JP2010267081A - Information search method, device and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To efficiently perform connection search to an XML data group that is structured data. <P>SOLUTION: An inquiry tree is developed to an execution tree by use of node information, an execution node is selected according to the execution tree, connection is performed about a connection key node group of a connection condition of the execution node, a connection ID is imparted in each connected pair and is stored in a search storage means, node information of a search node group that is a unit of the search is obtained from a node storage means, a corresponding node is matched to the search node group based on a range label of node information of the connection key node group, the matched search node group is obtained together with the connection ID of the connection key node group, the search node group and the connection key node group are connected and are stored in the search storage means, and a result is generated from information stored in the search storage means based on designation of a search query. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to an information search method, apparatus, and program, and in particular, in the field of databases and indexes for storing and searching a large number of XML data as structured data, the processing at the time of executing a combined search is performed efficiently and at high speed. The present invention relates to an information search method, apparatus, and program.

  XML is one of markup languages, and data described in XML is structured and has meaning in the structure. FIG. 13 is a diagram showing XML data and its structure. A mark used to identify an element constituting XML data is called a tag, and a value (text) is described between the tags. Tags can be nested.

  Therefore, XML data can be represented in a tree structure as a model, and each node is called a node. In particular, a value node is called a root element, a value (description content) is called a text node, and an element described in a tag is called an attribute.

  Each node is specified by a “path” in which a tag from the root element is connected by “/” after “/”. For example, if you want to show the title of a book, specify the path as “/ book / title”.

  When an efficient search is to be performed on XML data, it is necessary to be able to retrieve the description contents at an arbitrary place in the structure at high speed. In general, an index (structure index) describing the structure and description contents of XML data is constructed in advance. There are various methods for constructing a structure index. In many cases, an index that associates all the paths of XML data with the description contents specified in the path is constructed. As a technique for realizing efficient search of XML data as described above, a range label is attached to XML data summary information and retained, and at the time of search, the query is converted to a tree structure and the value part is deleted A query tree structure part is generated, a set of intermediate summary trees is generated by collating the summary information with the intermediate query tree, and a set of entity intermediate trees including data obtained by materializing the set of intermediate summary trees with respect to XML data There is a technique that realizes an efficient search of XML data by converting the XML data into a filter process using the portion of the value corresponding to the set of trees in the entity intermediate (see, for example, Patent Documents). 1).

JP 2006-228155 A

As a search pattern, there is a combination of two types (or the same type) of XML data groups. For example, when there are two types of XML groups “class information XML” and “student XML” as shown in FIG. 14, the “class information XML” indicating the ID of the student is displayed.
"/ course / students / student_id / text ()"("text()" indicates a text node)
And “Student XML”

をキーに２つの情報を結合させ、図１５に示すような受講者のプロファイル情報を加えたクラス情報を取得したといった検索が考えられる。

A search may be considered in which two pieces of information are combined using the key and the class information including the student's profile information as shown in FIG. 15 is acquired.

  In this example, the combined search query can be described as shown in FIG. 16 in XQuery, which is one XML query language.

  As a simple search processing method, there is a method in which child nodes are sequentially traced from the root element of each XML data to find a node group that matches the search condition. However, in the case of large data or a large number of data groups, the path is When it points to a deep place, scanning takes time. Therefore, as shown in the above-mentioned Patent Document 1, only a node group necessary for processing is directly taken out, and an efficient search is performed by determining a parent-child and ancestor-descendant relationship.

However, this method is not mentioned even in the join search, and an efficient method has not been proposed. Therefore, after performing efficient search as shown in Patent Document 1 for both XML data groups to be combined search, a method of matching the combination processing one by one with a simple method can be considered. Join condition ("$ a / students / student_id / text () = $ b / @ id")
If there are no other conditions, there are cases where the number of intermediate results is too large to improve efficiency.

  In addition, regarding the joining process, an efficient technique is realized in a relational database (RDB), but the joining of XML data differs in the following points, and thus cannot be applied as it is.

In the case of RDB, the unit (tuple) of the bond and the key serving as the join condition have a one-to-one relationship (see FIG. 17), but in the case of XML, the key has a one-to-N relationship. For example, in the case of the above example, the unit of combination of “class information XML” is
"/ course"
The following subtree, that is, the entire document, but the key that becomes the join condition
"/ course / students / students_id"
There are several in the document.

