JPH06124304A - Retrieving method using overlap code for extended item - Google Patents

Abstract

PURPOSE:To enable high-speed retrieval by introducing an encoding mean for the extended item and finding a matching item when the extended item is applied to a wild card or variable as index retrieval by pattern matching. CONSTITUTION:In an index generating process 10, hash values are found for pairs of respective attribute names and values of extended items and a function element are found and the index value 15-1 of a set to be retrieved and an index value 15-2 for inquiry are previously generated by OR operation 14 based upon those hash values. An extended item deciding process 20 for deciding extended items with possibility of pattern collating performed according to the logical operation result between the index value 15-1 of the set and the index value 15-2 for inquiry which are generated by the index generating process 10, and then a pattern collating process 30 is carried out.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a search method using a superposition code for an expansion term for high speed processing of information search for structured data.

[0002]

2. Description of the Related Art Conventionally, as a technique in such a field,
For example, some documents were described in the following documents.

Reference 1: The 35th Information Processing Society of Japan (Showa 62)
Second semester) Proceedings of National Convention, Morita et al., "Knowledge Base Machine Using MPPM (3)" 2C-7 Reference 2; Base machine Mu-X
(4) ”5E-7 Reference 3; IPSJ 78th Database System Research Group (1990-7-20) Yokota“ Outline of deduction / object-oriented database language Juan (The)
Outlook of a Destructive an
d Object-Oriented Databases
e Language Juan) ”Generally, in a system such as a deduction database or a knowledge base, a term including a variable is usually used as a basic element of a knowledge expression.

In recent years, not only simple terms used in Prolog etc. but also ψ terms, feature structures, extension terms, etc. can be described in random order and in an indefinite number of arguments by using attribute names instead of their positions. The term expanded to has become widely used. The process of searching for an extension term for pattern matching from a set having the extension term as an element is frequently used as a basic process in a system such as a deductive database or a knowledge base. However, there is no known technique for accelerating the search process for this extended term.

[0005] The closest one is a speed-up technique in term unification search as described in the above-mentioned document. This unification search is to search for a term that can be unified from the term set in the knowledge.
As the scale of knowledge increases, unification is not performed for all knowledge, but search candidates that have the possibility of unification are selected in advance. The term handled in the unification search of this term has a structure and is variable, unlike an ordinary constant. Therefore, a hash value (code) of each data in the structure is superposed according to the structure (SSCW: Structural Superi).
a coded code word) or a code that adds position information to a code and superimposes the code, and inspects the possibility of unification on the code in advance, thereby performing a high-speed unification process. Are known. Hereinafter, the method will be described with reference to FIGS. 2 and 3.

FIG. 2 is a diagram for explaining the creation of a conventional index (index value set), and FIG. 3 is a diagram for explaining a conventional search process.

First, terms are defined. The terms are defined as follows.

Term: = functor (term 1, term 2, ..., Term n) However, in the case of 0 argument, the term is only a functor (constant), but in this case also, the functor (constant) is a term. is there.

When an appropriate term is assigned to a variable for this term, the same term (term that can be unified) is searched for. The processing method is as follows.

Index values are generated for all the terms in the term set to be searched and the terms to be searched (terms that are inquiries). The index value of the term is a bit field of the hash value of the functor, and is the SSCW of term 1, term 2, ..., Term n.
It is generated by arranging so as to cover the bit field by and performing an OR operation (OR operation) on both. Here, when the index value of the term to be searched and the index value of the term to be queried are obtained, all bits are "1" for variables appearing in the term to be searched and all bits are "0" for variables appearing in the term to be queried. Gives the hash value of ".

By comparing the index values of all the terms in the term set to be searched with the index values of the term that is the query, unification candidate items are extracted. Let S be the index value of the item to be searched and Q be the index value of the query, then Q ∧
A term satisfying S = Q is a candidate term. Regarding the candidate term,
Check whether it is possible to unify with the term of the query.

For example, the item to be searched is shown in FIG.
(1) to (5), the search processing when the inquiry term is given as (a) in FIG. 3 is as follows.

Hashing (encoding) the functor and argument 1 and argument 2 of the term with 8 bits, 4 bits and 4 bits, respectively.
To do. The result of hashing is as shown in FIG. However, the hash value of all bits “1” is given to the variable existing in the argument of the item to be searched. Argument 1
And the hash value sequence (8 bits) of the argument 2 is ORed with the hash value (8 bits) of the functor to generate the index value of the term. Item (1) to be searched in FIG.
Regarding (5), the index value of each item is (c) in FIG.

