JPH07225771A - Retrieval expression preparation support system - Google Patents

Abstract

PURPOSE:To provide a retrieval expression preparation support system which is capable of easily obtaining a right conditional expression for retrieval. CONSTITUTION:A structured document 9 to be used for the illustration of a retrieval result is analyzed in a structure analysis part 1 and the structure is extended in a memory 7. A retrieval expression structure extraction part 2 successively takes out the routine structure patterns stored in a knowledge base 6 and finds out the partial structure pattern including the partial structure of the structured document illustrated by a user. A retrieval condition extraction part 3 fixes the value of the field to be used for a retrieval expression for each of the record included in the partial structure extracted in the retrieval expression structure extraction part 2. A retrieval expression synthesis part 4 synthesizes the expression of a form which is possible to be recognized as the retrieval expression. The synthesized retrieval expression is possible to be deformed in accordance with the edition instruction of the user and under such a condition that the illustrated partial structure is possible to be retrieved in a retrieval expression edition part 5.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a structured document processing apparatus for processing a structured document, and when a structured document is subjected to a retrieval process, a retrieval formula preparation for supporting a retrieval formula used for retrieval. It concerns support systems.

[0002]

2. Description of the Related Art Some conventional structured document processing apparatuses are capable of performing a search by giving a context containing a context in a structured document, that is, a pattern including information about a hierarchical structure as a search condition. Also, an apparatus that supports the creation of such a search formula has been considered. By using this device, the user can input and change the search condition by inputting the conditional expression in the form of a graph and editing it. However, in the conventional structured document processing device, the user himself has to consider the search condition and input it from the beginning. To do this, the user needs to know exactly what the internal representation of the structured document is.

A conventional device for assisting the creation of a search expression by means of a graph expression expresses a format recognizable by an input device in an easy-to-understand manner, and creates a grammatically correct expression even for a user who has no knowledge of the correct grammar. It enables you to do it. However, even such a device still requires the user to have accurate knowledge about the internal representation of the structured document.

Further, the conventional structured document processing device does not support anything in creating an expression having a correct meaning. Therefore, although search is a very useful function, it is not a function that anyone can easily use.

[0005]

SUMMARY OF THE INVENTION The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a search expression creation support system capable of easily obtaining a correct conditional expression for a search. To do.

[0006]

According to a first aspect of the present invention, there is provided a search formula creation support system for supporting creation of a search formula used when a search process is performed on a structured document, Storage means for storing the structured document; structure extraction means for extracting a partial structure including the exemplified partial structure based on the partial structure of the structured document exemplified by the user as one of the search results; The present invention is characterized by having a search formula synthesizing unit that synthesizes a search formula from the partial structure extracted by the structure extracting unit.

Further, in the invention described in claim 2,
When the search formula synthesized by the search formula synthesis means is modified in accordance with an editing operation instruction from the user, the modified search formula is used as an example of the structured document used in the synthesis. The present invention is characterized by having a search formula editing means for executing a modification process of a search formula on condition that the structure is included.

According to the third aspect of the present invention, in a search formula creation support system for supporting the creation of a search formula used when performing a search process on a structured document, search means for performing a search according to the search formula, An input means by which the user points out an error in the search result based on the search result by the search means, and a search formula based on the structure of the partial structure including the part pointed out according to the error point input by the input means and the structure of the search expression It is characterized by having a retrieval formula correction processing means for correcting.

According to a fourth aspect of the invention, in the retrieval formula creation support system according to the third aspect, information about each branch included in the structure of the retrieval formula in which an error is pointed out and a combination of the branches is provided. And a search formula correction processing unit refers to the database when correcting the search formula, and corrects a correction formula other than a correction candidate of the search formula including a combination of branches stored in the database. It is characterized in that the search formula is modified based on the candidates.

[0010]

According to the first aspect of the present invention, the user is made to exemplify one example of the search result, and the partial structure including the partial structure of the structured document obtained by the example is extracted by the structure extracting means, By automatically synthesizing the search expression from the extracted structure by the search expression synthesizing means, a grammatically correct search expression can be obtained by a simple operation of exemplifying.

In addition, there are cases where it is desired to strengthen or loosen the conditions of the synthesized search formula to perform the search. Therefore, according to the invention described in claim 2, it is possible to modify the search formula synthesized by the search formula editing means. However, free modification may create an unreasonable search expression that does not include the exemplified search results. The search expression editing unit executes the modification processing of the search expression on condition that the modified search expression includes the partial structure used in the composition. As a result, the user can always obtain a correct search formula that can always search the partial structure illustrated at the time of synthesis.

According to the third aspect of the present invention, when an unintended search result is obtained, the user can determine where the erroneous part of the search result is, not the search expression side. By exemplifying from the input means, the search expression correction processing means automatically corrects the search expression so as to correct the indicated error based on the partial structure including the illustrated portion and the structure of the search expression. The search formula correction processing means sequentially corrects the search formula even if the information of the error pointed out is incomplete, that is, not all errors are pointed out but only a part of the erroneous part is illustrated. You can go. At this time, as the number of indications regarding errors increases, it is possible to correct so that a more correct search formula that solves all the errors can be obtained.

According to the fourth aspect of the present invention, the information on the combination of the branches included in the structure of the search expression in which the error is pointed out is sequentially stored in the database, and the search expression correction processing means is used by the search expression. When correcting, the information stored in the database, that is, the search expression in which the error is pointed out, can be prevented from being corrected. As a result, a kind of learning effect of learning the past history can be obtained, and the retrieval formula can be corrected without repeating the same error known up to that point.