  In other words, a combination of “true” is required if any pair of N keys matches.

  The present invention has been made in view of the above points, and an object of the present invention is to provide an information search method, apparatus, and program capable of efficiently performing a combined search.

  FIG. 1 is a diagram for explaining the principle of the present invention.

The present invention (Claim 1) is an information search method in an apparatus for storing and searching XML data groups that are structured data,
Based on the search path and the search keyword specified by the input search query, the path index storage means for storing the XML structure in units of paths and the parent node for maintaining the relationship between each node of the XML are children. A query tree is generated by referring to node storage means storing node information whose node ID is a range label represented by a node ID including a node (step 1). Then, an execution tree including nodes of each processing type is generated (step 2), a process for acquiring a node group, a process for determining their ancestor-descendant relationship, a node group combining process, and a result generating process for generating a final combined result (step 2). ) Execution tree generation step;
If there is a child node in each execution node of the execution tree, the result of the child node is acquired, the process according to the processing type of the node is performed (step 3), and the final result is output (step 4). And a search step.

Further, the present invention (Claim 2), in the search step,
When an execution node of an execution tree has a join condition, the nodes specified in the join condition (join key node group) are joined, and a join ID is assigned to each joined pair and stored in the search storage means. And
Obtain node information of a node group (search node group) that is a unit of search from the node storage means, and match the corresponding one based on the range label of the node information of the combined key node group to the search node group, A matching search node group is acquired together with a combination ID of the combination key node group, the search node group and the combination key node group are combined, and stored in a search storage unit;
Based on the specification of the search query, a result is generated from the information stored in the search storage means.

  FIG. 2 is a principle configuration diagram of the present invention.

The present invention (Claim 3) is an information search device in an apparatus for storing and searching XML data groups that are structured data.
Path index storage means 1 storing the XML structure in units of paths;
A node storage unit 2 for storing node information in which a node ID is a range label represented by a node ID in which a parent node includes a child node, in order to maintain a relationship between XML nodes;
Based on the search path and the search keyword specified by the input search query, a query tree is generated with reference to the path index storage means and the node storage means, and a node is obtained for the path to be processed by the query tree. Execution tree generation means 41 for generating an execution tree including nodes of each processing type of a process for acquiring a group, a determination process for their ancestor-descendant relationship, a node group combination process, and a result generation process for generating a final combination result When,
Each execution node of the execution tree has a search execution means 42 that acquires the result of the child node if there is a child node, performs processing according to the processing type of the node, and outputs the final result .

The search execution means 42 of the present invention (Claim 4)
When an execution node of an execution tree has a join condition, the nodes specified in the join condition (join key node group) are joined, and a join ID is assigned to each joined pair and stored in the search storage means. Means to
Node information of a node group (search node group) that is a unit of search is acquired from the node storage means 2, and the corresponding one is matched to the search node group based on the range label of the node information of the join key node group Obtaining a matching search node group together with a combination ID of the combination key node group, combining the search node group and the combination key node group, and storing in the search storage unit;
Means for generating a result from information stored in the search storage means based on the designation of the search query.

  The present invention (Claim 5) is an information search program for causing a computer to function as each means constituting the information search apparatus according to claim 3 or 4.

  As described above, according to the present invention, it is possible to efficiently perform a combined search by directly handling only the minimum necessary node group without performing an intermediate process for sequentially tracing the hierarchical structure of XML.

It is a figure for demonstrating the principle of this invention. It is a principle block diagram of this invention. It is a block diagram of the information search device in one embodiment of the present invention. It is the example of a coupling | bonding of XML in one embodiment of this invention. It is an example of XQuery in one embodiment of this invention. It is an example of storing path information and node information of node management in an embodiment of the present invention. It is an example of the range label of the node management part in one embodiment of this invention. It is a flowchart of the search process in one embodiment of this invention. It is a flowchart of the node acquisition process in one embodiment of this invention. It is an example of the inquiry tree produced | generated by the search part in one embodiment of this invention. It is an example of the execution tree produced | generated by the search part in one embodiment of this invention. It is an intermediate result in the example of execution in one embodiment of the present invention. It is an example of XML data and its structure. It is an example of two XML data. This is an example of combining XML data. This is an XQuery example of join. This is an example of bonding in RDB.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

  The present invention is realized in RDB based on a technique (for example, Patent Document 1) that efficiently performs a join search by directly handling only the minimum necessary node group without sequentially following the hierarchical structure of XML. By combining the combination processing, the efficient combination processing for the XML data group is realized.