Next, an index value is created for the term that is the inquiry. The result is shown in FIG. However, the hash value of all bits “0” is given to the variable existing in the argument of the inquiry. In the index value S of the search target and the index value Q of the inquiry, terms satisfying Q∧S = Q are search candidates. In this case, as shown in (c) of FIG. ), (4),
Among them, the term (2) is a term that can be unified.

Thus, by narrowing down the candidates by a relatively simple (high speed) logical operation, a high speed search can be performed.

[0016]

However, the conventional method has the following problems.

In recent years, not only simple terms used in Prolog etc. but also ψ terms, feature structures, extension terms, etc. can be described in random order and in an indefinite number of arguments by using attribute names instead of their positions. The term expanded to has become widely used. From these sets with extension terms as elements,
The process of searching for the same extension term is important. However, conventionally, no technique has been known for increasing the search speed for the expansion term.

As a structure search method, a method using SSCW for unifying search for the above-mentioned terms is known. However, in the method using SSCW, the target is the term in Prolog, the number and position of the arguments are fixed, and it cannot be applied to the expansion term etc., so the search method is technically satisfactory. Was difficult to obtain.

The present invention provides a search method using a superposition code for an expansion term, which solves the problem that it is difficult to speed up the search for the expansion term as a problem that the above-mentioned prior art has. To do.

[0020]

In order to solve the above-mentioned problems, the present invention provides a system for expressing data / knowledge by an extension term which is a structure, wherein the extension term is an attribute name consisting of an arbitrary number and an attribute. A nest structure that includes a set of values and describes the extension term in the attribute name and attribute value parts can be handled, and as a search condition, a wildcard or variable is given to the attribute name, attribute value and functor, In the search method for searching the term set of the extension term that matches the attribute name, the attribute value, and the functor, the index value is configured as follows.

That is, the bit field of the function element, the attribute name and the attribute value is determined, and the arrangement of the bit field of the attribute name and the attribute value is arranged so as to be covered by the bit field of the function element, If the attribute name or attribute value is an extension term and the bit field is large enough, the bit field of the attribute name or attribute value is changed to a function of the attribute name or attribute value. An index value is constructed by a bit field of a child, an attribute name, and an attribute value.

Then, a hash value is obtained for each attribute name / attribute value pair and functor of the extension term by the index creation process, and the logical sum operation of the hash values is performed to obtain the hash value for the set to be searched. The configuration index value and the configuration index value for the inquiry are created in advance. Next, after performing expansion term determination processing for determining the expansion term that has a possibility of pattern matching based on the logical operation result of the index value of the set created in the index creation processing and the index value of the inquiry, the pattern I try to collate.

[0023]

According to the present invention, since the search method is configured as described above, a set index value and a query index value are created in advance by the index creation process, and the created result is sent to the extension term determination process. give. In the extension term determination processing, the extension term having a possibility of pattern matching is determined, the number of the extension terms is narrowed down, and then the pattern matching is performed.

That is, by introducing a coding means for the extension term and performing an index search by pattern matching, when a suitable extension term is applied to a wildcard or a variable, a matching one is obtained to search. Speed up. Therefore, the above problem can be solved.

[0025]

FIG. 1 shows an embodiment of the present invention and is a schematic flow chart of a search method using a superposition code for an expansion term.

First, the syntax of the extension term (Syntax) used in this embodiment will be shown. The extension term is an extension of the argument so that a plurality of arguments can be described in any order by using the attribute name or the like instead of the position, and the extension term is defined as the following syntax.

<Extended term> :: = <function element>["["<
Attribute> list “]”] <attribute> :: = <attribute name> “=” <attribute value><attributename> :: = <extended item><attributevalue> :: = <extended item> | “{” <Attribute value> list “}” However, in <extended item>, if there is no <attribute> list (in the case of 0 argument), only <function element> is included, but this <function element> includes a constant.

As can be seen from this syntax, this extended term has the following features.

(1) According to the <attribute> list, the order is
It is possible to describe an indefinite number of arguments. (2) Since the extension term can be described in the attribute name and attribute value portion of the attribute, a structure in which the extension term is nested can be handled. (3) When a rule or the like is described by using the extension term, variables are used as the functor, the attribute name and the attribute value.

In the present embodiment, as shown in FIG. 1, a set of extension terms to be searched and the attribute names, attribute values, and functor parts of the extension terms to be searched are wildcarded or Create an index for queries given variables. Then, the expansion term determination processing 20 searches for an expansion term in which the indexes perform pattern matching, and then the pattern matching processing 30 is performed.