[0014]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 is a block diagram of the first embodiment of the search expression creation support system of the present invention. In the figure, 1 is a structural analysis unit,
Reference numeral 2 is a retrieval formula structure extraction unit, 3 is a retrieval condition extraction unit, 4 is a retrieval formula synthesis unit, 5 is a retrieval formula editing unit, 6 is a knowledge base, 7 is a memory, 8 is a retrieval formula, and 9 is a structured document. .

The structure analysis unit 1 analyzes the structured document 9 used for exemplifying the search result and stores the structure in the memory 7.
Expand to. At this time, the hierarchical structure of the structured document 9 used as an example is stored in the memory 7 as a tree structure of records having pointers bidirectionally between those having a parent-child relationship and those having a sibling relationship. The search expression structure extraction unit 2 sequentially takes out the typical structure patterns accumulated in the knowledge base 6 and finds one that matches the partial structure including the partial structure of the structured document exemplified by the user. And
A mark is given to all the records that match the found typical structure pattern. Search condition extraction unit 3
Determines the value of the field used for the search expression for each of the records included in the partial structure extracted by the search expression structure extraction unit 2. The search formula synthesis unit 4 sequentially takes out the records of the partial structures to which the marks have been added, and synthesizes the expression in a form recognizable as the search formula. Search formula editor 5
In accordance with a user's editing instruction, the search formula combining unit 4 modifies the search formula combined to generate a new search formula. At this time, the search formula is modified only when the correctness that the partial structure designated by that time can be searched is not lost. In the knowledge base 6, a context (partial structure) used by a skilled user for searching is stylized and stored.

FIG. 2 is a flow chart showing the operation of search expression generation in the first embodiment of the search expression creation support system of the present invention. In S11, the user
A structured document editor or the like is used to exemplify the portion to be searched. For example, for example, a record displayed in the structured document editor may be input by pointing with a pointing device or the like. With this example, the identifier of the designated record can be obtained. A record having the same identifier is found from the structured document for the purpose of illustration developed in the memory 7, and the record of the structured document in the memory 7 is marked.

Next, in S12 and S13, the partial structure exemplified by the user, that is, the partial structure including the marked record of the structured document in the memory 7 is extracted using the knowledge base 6. . The knowledge base 6 stores a context used by a skilled user for a search as a substructure, and a part that a skilled user will use as a search condition for searching a substructure exemplified by the user. The structure is accumulated. Therefore, even with an instruction from an inexperienced user, a search condition similar to that used by a skilled user can be obtained. The extracted partial structure is a partial structure having the same shape as one of the structures stored in the knowledge base 6 including the partial structure exemplified by the user. Matches the extracted canonical substructure pattern,
A mark is given to all the records of the structured document in the memory 7. The marked record becomes a partial structure (partial tree) of the structure of the document developed in the memory 7. This partial structure is the structure that the search formula should have.

In step S14, the conditions of the retrieval formula corresponding to the extracted partial structure are obtained. The structure that the search expression should have is determined in the steps up to S13, but the attributes of each record, for example, conditions such as figures, titles, paragraphs, etc. are also required in the search. Therefore, the extracted partial structure of the structured document in the memory 7, that is, the attribute of the marked record is referred to and the search condition is obtained.

In step S15, the search expression combining unit 4 generates a search expression based on the partial structure and conditions for the search, and in step S16, the generated search expression is output. The user may perform the search process using the output search formula.

In the above-mentioned configuration, the rule stored in the knowledge base 6 is used when the standard structure pattern is extracted by the search expression structure extraction unit 2, but it is stored in the knowledge base 6. If there is no rule that contains the partial structure exemplified by the user, set the corresponding partial structure by the default rule, or by inference,
It is also possible to provide means for extracting a similar structure and presenting it to the user. Alternatively, the knowledge base 6 may be omitted, and the procedure for returning a partial structure corresponding to the partial structure exemplified by the user may be incorporated in the program.

FIG. 3 is an explanatory view of an example of a structured document, FIG.
FIG. 3 is an explanatory diagram of an example of a structured internal representation of a structured document. The document shown in FIG. 3 includes titles such as “Introduction” and “Main Thesis”, as well as respective texts and figures.
Such a document can be expressed as a tree structure as shown in FIG. That is, the document of FIG. 3 has two “sections” that indicate the entire introduction and the entire main body. For example,
The "section" indicating the introduction includes a "title" and a "text paragraph", and the content of the "title" includes a character string "introduction". In addition, for example, the figure is included in a “section” indicating the whole of the present theory, and is further included as a “figure area” in a “text paragraph” in the “section”. The figure is composed of the image part of the figure and the title of the figure. The same applies to the other parts. As described above, in the structured document, the document structure is represented by a tree structure.

FIG. 5 is an explanatory diagram of an example of the structure of the structured document stored in the memory 7. The structure analysis unit 1 analyzes the structure of the structured document 9 for illustration, and for example, the structured document having the structure shown in FIG. 4 is stored in the memory 7 in the data structure shown in FIG. It That is,
The hierarchical structure of the structured document is stored as a tree structure having pointers bidirectionally between those having a parent-child relationship and those having a sibling relationship.