  FIG. 3 shows the configuration of the information retrieval apparatus in one embodiment of the present invention.

  The information retrieval apparatus shown in FIG. 1 includes a path index unit 1 that holds the XML structure in units of paths, a node management unit 2 that grasps the relationship in the XML structure between each node of the XML, and constructs an index of the value of each node. It has a value index unit 3 to be managed and a search unit 4 that executes a (join) search. The path index unit 1 and the node management unit 2 correspond to the structure index in the above-mentioned background art section.

  The path index unit 1 constructs an index that associates all the paths of XML data with nodes specified in the path, and stores the index in an internal storage unit.

  The node management unit 2 holds the structural relationship of XML data between nodes in an internal storage unit. In general, in order to manage the structural relationship, an ID is assigned to each node of the XML data, and the relationship between nodes (such as an ancestor-descendant relationship) is held as a range label.

  The value index unit 3 constructs an index (B-Tree, hash, etc.) for the text node and attribute value, and holds it in the internal storage means. This is unnecessary if only searching, but is necessary to achieve high-speed searching.

  The processing of the search unit 4 is performed according to the following procedure. Hereinafter, as a term, a node (group) serving as a search unit is referred to as a “search node (group)”, and a node (group) serving as a join condition is referred to as a “join key node (group)”.

In the example of the query in FIG. 16 described above, “/ course” and “/ profile” of all the data set in “$ a” and “$ b” are the respective search nodes, and the conditions in the where clause Has been
"/ course / students / student_id / text ()"
as well as,

が結合キーノード群となる。

Is a group of combined key nodes.

  (1) For both combined key node groups, the combined process is performed in units of the node groups (hereinafter, this process is referred to as “key combined” process).

  This can be realized by a processing method equivalent to RDB. At this time, a binding ID (both or one of the node IDs) that can identify the matched pair is assigned and held in a memory (not shown) in the search unit 4.

  (2) Acquire both search node groups. (At that time, it is more efficient to obtain only the XML data including the combination key node group obtained in (1)).

  (3) For each XML data group, the search node group obtained in (2) is matched with the corresponding one from the join key node group obtained in (1) (using parent-child and ancestor-descendant relationship determination by node ID). , A matching search node group is acquired together with a binding ID (group).

  (4) The correspondence between both search node groups is determined by the combination ID (group), and the connection is performed. However, in this case, the result unit (how to summarize) depends on the query specification.

  (5) Using the search node as a base, obtain the node necessary for the result (specified in the return clause), generate the result, and return it.

  In other words, candidates can be narrowed down by performing the same process as the RDB join process at the join key node level for the join condition, and by further saving the join ID, the final join process between search nodes Can also improve efficiency.

  Supplementing the above (4), in the case of the query of FIG. 16, since “$ a” is specified by the “for” statement, the results are collected in units of the search node “$ a” (FIG. 15). On the other hand, as shown in Fig. 4, it is also possible to obtain the result in units of "$ b". In that case, the XQuery specifies the "$ b" in the "for" statement as shown in Fig. Results are collected in units of search node "$ b" which is "for".

  Further, when there are a plurality of binding conditions, it can be dealt with by holding a binding ID (group) for each binding condition and determining the combination in (4).

  In addition, if there are other narrowing conditions ("$ a / meta / grade = 'Beginning class'") as shown in the query in FIG. 5, narrowing down at appropriate timings before and after (1) to (3). (The appropriate order varies depending on the narrowing-down condition and the target data group).

  Hereinafter, an example based on the technique of Patent Document 1 will be described as one form of the realizing method.

  Each component shown in FIG. 3 will be described in detail.

  The path index unit 1 is a module that manages path information by an internal storage unit, and includes the maintenance of the relationship between the path and each data corresponding to the path to the node. A storage example of path information is shown in FIG. Holds the path and the ID for that path.