In the index creating process 10, in step 11,
Allocates the bit field of the function, attribute name, and attribute value, and calculates the hash value of each attribute name (step 1
2-1), calculation of hash value of each attribute value (step 12)
-2), and calculation of the hash value of the functor (step 12)
-3) is performed.

After the hash value of each attribute name is calculated,
In step 13-1, it is determined whether the attribute name is an extension term (step 13a). If it is an extension term, the index value for the bit field of the attribute name is calculated (step 13b), and if it is not an extension term, the attribute name is hashed (step 13c). The process after the calculation of the hash value of each attribute value is the same as the process of the attribute name in step 13-1.

Each of the processing results of steps 12-3, 13-1, 13-2 is subjected to the logical sum operation in step 14,
The index value of the set (step 15-1) and the index value of the inquiry (step 15-2) are obtained.

Based on the processing result of the index creation processing 10, the expansion term determination processing 20 is performed, and then the index search is performed by the pattern matching processing 30, which enables high-speed search.

The contents of each of the processes 10, 20, 30 in FIG. 1 will be described in detail below with reference to FIGS.

For example, in FIG. 4 showing the search target and the extension term of the inquiry, the extension terms to be searched of (1) to (8) are pattern-matched with the extension term which is the inquiry shown in (9). The process of searching for an extension term is as follows.

First, each index value is obtained for the expansion terms (1) to (8) which are the retrieval targets shown in FIG. According to the above syntax, an extension term is defined by extension term: = function element (attribute name 1 = attribute value 1, attribute name 2 = attribute value 2, ..., Attribute name n = attribute value n) To do. This index value is, as shown in the structure of the index value of FIG. 5, the attribute name of the function and each argument, and the attribute value (attribute name 1, attribute value 1, attribute name 2, attribute value 2,
.., attribute name n, attribute value n) is hashed into the bit field of the index value, and the hash value is ORed to obtain.

In the structure of index values (1) to (8) shown in FIG. 4, the bit fields of the functor, the attribute name, and the attribute value are 18-bit, 8-bit, and 8-bit hash values, respectively, and the attribute is Arrange so that the sequence of the name and attribute value bit fields is covered by the functor bit field.

The attribute name (attribute value) is, as shown in the syntax of the extension term, the following two formats: (a) an attribute name (attribute value) whose value is unique, and (b) an extended value. There is an attribute name (attribute value) that is a term, and only for the attribute value,
(C) Since there is an attribute value whose value is an attribute value list (set), hashing is performed as follows.

(A) When the value is unique, the value is hashed with 8 bits, and the hash value is assigned to the bit field of the attribute name (attribute value).

(B) If the value is an extension term, a bit field of the attribute name (attribute value) and a bit field of the attribute name and the attribute value are further allocated as the bit field of the attribute name (attribute value). For example, the attribute w in (1) of FIG.
orks = cs [dname = "CS", management
The attribute value cs [dname = "C" in r = phil]
S ", manager = phil], the bit field of cs is 8 bits, dname," CS ",
Each of the bit fields of manager and phil is set to 3 bits, and each is hashed.

(C) When the value is the attribute value list (set), the element of each attribute value is hashed with 8 bits, and the result of the OR operation of the hash value of each element is assigned to the bit field of the attribute value. . For example, the attribute h in (1) of FIG.
obby = {music, sport} hash value h
(Hobby = {music, sport}) is h
It is the result of the OR operation of (music) and h (sport).

Function of extension term to be searched, attribute name,
For the hash value of the variable existing in the attribute value, a bit string of all bits “1” having the length of each bit field is given.

FIG. 6 shows hash values of the function elements existing in the extension terms (1) to (8). In FIG. 7, the expansion term (1)-
The hash value of the attribute name existing in (8) is shown as the result obtained by dividing into two cases, the attribute name having the unique value (a) and the attribute name having the expansion term (b). FIG. 8 shows the hash values of the attribute values existing in the extended terms (1) to (8) in (a).
The following shows the results obtained by dividing into three cases: an attribute value having a unique value, (b) an attribute value whose value is an extension term, and (c) an attribute value whose value is an attribute value list (set). . In FIG. 9 and FIG. 10, the expansion term (1) to be obtained by using FIG. 6 to FIG.
The index value for (8) is shown.

Next, an index value is created for the extended term that is the inquiry.

As shown in (9) of FIG. 4, the hash value of the functor existing in the inquiry is calculated in 18 bits, and the attribute name and the attribute value are calculated in 8 bits. The result is shown in FIG. However, for the hash value of the variable existing in the function of the extension term which is the inquiry, the attribute name and the attribute value, a bit string of all bits “0” having the length of each bit field is given. Find the index value of the extended term that is a query. The result is shown in FIG.