A portion of the structured document tree structure that serves as a node is represented by a record. In FIG. 5, each rectangle is a record. Each record has several fields. In Figure 5, the record ID
"No.", "Attribute" that indicates the type of record, "Value" that indicates the contents that actually appear in the document such as text data, and "Sub-expression" that stores the information used for the search expression. Shows. Of course, other fields may be provided, and for example, the field "subexpression" may be managed by a record of a different type from this record.

The part or structure having the character string "main theme" in FIG. 4 corresponds to the record having "1.1.1" as the field "No." in FIG. This record has "text" as a field "attribute" and a character string "main text" as a field "value". The field “No.” is configured to represent a hierarchical structure. When the value is “1.1.1”, the child record “1.1” of the record having the value “1”. It is shown that it is a child of. As mentioned above, each record also has pointers to parents, children and siblings.

FIG. 6 is an explanatory diagram of an example of a typical structural pattern stored in the knowledge base 6. The knowledge base 6 stores various typical partial structure patterns that a skilled user uses for searching. For example, in pattern 1 of FIG. 6, a record whose field “attribute” value is “section”, a record whose field “attribute” value is “title”, and a field “attribute” value is “text paragraph” 2 shows the partial structure in which any partial structure may be connected as a child of the “title” record.
Similarly for pattern 2, the value of the field "attribute" is
A "body paragraph" record exists as a child of a "section" record, a "chart area" record exists as a child, an attribute whose attribute is undetermined, and another "area title" Record exists,
The record of “area title” indicates a partial structure having a child record. A method of deciding the attribute of the above-mentioned undecided attribute will be described later, but this means that any attribute may be used as a knowledge base rule. However, this attribute is set to be equal to the attribute of the concretely exemplified structure before the search expression is synthesized, and the attribute is used as a condition at the time of search.

The specific operation of the first embodiment of the retrieval formula creation support system of the present invention will be described with reference to FIGS. For example, the structured document 9 for illustration as shown in FIG. 4 is expanded by the structure analysis unit 1 into the structure as shown in FIG. 5 and stored in the memory 7, and the structured document 9 as shown in FIG. It is assumed that the pattern is stored in the knowledge base 6.

Here, as an example of the search result of the structured document shown in FIG. 4, consider a case where the user has designated "area title". The search expression structure extraction unit 2 refers to the contents of the knowledge base 6 shown in FIG. 6 and extracts the pattern 2 shown in FIG. 6 as the partial structure pattern including the designated “area title”. At this time, a plurality of partial structure patterns may be extracted. In that case, a partial structure pattern including the plurality of extracted partial structure patterns may be generated,
Alternatively, it can be configured to select one according to some criteria.

When pattern 2 is extracted, pattern 2
A mark is added to the record that matches with. That is, in the example shown in FIG. 5, a mark is added to the record indicated by the thick line. The marked record is a partial structure of the structure developed in the memory 7, and is the structure that the search expression should have.

In the extracted pattern 2, "attribute =?"
Or, there is an undetermined record such as "?". The search condition extraction unit 3 extracts the necessary information from the field storing the content of the corresponding record in order to determine the applicable standard partial structure pattern found from the knowledge base 6, and the field "subexpression" To store. Figure 6
In the pattern 2 of “attribute =?”, The field “N
o. "" Corresponds to the record having "1.2.2.1". From this, it can be understood that the "attribute" portion should be used as a condition in addition to the structure of the pattern 2 stored in the knowledge base 6. Therefore, the content of the field "attribute" of the record in which the field "No." is "1.2.2.1" is set in the field "subexpression" as "attribute = figure".

Although not present in this example, if there is a record that is not marked between the marks that are marked, from the number of records and the partial structure pattern of the knowledge base related thereto, The degree of freedom regarding the depth of the structure of the parent-child relationship and the degree of freedom regarding the number of other structures that may exist during the sibling relationship are determined and also stored as part of the field "subexpression".

Finally, the retrieval formula synthesizing unit 4 sequentially retrieves the records of the partial structure marked with the bold line in FIG. 5, and the retrieval formula is expressed in a format that can be recognized by the device that conducts the retrieval as the retrieval formula. To synthesize. FIG. 7 is an explanatory diagram of an example of the generated search formula. As shown in FIG. 7, although the search expression itself is an incomprehensible expression, by using the search expression creation support system of the present invention, the search expression as shown in FIG. Is automatically generated. The generated search formula may be any expression that can be understood by the device that actually performs the search. The user conducts a search using a device that actually performs the search using the synthesized search formula.

As shown concretely with reference to FIG. 7, a search expression in a structured document is generally expressed in a tree structure format. 1. The condition of the attribute of the node to be searched, and 2. The condition of the attribute of the node that should appear around the search target, and The node to be searched and 2. The positional relationship of the nodes of is specified. For example, in the search formula shown in FIG. 7, since the search target is “area title”, “(attribute =
The expression "area title)" is the condition of the attribute of the node to be searched, and the other expression "(attribute = ...)" is the condition of the attribute of the node that should appear in the vicinity of the search target. “/”, “#”, And the like that are output are conditions that specify the positional relationship. The condition for specifying the positional relationship includes a constraint condition regarding the actual length of the branch connecting the nodes in the search expression in the document tree structure (the number of nodes that appear between the corresponding nodes). .

The search result is 1. 1. A node in the document structure that satisfies the attribute condition of 2.
The nodes in the document structure satisfying the conditions of the node attribute shown in 3) exist in the vicinity, and further, it is 3. It exists so as to satisfy the positional relationship described in. That is, the search result is one node in the partial structure (subtree) having the same shape as the search expression in the document tree structure.

