string data 150 in units of text records (several 100 bytes to several kilobytes) from the text data file memory. 11
Start by entering 0. Each time the character string data 150 matches a search string registered in the string matching means 110, the first storage means 120 is accessed, and the concept code of the matched search string is stored in the mark bit register 1.
Sent out at 30. This control is called a string search process.

The mark bit register 130 stores the matched concept codes in units of text records as mark pits after being decoded. Register 13 corresponding to registered concept code
The contents of 0 are reset to O when the text record is switched. The contents of other registers 130 remain at 1.

Immediately before being reset, the second storage means 140 is accessed, and the stored content (0 or 1) of the address selected by the output of the mark bit register 130 is output as a result of matching against the search condition logical expression of each text recorder. Ru. These controls are called a logical expression search process and a reset process. After the full-text search process, changes or additions to search conditions occur. The following process is then used:

First, when a search string included in a new search condition logical expression does not match a search string registered in the string matching means 110, it is set as a new search string, and the string matching means 110 associates it with the concept code. and is additionally registered in the J-th storage means. The registration address of a new search string in the string matching means 110 is determined after the previous search string is registered, and is held until additional registration. on the other hand,
Additional registration to the first storage means 120 is performed by the character string matching means 11.
This is achieved by writing a logic signal "'1" at the intersection of the address selected by the output for the new search character string of 0 and the bit string selected by the concept code given from the input terminal 121. This control is called the character string addition registration process.

Next, the logical expression of the new search condition is additionally registered in the second storage means 140. To do this, the new search condition logical expression expressed in standard form is decomposed into multiple product terms, the search strings that are true and included in each term are sequentially fed to the string matching means 110, and the The concept codes output from the first storage means 120 are sequentially stored in the mark bit register 130, and the second storage means 14 stores the concept codes output from the second storage means 120 at an address selected by the output for each product term.
A logic signal 1 is written to 0. At this time, the selection code given to the Y decoder 142 needs to be newly set unless the search condition logical expression is the same as the registered one. For example, when registering the logical formula A-B-C+A-B-C, first, the concept codes A, B, C
The character strings corresponding to A,
When the concept codes B and C are output and stored in the mark focus register 130, a logic signal 1 is written in the second storage means 140 at the address selected by the output thereof.

Next, the character strings corresponding to A and B are checked by the character string matching means 110.
and outputs the concept codes A and H from the first storage means 120, and when they are stored in the mark bit register 130, 1 is written in the second storage means 140 at the address selected by the output thereof. A-B- with two write operations
Finish registering C+A-B-C. This control is called the logical expression addition registration process.

As mentioned above, the control method is divided into registration control and search control, and registration control consists of four processes: initial string registration, initial logical expression registration, additional character string registration, and additional logical expression registration.
Search control consists of three processes: character string search and logical expression search.

(Effects of the Invention) As can be seen from the explanation of the embodiment of the character string search device using logical expressions shown in FIG. They can be registered sequentially, and when there is variety in the expression of the search string, it can be registered as a representative concept code.

Furthermore, when performing a full-text search on character string data stored in the text file memory, in addition to character string matching, the judgment result of whether each text record matches the logical expression of the concept code is used to search the entire text of the text file memory. Column data can be output while being read continuously at high speed.

The effects of the present invention are that when registering search condition logical expressions, the storage capacity of the second storage means 140 for storing logical expressions can be greatly reduced, and that full-text searches using logical expressions of character string data in a text file memory can be performed. The advantage is that it can be done all at once at high speed.

When registering a search condition logical expression, if there are many synonyms in the search string, registering each synonym as a separate logical expression can save the memory of the second storage means 140 instead of reducing the need for the first storage means 120. The capacity becomes really large.

For example, suppose one search condition logical expression contains 5 search strings, and each search string has 5 synonyms.
, there are 55 (3125) combinations of search condition logical expressions. At that time, it is necessary to register 25 search character strings in the character string matching means 110, give a 25-pinto address code to the X decoder 141 of the second storage means 140, and give the Y decoder 142 a lo
It is necessary to provide a selection code of g23125 or more, that is, as many as 12 bits. The number of cells in the memory matrix is 22
Must be 5x3125 pieces. This amounts to about 100 billion, which is impossible. If the first storage means 120 is used to combine 25 search character strings into 5 concept codes, the second storage means 140 receives a 5-bit address code from the X decoder 141 and stores the logic on one bit line. Since formulas are stored, only 125 memory cells are used. This is less than 1/500 million compared to 100 billion. Using the present invention, even if the number of logical expressions increases to 8000, it can be handled with one IMb RAM chip. Assuming that there are about 64 logical expressions, the number of conceptual codes can be increased to 14 with one IMb chip.

Even if the number is increased to 16, it will still fit in one 4Mb chip.

The effect is astronomical.

As described above, according to the present invention, the problem of requiring software to intervene in the search of character string data using logical expressions in conventional character string search devices can be easily solved.

[Brief explanation of the drawing]

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

Claims (3)

[Claims] (1) a string matching means for storing a plurality of search strings and determining which of the plurality of search strings an input string of a text record matches; an access address is determined by the output of the string matching means; a first storage means for storing a concept code associated with each search string; a mark bit register for temporarily storing which concept code is included in each text record; an access address is determined by the output of the mark bit register, and the search 1. A character string search device using a logical formula, comprising: second storage means for storing a logical formula of a condition. (2) A control method for a character string search device using a logical formula according to claim 1, wherein a search character string included in the logical formula and its concept code are associated with each other and registered in a character string matching means and a first storage means. Character string initial registration process, character string matching means for search strings included in true (non-negated form) in each product term of the logical expression converted to product-sum standard form, output of first storage means, and mark bit register is used to determine the access address of the second storage means and write the true logical signals in order of product terms, and the search string included in the new logical formula is registered in the string matching means. A process for adding a search string and registering it as a new search string when it does not match a search string that has been previously searched, and adding it to the string matching means and first storage means in association with the concept code; The access address of the second storage means is determined by using the character string matching means for the search string included in the product term as true, the output of the first storage means, and the mark bit register, and the true search string is stored in another storage area in the order of the product terms. 1. A control method for a character string search device using logical expressions, characterized in that registration control is performed including a logical expression additional registration process of writing logical signals. (3) A control method for a character string search device using a logical formula according to claim 1, wherein during a full text search after registering a search condition logical formula, the character string data is sequentially checked in text record units by a character string matching means. A character string matching process that accesses the first storage means and sets the output in the mark bit register when there is reception and character string matching, and a second storage means that uses the output of the mark bit register when switching text records A logical expression retrieval process that accesses the storage means and outputs the matching result for the logical expression; and after accessing the second storage means, only the mark bit register corresponding to the registered concept code is returned to the reset state, and the rest are returned to the marked state. 1. A control method for a character string search device using a logical expression, characterized in that a search control including a reset process in which the string remains unchanged is performed.