In the index value S to be searched and the index value Q to be inquired, the term satisfying Q∧S = Q is a term (which is a search candidate) having a possibility of pattern matching. in this case,
As shown in the search process of FIG. 12, as the search candidates,
(1) and (8) are mentioned, and among them, (1) is a term for which pattern matching is possible. For terms that satisfy Q∧S ≠ Q, pattern matching is not attempted because pattern matching cannot be performed, so the search time can be shortened and high-speed search can be performed.

The present invention is not limited to the illustrated embodiment. For example, the index creation processing 10 of FIG. 1 may be executed in another step, or the expansion term shown in FIGS. Various modifications are possible.

[0049]

As described above in detail, according to the present invention, the index value of the set and the index value of the inquiry are created in advance by the index creating process, and the pattern matching is performed after the expansion term determining process is performed. I do. In this way, since the number of extension terms that have the possibility of pattern matching can be narrowed down by using the index, no attempt is made to determine whether or not pattern matching is possible for terms that are not narrowed down. Therefore,
The search time can be shortened by that amount, and high-speed search can be performed.

[Brief description of drawings]

FIG. 1 is a schematic flowchart showing a search method according to an embodiment of the present invention.

FIG. 2 is a diagram illustrating the creation of a conventional index (index value set).

FIG. 3 is a diagram illustrating a conventional search process.

FIG. 4 is a diagram showing search items and extended items of an inquiry according to the present embodiment.

FIG. 5 is a diagram illustrating a configuration of index values according to the present embodiment.

FIG. 6 is a diagram showing a hash value of a function element according to the present embodiment.

FIG. 7 is a diagram illustrating a hash value of an attribute name according to this embodiment.

FIG. 8 is a diagram illustrating a hash value of an attribute value according to this embodiment.

FIG. 9 is a diagram illustrating index values of extension items (1) to (6) of the present embodiment.

FIG. 10 is a diagram illustrating index values of extension items (7) to (8) of the present embodiment.

FIG. 11 is a diagram illustrating an index value of an extension term that is an inquiry in this embodiment.

FIG. 12 is a diagram illustrating a search process according to the present embodiment.

[Explanation of symbols]

10 Index Creation Processing 11 Allocation of Bit Field of Function Element, Attribute Name and Attribute Value 12-1 Calculation of Hash Value of Each Attribute Name 12-2 Calculation of Hash Value of Each Attribute Value 12-3 Calculation of Hash Value of Function Child 14 OR operation 15-1 Set index value 15-2 Query index value 20 Extended term determination process 30 Pattern matching process

[Procedure amendment]

[Submission date] July 8, 1992

[Procedure 3]

[Document name to be amended] Statement

[Correction target item name] 0035

[Correction method] Change

[Correction content]

The contents of each processing 10, 20, 30 in FIG. 1 will be described in detail below with reference to FIGS. 4 to 13.

[Procedure amendment 4]

[Document name to be amended] Statement

[Correction target item name] 0044

[Correction method] Change

[Correction content]

FIG. 6 shows hash values of the function elements existing in the extension terms (1) to (8). In FIG. 7, the expansion term (1)-
The hash value of the attribute name existing in (8) is shown as the result obtained by dividing into two cases, the attribute name having the unique value (a) and the attribute name having the expansion term (b). FIG. 8 shows the hash values of the attribute values existing in the extended terms (1) to (8) in (a).
The following shows the results obtained by dividing into three cases: an attribute value having a unique value, (b) an attribute value whose value is an extension term, and (c) an attribute value whose value is an attribute value list (set). . In FIG. 9 and FIG. 11, the expansion term (1) to be obtained by using FIGS.
The index value for (8) is shown.

[Procedure Amendment 5]

[Document name to be amended] Statement

[Correction target item name] 0046

[Correction method] Change

[Correction content]

As shown in (9) of FIG. 4, the hash value of the functor existing in the inquiry is calculated in 18 bits, and the attribute name and the attribute value are calculated in 8 bits. The result is shown in FIG. However, for the hash value of the variable existing in the function of the extension term which is the inquiry, the attribute name and the attribute value, a bit string of all bits “0” having the length of each bit field is given. Find the index value of the extended term that is a query. The result is shown in FIG.