Next, the operation relating to the editing of the retrieval formula in the first embodiment will be described. As mentioned above, the search condition is
It is a partial structure (partial tree) of the structure of the document given for illustration. In general, modifying the search conditions is equivalent to transforming this subtree. For example, if the condition is to be strengthened, a tree branch may be added or the branch length may be fixed. On the contrary, when the condition is weakened, the branches of the tree may be cut off or the length of the branches may be freely changed.

However, if the branches are added or pruned freely to the subtree of the conditional expression in this way, there is a risk that even the illustrated portion cannot be searched. Figure 8
It is explanatory drawing of the problem at the time of editing a search formula. Figure 8
The portion indicated by the solid line indicates the tree structure of the search condition in which the partial tree that is the search expression with the mark added in FIG. 5 is extracted. For example, if a new branch (branch of "attribute = text") shown by the broken line in FIG. 8 is added to this tree structure, it cannot be overlapped with the tree structure of the structured document shown in FIG. Therefore, the partial structure illustrated when the search formula is generated becomes unsearchable.

If only the grammatical correctness is checked, such an error cannot be found because the condition of adding the branch shown by the broken line in FIG. 8 is correct. Therefore, in the present invention, when the editing work corresponding to adding a branch to the tree structure and designating the length is performed by referring to the structure shown in FIG. Is allowed only when it can be overlapped with the structure of FIG. As described above, the search expression editing unit 5 includes an error prevention function.

FIG. 9 is an explanatory diagram of an example of the editing operation in the search expression editing unit 5. The modification that can be performed by the editing operation is, for example, as shown in FIG. 9, adding a branch, deleting a branch, freedom of branch length, between two branches (or further outside of the outermost branch). There can be five operations: the degree of freedom of the number of other branches that may exist, and the change of the subexpression pattern. For these editing operations, the functions of the search formula structure extraction unit 2, the search condition extraction unit 3, and the like can be used.

When an attempt is made to edit each of these operations at a certain node, it is determined as follows whether or not the operation is correct.

First, in the case of adding a branch, each node corresponds to the record shown in FIG. 5, and since the parent-child relationship and the sibling relationship of the structure are linked, in the direction instructed to add a branch. Following this link, if the record exists, it can be determined that the addition of the branch is correct. In the case of branch deletion, there is no need to make a decision and it is always correct. However, the degree of freedom regarding the number of branches that may exist between the branches on both sides of the branch to be deleted is set to infinity.

In the case of changing the degree of freedom of the branch length, it is always correct to increase the degree of freedom first. When the degree of freedom is reduced, the parent-child relationship link in FIG. 5 is followed, the length is checked, and if the set degree of freedom includes this length, it is determined to be correct.

In the case of the degree of freedom of the number of branches that may exist between the branches (further outside the outermost branch), it is always correct to increase the degree of freedom. When the degree of freedom is reduced, the sibling links in FIG. 5 are followed, the number of branches between them is checked, and if the designated degree of freedom includes this number, it is determined to be correct.

In the case of changing the pattern of the sub-expression, the field of the corresponding record is referred to, and if it matches the changed pattern, it is determined to be correct.

As described above, the retrieval formula editing unit 5 edits the retrieval formula generated by the retrieval formula combining unit 4 in accordance with the instruction of the editing operation from the user. By executing only the operation that is determined to be correct by the determination method, it is possible to always perform only the transformation of the expression that is not only grammatically correct but also semantically correct.

By using the search expression creation support system described in the first embodiment, the user can create a search expression by only exemplifying one of the search results. As a result, even a general user who does not know the internal structure of a structured document or the grammar of an expression can easily obtain the search expression. However, in essence, the search formulas that can search the exemplified ones are not the only ones, and there are many possible. For this reason, it is highly possible that the search results obtained by the automatically-synthesized search formulas are deficient or insufficient with respect to the intended one. In such cases, the search formula must be modified.

In the above-described first embodiment, the retrieval formula editing means is used to modify the retrieval formula. As a result, the search formula can be edited while always maintaining the correct search conditions. However, in the above-described embodiment of the search formula editing means, when the user wants to correct the automatically obtained search formula, the user modifies the search formula itself. At this time, the user has to think for himself how to modify. That is, the user will again be required some knowledge of the internal structure of the structured document and the grammar of the expression. As such, it is not easy to modify the search formula.

In the second embodiment described below, when an unintended retrieval result is obtained due to an incorrect conditional expression, the erroneous portion of the retrieval result is pointed out rather than the retrieval expression side. By doing so, a search formula creation support system capable of automatically correcting a search formula will be described.

As described in the first embodiment, in general,
The search expression is represented by a tree structure, and modifying the search condition is equivalent to modifying the tree structure of the search expression. As described with reference to FIG. 9, assuming that the subtree shown by the solid line in FIG. 9 is the original search condition, a change related to the state of the branch as shown by the broken line in FIG. 9 is added to this. By doing so, the search condition can be modified. If you want to strengthen the condition, you can add a tree branch or fix the branch length. That is, if the tree that is the structure of the search result of the original document cannot be overlapped, this part is not included in the search result. On the contrary, if you want to weaken the condition, cut the branches of the tree,
The length of the branches may be freely changed.