  The node management unit 2 stores and manages information of each node in an internal storage unit. As shown in FIG. 6B, the node information includes a path ID corresponding to the path ID of the path index unit 1, an ID (document ID) for each data, a node ID (range label), a tag, a value, and the like. Composed. The range label is represented by a binary value for each node as shown in FIG. 7, and is given so that the parent node includes a child node. This makes it possible to determine the parent-child and ancestor-descendant relationship between arbitrary nodes. In this example, only the document IDs 101 and 201 are shown. However, in this case, since the XML groups have the same structure, the same range labels are assigned to other XMLs (general). (In other words, the same range label is not always the same for the same path).

  The value index unit 3 constructs an index (B-Tree, hash, etc.) for the text node and attribute value, and stores and manages it in the internal storage means. At least the path and its value are limited, and the search is possible. For example, one index that covers all paths and values as keys may be created, or a separate index (value is a key) may be created for each path.

  The search unit 4 receives the search query, specifies a node that matches the specified search path and search keyword from the storage means of the path index unit 1, the node index unit 2, and the value index unit 3, and if necessary, A return result is acquired from a data management unit (not shown) such as a database and returned.

  The combined search process in the search unit 4 is shown below.

  FIG. 8 is a flowchart of join search processing in one embodiment of the present invention.

  Step 110) The search unit 4 generates a query tree from the acquired search query.

  Step 120) Next, using the node information of the node manager 2, an execution tree for the query tree is created.

  Step 130) According to the execution tree, an execution node such as node type “node acquisition”, “relationship determination”, “join”, “result generation”, etc. is selected.

  Step 140) If the execution node type is “node acquisition”, the process proceeds to step 150; if it is “relationship determination”, the process proceeds to step 160; if it is “join”, the process proceeds to step 107; In the case of “generation”, the process proceeds to step 180.

  Step 150) When the execution node type is “node acquisition”, node information is acquired from the path index unit 1 and the node management unit 2, and the process proceeds to step 190.

  Step 160) When the execution node type is “relationship determination”, the node information is acquired from the node management unit 2, and the parent-child / ancestor-descendant relationship is determined according to the range label (document ID matching, structure join by range label). The process proceeds to step 190.

  Step 170) When the execution node type is “join”, the node information is acquired from the node management unit 2, the value index is also used as necessary, and the join ID is given. Further, the same combination IDs are combined and the process proceeds to step 190.

  Step 180) If the execution node type is “result generation”, a result to be returned is generated, and the process proceeds to step 190.

  Step 190) The processing after Step 130 is repeated for all execution nodes.

Next, step 150 will be described in detail. FIG. 9 is a flowchart of the node acquisition process according to the embodiment of the present invention.

  Step 151) When the execution node type is “node acquisition”, it is determined whether or not there is a join condition. If so, the process proceeds to step 152. If not, the process proceeds to step 156.

  Step 152) If the key combination has been executed, the process goes to Step 155, and if not yet executed, the process goes to Step 153.

  Step 153) The corresponding combined key node information is acquired from the path index unit 1 and the node management unit 2.

  Step 154) The value index unit 3 is inquired based on the combined key node information acquired in step 153, a key ID corresponding to the combined key node information is acquired, and key combination processing is performed.

  Step 155) The node information is acquired together with the combination ID from the key combination result, and the process is terminated.

  Step 156) If there is no join condition, the corresponding node information (group) is acquired from the path index unit 1 and the node management unit 2.

  Step 157) The value index unit 1 is inquired, only those satisfying the value constraints are narrowed down, and the process is terminated.

  In the following, as a specific example of the processing of FIGS. 8 and 9, when the data group of FIG. 14 is labeled in FIG. 7 and stored in the memory (not shown) of the search unit 4, FIG. 5 shows a procedure for executing the query 5.

  The search unit 4 generates a query tree. The query tree created here is as shown in FIG. In the figure, the double line part ("course" "profile") represents the path to be processed, and the path indicated by the dashed arrow indicates that the value is constrained .

  Next, the search unit 4 creates an execution tree as shown in FIG. For a path to be processed of the query tree, a process for acquiring a node group (node acquisition) is set as a leaf, and the determination of the ancestor-descendant relationship is set as a clause. In node acquisition, a value constraint may be given, and a joint condition is also commonly given to both leaves as a kind of constraint. At the root of the execution tree, there is a result generation process that generates the final combined result. In the query example, there are two search node groups set to "$ a" and "$ b" in the query example. To do.