[Procedure correction 6]

[Document name to be amended] Statement

[Correction target item name] 0047

[Correction method] Change

[Correction content]

In the index value S to be searched and the index value Q to be inquired, the term satisfying Q∧S = Q is a term (which is a search candidate) having a possibility of pattern matching. in this case,
As shown in the search process of FIG. 13, as the search candidates,
(1) and (8) are mentioned, and among them, (1) is a term for which pattern matching is possible. For terms that satisfy Q∧S ≠ Q, pattern matching is not attempted because pattern matching cannot be performed, so the search time can be shortened and high-speed search can be performed.

[Procedure Amendment 7]

[Document name to be amended] Statement

[Correction target item name] 0048

[Correction method] Change

[Correction content]

It should be noted that the present invention is not limited to the illustrated embodiment, and for example, the index creation processing 10 of FIG. 1 is executed in another step, or the expansion term shown in FIGS. Various modifications are possible.

[Procedure Amendment 8]

[Document name to be amended] Statement

[Name of item to be corrected] Brief description of the drawing

[Correction method] Change

[Correction content]

[Brief description of drawings]

FIG. 1 is a schematic flowchart showing a search method according to an embodiment of the present invention.

FIG. 2 is a diagram illustrating the creation of a conventional index (index value set).

FIG. 3 is a diagram illustrating a conventional search process.

FIG. 4 is a diagram showing search items and extended items of an inquiry according to the present embodiment.

FIG. 5 is a diagram illustrating a configuration of index values according to the present embodiment.

FIG. 6 is a diagram showing a hash value of a function element according to the present embodiment.

FIG. 7 is a diagram illustrating a hash value of an attribute name according to this embodiment.

FIG. 8 is a diagram illustrating a hash value of an attribute value according to this embodiment.

FIG. 9 is a diagram illustrating index values of extension items (1) to (3) of the present embodiment.

FIG. 10 is a diagram illustrating index values of extension items (4) to (6) of the present embodiment.

FIG. 11 is a diagram illustrating index values of extension items (7) to (8) of the present embodiment.

FIG. 12 is a diagram illustrating an index value of an extension term which is an inquiry in this embodiment.

FIG. 13 is a diagram illustrating a search process according to this embodiment.

[Explanation of Codes] 10 Index Creation Processing 11 Function Element, Attribute Name and Bit Field Allocation of Attribute Value 12-1 Calculation of Hash Value of Each Attribute Name 12-2 Calculation of Hash Value of Each Attribute Value 12-3 Function Child Hash value calculation 14 OR operation 15-1 Set index value 15-2 Query index value 20 Extended term determination process 30 Pattern matching process

[Procedure Amendment 9]

[Document name to be corrected] Drawing

[Correction target item name] Figure 9

[Correction method] Change

[Correction content]

[Figure 9]

[Procedure Amendment 10]

[Document name to be corrected] Drawing

[Name of item to be corrected] Fig. 10

[Correction method] Change

[Correction content]

[Figure 10]

[Procedure Amendment 11]

[Document name to be corrected] Drawing

[Name of item to be corrected] Figure 11

[Correction method] Change

[Correction content]

FIG. 11

[Procedure Amendment 12]

[Document name to be corrected] Drawing

[Name of item to be corrected] Fig. 12

[Correction method] Change

[Correction content]

[Fig. 12]

[Procedure Amendment 13]

[Document name to be corrected] Drawing

[Name of item to be corrected] Fig. 13

[Correction method] Added

[Correction content]

[Fig. 13]

Claims (1)

[Claims] 1. In the system for expressing data / knowledge by an extension term which is a structure, the extension term includes an arbitrary number of attribute name and attribute value pairs, and the attribute name and the attribute value are included in the portion. A nested structure that describes the extension term can be handled, and as a search condition, a wildcard or variable is given to the attribute name, attribute value, and functor, and the extension term that matches the attribute name, attribute value, and functor In a search method for searching a term set, a bit field of the functor, the attribute name and the attribute value is defined, and the arrangement of the bit field of the attribute name and the attribute value is arranged so as to be covered by the bit field of the functor. If the attribute name or attribute value is an extension term and the bit field is large enough, the bit field of the attribute name or attribute value is The index value is constructed by a bit field of the attribute name or attribute value and a function field of the attribute name or attribute value, and by the index creation process, each attribute name and attribute value pair and functor of the extension term On the other hand, a hash value is obtained, and by the logical sum operation of those hash values, the index value of the configuration for the set to be searched and the index value of the configuration for the inquiry are created in advance, and created by the index creation processing. An extension characterized by performing pattern comparison after performing extension term determination processing for determining the extension term having a possibility of pattern matching based on the logical operation result of the index value of the set and the index value of the inquiry Search method using superposition codes for terms.