FIG. 10 is a block diagram showing a second embodiment of the search expression creation support system of the present invention. 21 in the figure
Is a retrieval formula correction processing unit, 22 is an inference unit, 23 is a structure analysis unit, 24 is a database, and 25 is a structured document. In addition to this, there are an input unit and a search unit (not shown).

The retrieval formula correction processing unit 21 receives an error indication from the user from the input unit, and uses the inference unit 22 and the structure analysis unit 23 to include the indicated partial structure of the structured document as a retrieval result. Or modify the search expression so that it is excluded from the search results.

According to the request from the search formula correction processing unit 21, the inference unit 22 determines whether or not to correct the search formula, presents a correction plan for the search formula, and registers and searches in the database 24. The database 24 stores information about each branch of the structure included in the inquiry to the inference unit 22. For each branch, all information about whether or not the branch is included in the current search formula and the degree of freedom of the length of the branch is registered as a hypothesis. In addition, regarding a combination of hypotheses that is found to be incorrect due to a user's indication or the like, information indicating that the combination is incorrect and the reason for the incorrectness are also registered. The reason for incorrectness can be to register the node that the user has pointed out to be incorrect. Further, the database 24 is pre-registered with an evaluation function for evaluating the suitability of a change to the branch based on the size of the entire tree of the search expression, the number of branches to be removed, the shape of the tree, and the like. The evaluation function includes a tree shape as an input parameter. Therefore, it is defined in the form of a pair of a tree and a mathematical formula.

The structure analysis unit 23 analyzes the structure of the search expression according to the instruction of the search expression correction processing unit 1, compares the partial structure of the structured document 25 with the structure of the search expression, and outputs the comparison result. Alternatively, a predetermined partial structure is extracted. For example, whether or not the correction proposal presented by the inference unit 22 includes the branch pointed out by the user is compared with the partial structure of the structured document 25 and the correction proposal, and the result is reported. In addition, when there is a node to be added, when there is a node that does not exist around that node but exists only in the search expression, or when there is a node to delete, there is a node around that node that should be deleted. Extracts branches that do not exist, but exist around nodes that are known to be correct.

FIG. 11 is a flow chart for explaining an example of the operation in the second embodiment of the search expression creation support system of the present invention. First, the search formula correction processing unit 21 first executes S
At 31, the search formula automatically synthesized in the first embodiment or the search formula directly given by the user is obtained as the current search formula. Then, the structure analysis unit 23 is requested to analyze the state related to each branch and the length of the tree structure of the search expression in S32 for this search expression.

In S33, the inference unit 22 is notified based on the response from the structure analysis unit 22 as a hypothesis that the length of each branch and the fact that the branch is included in the search formula are correct. Based on this notification, the inference unit 22 registers information about each branch of the search expression in the database 24.

At S34, using the current search formula,
A search unit (not shown) performs a search and presents the search result to the user. The user refers to the presented search result and determines in S35 whether or not there is an excess or deficiency with respect to what the user intended. If there is an excess or deficiency, the user sequentially points out the incorrect portion in S36. If this indication indicates that the search result is insufficient, the process proceeds from S37 to S38, and the process of modifying the search expression is performed so that the indicated node is added to the search result. If the user's indication indicates a redundant search result, the process proceeds from S37 to S39.
A process of correcting the search formula is performed so that the specified node is excluded from the search. Then, with the modified search expression as the current search expression, it is possible to return to S34 and make further modifications.

FIG. 12 is a flow chart showing an example of the correction processing of the search expression by specifying the additional node. A case will be described in which there is a missing node in the search results obtained by the current search formula. In this case, the user points out an example of what must be included as a search result. Hereinafter, this pointed out portion will be referred to as an additional node. The number of additional nodes to be pointed out is not limited to one, but a plurality of points can be pointed out.

In S41, the retrieval formula modification processing unit 21 compares the tree structure of the retrieval formula with the structure of the additional node and its surroundings in the structured document 25 to the structure analysis unit 23, thereby determining the periphery of the additional node. To find all branches of the search expression that do not exist in, but exist only in the structure of the search expression. The structure analysis unit 23 checks, for each branch of the search expression, whether or not there is a corresponding branch around the additional node. At this time, each branch of the search formula is examined as if the length of the branch can be freely changed, and a reply is sent to the search formula correction processing unit 21. When the length of the branch is changed from the fixed state to the freely changeable state and the check is performed, information about the branch whose length is freely changed is also included in the reply.

In S42, the retrieval formula correction processing unit 21
Asks the inference unit 22 whether or not it is possible to remove from the search formula all the branches obtained only in the structure of the search formula, which are obtained in response from the structure analysis unit 23. When making an inquiry to the inference unit 22, the additional node is notified as the reason why those branches are to be removed. The inference unit 22 evaluates whether or not the branch may be removed, and notifies the search expression correction processing unit 21 of the possibility. Similarly, in S43, the inference unit 22 is inquired as to whether or not the branch whose free length is changed by the structure analysis unit 23 can be freely changed, and the reply from the inference unit 22 is received. .

In S44, if all the inquiries are given from the inference unit 22, the retrieval formula can be modified. At this time, the inference unit 22 not only issues the permission, but also issues an instruction regarding addition of another branch to the search expression and designation of the branch length.
According to this, the search formula correction processing unit 21 changes the tree structure of the search formula to that state. That is, the branch permitted by the inference unit 22 is deleted in S45, the branch instructed by the inference unit 22 is added in S46, and the length of the branch instructed by the inference unit 22 is changed in S47. . With these processes, the process of changing the tree structure of the search formula by the additional node in the search formula modification processing unit 21 is completed.