  The tree shape represents the execution order. Since there are a plurality of choices, it is desirable to perform some kind of execution plan optimization process, but since it is out of scope in the present invention, its description is omitted. Select an execution node and proceed from the leaves of the execution tree toward the root.

  When the node ID is expressed in the format of [document ID, (range label)], it is processed in the following procedure.

(1) P5: "/ course / meta / grade / text () Constraint: =" Beginner class "node acquisition Since there is no join condition, the node corresponding to the path from the path index part 1 and the node management part 2 Information is acquired, and the value index unit 3 narrows down to only nodes that satisfy the constraint. Those having the value of “beginning class” in the corresponding path become nodes of the documents 101 and 102, so that the following is obtained.

'Beginner class' [101, (6,7)], 'Beginner class' [102, (6,7)]
(2) Node acquisition of P6: “/ course” Since there is no join condition, node information corresponding to the path is acquired from the path index unit 1 and the node management unit 2 (because there is no constraint, the value index unit Since the same path exists in all documents), the following is obtained.

[101, (1,36)], [102, (1,36)], [103, (1,36)] *
However, the node [103, (1, 36)] * can be removed at this stage using the document ID from the result of (1). In the following, “*” is given to nodes that can be similarly removed using the document ID of the previous processing result.

(3) P4: Determination of ancestor-descendant relationship between P5 and P6 In this example, P6 is an ancestor and P5 is a descendant node in the form of leaving the ancestor P6 node.

[101, (1, 36)] (← offspring 101, [6, 7]), [102, (1, 36)] (← offspring 102, [6, 7])
(4) P7: “/ course / students / student_id / text () constraint: key join” node acquisition Since it has a join condition and the key join has not yet been executed, the key join is executed first. The node information of both P7 and P9 is acquired from the path index unit 1 and the node management unit 2, and the value index unit 3 is also used as necessary to perform the combination and also give the combination ID. As a result of key combination, the following is obtained.

本例では、結合ＩＤとしてＰ７の［文書ＩＤ，範囲ラベルの第１値］を利用することとしている（こうすることで、後述の（９）．Ｐ２の結合処理時に効率的な処理が期待できる）。

In this example, [document ID, first value of range label] of P7 is used as the combination ID (this makes it possible to expect an efficient process at the time of (9) .P2 combination process described later. ).

  Next, the node ID and the combination ID of P7 are extracted from the obtained key combination result, and the node acquisition result is obtained.

（５）Ｐ３：Ｐ４とＰ７の先祖子孫関係判定
先祖側のＰ４のノードを残す形で、Ｐ４が先祖でＰ７が子孫の関係のノードを取得する。その際、結合ＩＤも検索部４の内部のメモリ（図示せず）に保持しておく。

(5) P3: Determination of ancestor-descendant relationship between P4 and P7 In the form of leaving the ancestor P4 node, P4 is an ancestor and P7 is a descendant. At that time, the combination ID is also stored in a memory (not shown) inside the search unit 4.

（６）Ｐ９："/profile/@id制約：キー結合"のノード取得
結合条件を持っており、キー結合結果も（１）で得られているため、キー結合結果から、Ｐ９のノードＩＤと結合ＩＤを取り出し、ノード取得結果を得る。

(6) Node acquisition of P9: “/ profile / @ id constraint: key join” Since it has a join condition and the key join result is also obtained in (1), the node ID of P9 is obtained from the key join result. The binding ID is taken out and the node acquisition result is obtained.

（７）Ｐ１０："/profile"のノード取得
結合条件は持っていないため、パスインデクス部１、ノード管理部２から、パスに該当するノード情報を取得する（制約を持たないため、値インデクス部３により絞り込んでも同じ）全ての文書に該当パスを持つノードが存在するため、以下が得られる。

(7) P10: Node acquisition of “/ profile” Since there is no join condition, node information corresponding to the path is acquired from the path index unit 1 and the node management unit 2 (there is no constraint, so the value index unit Since the same path exists in all documents), the following is obtained.

[201, (1,2)], [202, (1,12)], [203, (1,12)], [204, (1,12)] *, [205, (1,12)] *, [206, (1,12)] *,
[207, (, 12)]
(8) P8: Determination of ancestor-descendant relationship between P9 and P10 In the form of leaving the ancestor-side P10 node, the node of the path index unit 1 and node management unit 2 acquires a node in which P10 is an ancestor and P9 is a descendant. At that time, the binding ID is also held.