In S44, if there is any branch for which the inference unit 22 has not given permission, it is judged that the branch should be searched by another formula, and the correction is given up. Then, measures such as notifying the user of that fact and calling a process for automatically synthesizing a search expression for the part are taken.

FIG. 13 is an explanatory view of an example of a partial structure of a structured document for explaining a concrete example of correction of a search expression, and FIG.
FIG. 7 is an explanatory diagram showing an example of the structure of a search expression to be modified. For example, consider a case where a structured document having a structure as shown in FIG. 13 is searched by a search formula having a structure including a broken line part as shown in FIG. The * mark in the figure indicates the extraction position. In this case, the search result is node a in FIG.
"I think the introduction wasn't interesting ..." Here, the user points out that the node b “Introduction ...” in FIG. 13 should also appear as a search result. Then, the structure analysis unit 23 compares the structure of the search expression shown in FIG. 14 with the surrounding partial structure including the node b of FIG. 13, and does not exist around the node b, and only the structure of the search expression is present. Extract existing branches. In this example, since the branch k indicated by the broken line in FIG. 14 does not exist around the node b in FIG. 13, the structure analysis unit 23 extracts the branch k.
Then, if the inference unit 22 permits this branch k to be removed from the structure of the search expression, the search expression correction processing unit 21 removes the branch k from the structure of the search expression, and only the structure shown by the solid line in FIG. 14 is obtained. The search formula is modified as follows. Then, by performing the search using the modified search formula, the search result includes both the parts of the node a and the node b in FIG.

FIG. 15 is a flow chart showing an example of the correction processing of the search expression by designating the deleted node. When an extra result is obtained as a search result, the user points out an example of the extra result. Hereinafter, this pointed out portion will be referred to as a deletion node.

When the user designates a deletion node, S51
In the above, the retrieval formula correction processing unit 21 instructs the inference unit 22 to
The current search formula is incorrect and requires you to give instructions on how to correct it. When a correction instruction is received from the inference unit 22, the process proceeds from S52 to S53, and the search formula correction processing unit 21
Requests the structure analysis unit 23 to check whether or not the deleted node will not be included in the search result by modifying the search expression based on the instruction received from the inference unit 22. S5
In step 4, when the reply from the structure analysis unit 23 indicates that the deleted node will not be included, in S55, the correction instruction previously obtained from the inference unit 22 is adopted, and according to the instruction Change the structure of the search expression to that state.

If the deleted node is still included in the search result in accordance with the correction instruction received from the inference unit 22 in S54, the inference unit 22 is executed in S56.
Request that a new correction method be instructed because the previous correction plan is incorrect. Then, the process returns to S52, and the retrieval formula correction processing unit 21 repeats this until the inference unit 22 gives a correct correction plan.

When the response from the inference unit 22 indicating that the correction cannot be made is received in S52, the processes in and after S57 are performed. In S57, the search formula modification processing unit 21 instructs the structure analysis unit 23 to create a partial tree that overlaps the search formula tree structure with respect to the structure around the node and the structure around the deleted node that are already known to be correct as the search target. It is instructed to find all branches that are in contact with each other and that exist around the node that is known to be correct, but do not have a corresponding branch around the deleted node. Then, in S58, the inference unit 22 is inquired about which of the branches extracted by the structure analysis unit 23 should be added to the search formula. At this time, the deleted node is notified as the reason for adding. The inference unit 22 returns an instruction to the search formula correction processing unit 21 so as to select one from the branches extracted by the structure analysis unit 23, specify its length, and add it to the search formula tree. When an additional instruction is returned from the inference unit 22, the process proceeds from S59 to S60, and the retrieval formula correction processing unit 21
Change the structure of the search expression tree to that state according to additional instructions from.

Alternatively, in S59, the inference unit 22 determines that it is better not to add any branch, and when such a reply is received, that part is changed by another method, for example, the node attribute or the like is changed. And decide that it should be excluded, and give up the correction. Then, measures are taken such as notifying the user of that fact and calling a process for automatically obtaining a node attribute for excluding the part.

A concrete example will be described with reference to FIGS. 13 and 14 described above. For example, consider a case where a structured document having a structure as shown in FIG. 13 is searched by the structure of the search formula of only the solid line part shown in FIG. The search result at this time is the part of the node a “Introduction ...” and the node b “I think the introduction was not interesting ...” in FIG. Here, when the user points out that the part of the node b in FIG. 13 should not appear as a search result, the search formula correction processing unit 21 requests the inference unit 22 to instruct the correction method. Here, it is assumed that the inference unit 22 has received a reply that the correction cannot be made.

The retrieval formula modification processing unit 21 requests the structure analysis unit 23 to examine a subtree around the nodes a and b in FIG. 13 corresponding to the solid line tree in FIG.
In the case of this example, the branches l, m, and n that are in contact with the subtree that overlaps the tree structure of the search expression of the node a that is known to be correct are obtained. Among them, since the branch m has a branch corresponding to the periphery of the node b, the branches l and n are extracted here. The extracted branches l and n are sent to the inference unit 22 and one of them is selected. For example, when the inference unit 22 determines that it is best to add the branch l to the branch of the search expression, the search expression correction processing unit 21 adds the branch l to the branch l of FIG. 13 as indicated by the broken line in FIG. Modify the structure of the search expression to include the corresponding branch k. By this modification, the search result using the modified search formula does not include the node b portion in FIG.

As can be seen from the above, the user does not have to point out all the erroneous portions from the beginning, and if he points appropriately, a search is executed and this process is repeated to obtain the intended result. By the way, you can stop the correction. Executing the search on the way does not affect the search formula modification process of the present invention.

16 to 19 are flowcharts for explaining an example of the operation of the inference unit 22. The request to the inference unit 22 is as follows, as can be seen from the above description. Request 1. The information that each branch is included in the search formula and the information regarding the length thereof are registered in the database 24 as hypotheses. (S33 of FIG. 11) Request 2. Decide whether to remove the branch from the search expression or to free the length. (S42 and S4 in FIG.
3) Request 3. Submit a suggestion when the current search formula is incorrect. (S51 in FIG. 15) Request 4. Select one branch from the branch candidates that can be added to the search formula. (S58 in FIG. 15)

In S61 of FIG. 16, the inference unit 22 receives the request. In S62, it is determined which of the above-mentioned request 1 to request 4 the received request is, and in accordance with the request, any one of the processes of S63 to S66 is selectively executed. Then, in S67, an answer is returned to the request source.

The processing for each request will be described below.
An example will be described. Regarding request 1 described above, only the process of registering in the database 24 is performed by using the information that each branch is included in the search expression and the information regarding its length as a hypothesis.

Regarding the request 2, in S71 of FIG. 17, first, the evaluation function corresponding to the tree which is the subtree of the search expression is searched from the database 24. Here, the structure analysis unit 23 determines whether or not it is a subtree. By storing the tree structure in the database 24 in consideration of the priority order in advance, it is possible to adopt the first searched data. A mathematical expression is extracted from the evaluation function corresponding to the searched tree structure. Then, in S72, the mathematical expression is given an argument other than the tree shape, for example, information such as the overall size of the search expression and the number of branches to be removed.

In S73, it is determined whether or not the calculated evaluation value is within the preset range. If it is not within the range, in S74, the branch is removed from the search expression or the length is increased. It is returned to the request source as an answer not to allow freeness. If the evaluation value is within the range, in S75, a reply to the effect that the request is permitted is returned to the request source. Then, in S76, the fact that the combination of branch states included in the current search formula is incorrect is registered in the database 24 with a reason. Further, in S77, the information in the database 24 is updated to the state in which the hypothesis is not established regarding the branch which has been removed or whose length has been freely set.

For example, in the structure of the search formula shown in FIG. 14 described above, when the branch k indicated by the broken line is removed from the branch of the search formula, the data which is the subtree of the structure of the search formula shown in FIG. Search and retrieve from database 24. The formula paired with the extracted data is, for example, (1- (the number of branches to be deleted) / (the number of branches of the entire tree)) *
It is assumed that the expression is 100. The number of branches to be deleted is one, and the number of branches of the entire tree is five. By substituting these values, the evaluation value becomes (1-1 / 5) * 100 = 80. Then, the deletion permission is issued if the value is defined to be 60 or more. Of course, if the deletion permission is defined to be 90 or more, the deletion permission is not issued.

By the way, in this embodiment, two different evaluation functions are used below. The implementation method is the same for them. However, since the following two evaluation functions are merely devised to reach a correct search expression sooner, they may return a random number or a constant, for example.
In that case, it may just be that the user needs a little more error indication to reach the intended expression.

Regarding the request 3, first, S81 in FIG.
In (2), the fact that the combination of branch states included in the current search formula is incorrect is registered in the database 24 with a reason. In S82, one of the hypotheses registered in the database 24 is reversed to obtain a new search expression candidate. At this time, all the states other than the states that include combinations of branches that are known to be incorrect and are registered in the database 24 are obtained by the request to the inference unit 22 up to this point. As a result, it is possible to obtain the learning effect using the history that was determined to be incorrect in the past. There are multiple calculated states. Therefore, in S83, evaluation is performed using the evaluation function prepared for each state. In S84, the highest evaluation value is selected based on the calculated evaluation values. In S85, it is determined whether or not the evaluation value is within a predetermined range.
At 86, the fact that there is no correction plan is returned to the request source as an answer. If the highest evaluation value is within the predetermined range, in S87,
The information of the hypothesis that was converted to obtain the state of the highest evaluation value is returned to the request source as a correction proposal. Then, in S88,
The converted hypothesis is reflected in the database 24.

With respect to the request 4, in S91 of FIG. 19, the database 24 is searched based on the shape of the tree or the like to obtain an evaluation function for evaluating whether or not to add. And S9
In step 2, the obtained evaluation function is used to evaluate, with respect to each branch of the addition candidate, the size of the branch and the entire tree of the search expression as an input. In S93, the best evaluation result by the evaluation function is selected. In S94, it is determined whether or not the selected evaluation result is valid, and if it is not valid, it is returned to the request source that the addition is not permitted in S95. If the evaluation result is valid, in S96, a reply is made to the request source to add a branch having the selected evaluation result.
In S97, the fact that the combination of the branch states included in the current search formula is incorrect is registered in the database 24 with a reason. Further, in S98, the information in the database 24 is updated for each branch of the addition candidate. At this time, the selected branch is registered as a state in which the hypothesis holds, and the branch not selected is registered as a state in which the hypothesis does not hold.