（９）Ｐ２：Ｐ３とＰ８の結合
同じ結合ＩＤを含むもの同士で結合を行うが、本例では、"$b"にあたるＰ８側の単位で結合する。

(9) P2: Binding of P3 and P8 Bindings that include the same binding ID are performed with each other, but in this example, they are combined in units of the P8 side corresponding to “$ b”.

参考までに、個々までの処理の中間結果を記したものを図１２に示す。

For reference, FIG. 12 shows intermediate results of individual processes.

(10) P1: Result generation When a return result is created according to the return phrase in the last P1, it is as shown in FIG. Acquisition of the node to be returned can also be performed by a method similar to the above procedure, but since it is out of scope in the present invention, description thereof is omitted.

  It is possible to construct the operation of the search unit 4 shown in FIG. 3 as a program, install it on a computer used as an information search device and execute it, or distribute it via a network.

  Further, the constructed program can be stored in a portable storage medium such as a hard disk, a flexible disk, or a CD-ROM, and can be installed or distributed in a computer.

  The present invention is not limited to the above-described embodiments and examples, and various modifications and applications can be made within the scope of the claims.

  The present invention can be applied to a technique for accumulating and retrieving a large number of XML data as structured data.

1 path index storage means, path index section 2 node storage means, node management section 3 value index storage means, value index section 4 search section 41 execution tree generation means 42 search execution means 45 memory

Claims (5)

An information retrieval method in an apparatus for storing and retrieving XML data that is structured data,
Based on the search path and the search keyword specified by the input search query, the path index storage means for storing the XML structure in units of paths and the parent node for maintaining the relationship between each node of the XML are children. A query storage is generated by referring to a node storage unit that stores node information whose node ID is a range label represented by a node ID including a node, and a node group is determined for a path to be processed by the query tree. An execution tree generation step for generating an execution tree including nodes of each processing type of a process to acquire, a determination process of their ancestor-descendant relationship, a node group combination process, and a result generation process for generating a final combination result;
In each execution node of the execution tree, if there is a child node, a result of the child node is obtained, processing according to the processing type of the node is performed, and a final result is output;
An information retrieval method characterized by: The search step includes
When the execution node of the execution tree has a join condition, the join group is joined with respect to the node group (join key node group) designated by the join condition, and a join ID is assigned to each joined pair, and the search storage means Store and
Node information of a node group (search node group) as a unit of search is acquired from the node storage means, and the corresponding one based on the range label of the node information of the combined key node group is obtained for the search node group A matching search node group is obtained together with a combination ID of the combination key node group, the search node group and the combination key node group are combined, and stored in a search storage unit;
The information search method according to claim 1, wherein a result is generated from information stored in the search storage unit based on designation of the search query. An information search apparatus in an apparatus for storing and searching XML data group that is structured data,
A path index storage means for storing the XML structure in units of paths;
Node storage means for storing node information whose node ID is a range label represented by a node ID whose parent node includes a child node, in order to maintain a relationship between nodes of XML;
Based on the search path and the search keyword specified by the input search query, the query index is generated by referring to the path index storage means and the node storage means, and the query tree is processed. An execution tree generation that generates an execution tree including nodes of each processing type of a node group acquisition process, a determination process of their ancestor-descendant relationship, a node group combination process, and a result generation process that generates a final combination result Means,
In each execution node of the execution tree, if there is a child node, a result of the child node is obtained, processing according to the processing type of the node is performed, and a search execution unit that outputs a final result;
An information retrieval apparatus comprising: The search execution means includes
When the execution node of the execution tree has a join condition, the join group is joined with respect to the node group (join key node group) designated by the join condition, and a join ID is assigned to each joined pair, and the search storage means Means for storing;
Node information of a node group (search node group) as a unit of search is acquired from the node storage means, and the corresponding one based on the range label of the node information of the combined key node group is obtained for the search node group Means for matching, obtaining a matching search node group together with a combination ID of the combination key node group, combining the search node group and the combination key node group, and storing in the search storage unit;
Means for generating a result from information stored in the search storage means based on the designation of the search query;
The information search device according to claim 3, comprising:   An information search program for causing a computer to function as each means constituting the information search device according to claim 3 or 4.

Images