As can be seen from the above flow chart,
In S76 of FIG. 17, S81 of FIG. 18, S97 of FIG. 19, etc., the fact that the set of branches included in the search formula is incorrect is registered in the database, and in S82 of FIG. Search formula candidates that include sets are excluded. As a result, corrections are made without repeating the same mistakes as twice, so that if the number of points pointed out increases, it is possible to gradually approach the correct expression. In addition, because there is a reason for the incorrect combination of branch states, it is possible to return to the original state if this reason itself is canceled, and the user's error pointing operation was incorrect. You can also deal with it.

In the second embodiment, it is of course possible to use this configuration alone, and in that case, the retrieval formula is given in advance. It can also be incorporated as the search expression editing unit 5 in the first embodiment. In this case, the functions of the structure analysis unit 23 and the like can be realized by using the search expression structure extraction unit 2, the search condition extraction unit 3, and the like. It is also possible to share the database 24 with the knowledge base 6.

[0080]

As is apparent from the above description, according to the present invention, a search expression can be automatically created by a simple operation. For example, the internal structure of a structured document and the expression grammar can be known. Even a non-user can easily obtain the intended expression. This enables general users to effectively utilize the advantages of structured documents. In addition, it is also useful for experienced users, inadvertent mistakes can be almost completely eliminated, and the intended formula can be created accurately in a very short time. Further, in the configuration having the search formula editing means, there is an effect that the search formula can be edited while always maintaining the correctness of the search condition.

Further, since the retrieval expression of the structured document can be automatically corrected by a simple operation of pointing out an error in the retrieval result, even a user who does not know the grammar of the retrieval expression of the structured document intends. You can easily get the formula This enables general users to effectively utilize the advantages of structured documents. Further, it is also useful for a proficient user, and there is an effect that an error due to carelessness can be almost completely eliminated and an intended expression can be accurately created in a very short time.

[Brief description of drawings]

FIG. 1 is a configuration diagram of a first embodiment of a search expression creation support system of the present invention.

FIG. 2 is a flowchart showing an operation of generating a search expression in the first embodiment of the search expression creation support system of the present invention.

FIG. 3 is an explanatory diagram of an example of a structured document.

FIG. 4 is an explanatory diagram of an example of a structured internal representation of a structured document.

5 is an explanatory diagram of an example of a structure of a structured document stored in a memory 7. FIG.

FIG. 6 is an explanatory diagram of an example of a typical structural pattern accumulated in the knowledge base 6.

FIG. 7 is an explanatory diagram of an example of a generated search expression.

FIG. 8 is an explanatory diagram of a problem when a search expression is edited.

FIG. 9 is an explanatory diagram of an example of an editing operation in the search formula editing unit 5.

FIG. 10 is a block diagram showing a second embodiment of the search expression creation support system of the present invention.

FIG. 11 is a flowchart illustrating an example of an operation in the second embodiment of the search expression creation support system of the present invention.

FIG. 12 is a flowchart showing an example of a search expression correction process by specifying an additional node.

FIG. 13 is an explanatory diagram of an example of a partial structure of a structured document for explaining a specific example of correction of a search expression.

FIG. 14 is an explanatory diagram showing an example of the structure of a search expression to be modified.

FIG. 15 is a flowchart showing an example of a correction process of a search expression by designating a deleted node.

16 is a flowchart illustrating an example of the overall operation of the inference unit 22. FIG.

FIG. 17 is a flowchart illustrating an example of an operation for determining whether to remove a branch from a search expression or to make the length free in the inference unit 22.

FIG. 18 is a flowchart illustrating an example of an operation of presenting a correction plan in the inference unit 22.

FIG. 19 is a flowchart illustrating an example of an operation of selecting one branch from addition candidate branches in the inference unit 22.

[Explanation of symbols]

1 ... Structural analysis unit, 2 ... Search formula structure extraction unit, 3 ... Search condition extraction unit, 4 ... Search formula synthesis unit, 5 ... Search formula editing unit, 6 ... Knowledge base, 7 ... Memory, 8 ... Search formula, 9 … Structured documents,
21 ... Retrieval expression correction processing unit, 22 ... Inference unit, 23 ... Structure analysis unit, 24 ... Database, 25 ... Structured document.

Claims (4)

[Claims] 1. A search formula creation support system for supporting the creation of a search formula used when performing a search process on a structured document, and a storage means for storing the structured document,
Structure extracting means for extracting a partial structure including the exemplified partial structure based on the partial structure of the structured document exemplified by the user as one of the search results, and searching from the partial structure extracted by the structure extracting means A search expression creation support system having a search expression synthesizing means for synthesizing expressions. 2. When the search formula synthesized by the search formula synthesis means is transformed in accordance with an editing operation instruction from a user, the transformed search formula includes a partial structure used in the synthesis. The search expression creation support system according to claim 1, further comprising a search expression editing unit that executes a modification process of the search expression under the condition. 3. A search formula creation support system for supporting the creation of a search formula used when performing a search process on a structured document, and based on a search means for performing a search according to the search formula and a search result by the search means. An input means for the user to point out an error in the search result, and a search formula correction processing means for correcting the search formula based on the structure of the partial structure including the part pointed out according to the error point input by the input means and the structure of the search formula. A search expression creation support system characterized by having. 4. A database that sequentially stores information about each branch included in the structure of the search expression in which an error is pointed out and the combination of the branches, and the search expression correction processing means corrects the search expression. When referring to the database,
4. The search expression creation support system according to claim 3, wherein the search expression is corrected based on correction candidates other than the search expression correction candidates including the combination of branches stored in the database.