JP2005339150A - Document retrieval device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a document retrieving device capable of shortening the retrieval time with respect to various retrieval conditions. <P>SOLUTION: When the retrieval instruction for the neighborhood conditions of a plurality of keyword groups is issued, a neighborhood condition deciding part 108 of a retrieval processing executing part 105 acquires the appearing positions in a document of the plurality of keyword groups from a keyword group collating part 107, and decides whether the instructed predetermined neighborhood conditions are satisfied by the acquired appearing positions. The acquisition processing of the appearing positions and the neighborhood condition decision processing are executed alternately, and when it is decided that the neighborhood conditions are true by the neighborhood condition deciding part 108, the decision result is outputted as a retrieval result at that point of time. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a document search apparatus that outputs a document that matches a predetermined search condition for a keyword set.

  Conventionally, as a document search apparatus, for example, as shown in Patent Document 1, wasteful data transfer of appearance position information in a keyword document is reduced, and the determination of the compound condition determination is accelerated. there were. Here, the compound condition is a neighborhood condition, a range condition, an attribute condition, a context condition, and their logical conditions.

JP-A-4-293161

  However, the conventional document search device described in Patent Document 1 makes it efficient to transfer data of appearance position information in a keyword document and speeds up determination of composite condition determination. In this case, there is only a brute force determination of the appearance positions of two or more keywords, and there is a problem that the amount of calculation required for the determination increases as the appearance positions of the keywords in the document increase. It was.

  The present invention has been made to solve the above-described problems, and an object of the present invention is to provide a document search apparatus capable of shortening the search time even for various search conditions.

  The document search device according to the present invention obtains the appearance positions of a plurality of keyword sets in a document in the order of the appearance positions of the keyword sets, and whether the appearance positions of the plurality of keyword sets satisfy a predetermined neighborhood condition. And a proximity process determination unit that alternately executes a proximity condition determination process that determines whether or not, and outputs the determination result as a search result when the proximity condition is determined to be true.

  The document search apparatus according to the present invention alternately executes the appearance position acquisition process in the keyword set matching unit and the neighborhood condition determination process in the neighborhood condition determination unit, and the determination result is obtained when the neighborhood condition is determined to be true. Since the search result is output, the search result can be output when the neighborhood condition is satisfied. Therefore, even when the number of keywords and the appearance positions of the keywords in the document are large, the search time can be shortened. be able to.

Embodiment 1 FIG.
FIG. 1 is a block diagram showing a document search apparatus according to Embodiment 1 of the present invention.
This document search apparatus is a document search apparatus that outputs a document that meets a proximity condition specified as a search condition from a database that stores one or more documents.

  In the figure, a document search apparatus includes a search condition input unit 103, a search condition analysis unit 104, a search processing execution unit 105, a search result output unit 106, a keyword set collation unit 107, a neighborhood condition determination unit 108, a keyword collation unit 109, a database. 110 and a document index 111.

  The search condition input unit 103 is a functional unit that receives an input of the search condition 101 from a search user. In the search condition 101, information on two or more keyword sets to be searched and a neighborhood condition between the keyword sets are specified. The search condition analysis unit 104 is a functional unit that analyzes the search conditions input from the search condition input unit 103 and generates a search execution plan such as a predetermined neighborhood condition search. The search processing execution unit 105 is a functional unit that executes search processing according to the search execution plan output from the search condition analysis unit 104. The search result output unit 106 is a functional unit that outputs the search result 102 obtained by the search processing execution unit 105 to a search user.

  The search processing execution unit 105 includes a keyword set matching unit 107 and a neighborhood condition determination unit 108. The keyword set matching unit 107 has a function of acquiring the appearance position in the document of the keyword included in the keyword set specified in the search condition by repeatedly calling the keyword matching unit 109. The keyword matching unit 109 has a function of acquiring the appearance position of the keyword in the document while referring to the document index 111 accumulated in the database 110. In addition, the neighborhood condition determination unit 108 has a function of determining the neighborhood condition specified as the search condition based on the appearance position information in the documents of two or more keyword sets output from the keyword set matching unit 107. Have. That is, the search processing execution unit 105 alternately executes the appearance position acquisition process in the keyword set matching unit 107 and the neighborhood condition determination process in the neighborhood condition determination unit 108, and when the neighborhood condition is determined to be true, The determination result is output as a search result.

  Note that the document search device described above is realized by a computer, and the search condition analysis unit 104 and the search processing execution unit 105 include software corresponding to each function, and hardware such as a CPU and a memory for executing the software. It is composed of

  A keyword set is a set including one or more keywords. A keyword set that does not contain any keywords does not have any appearance position in the document, and is not considered here.

  The keyword k appears in the document D when the same character sequence as the keyword k exists in the document D. When the keyword k appears in the document D, the appearance position of the keyword k is represented by the number of characters from the first character of the document D to the first character of the keyword k. Since the keyword k appears 0 or more times in the document D, the keyword k can also be regarded as a set of appearance positions in D of k.

FIG. 2 is an explanatory diagram showing the relationship between documents and keywords.
A set of appearance position in the document D of keyword k represents a k ^D. In the example of FIG. 2, when the keyword k = “keyword”, the appearance position of k in the document 201 is k ^D = {1, 18, 35} as indicated by 202 in the figure. The number of characters of the keyword k is called the keyword length and is represented as LEN (k). A process of acquiring one appearance position of the keyword k in the document D is represented as STR ^D (k). Further, the position of the last character of k is represented as END ^D (k). The relationship between STR ^D (k) and END ^{D (} k) is shown below.

図２の文書２０１の例では、最初の”キーワード”と”出現”の距離は３である。 When there are two keywords _{k 1} and _k 2 ^(STR _D (k 1) and _{^{<STR D (k 2))}} , the distance between the keyword, the end of the keyword _{k 1} character and the keyword _{k 2} beginning of It is represented by the number of characters between characters, and is represented as DIST ^D (k ₁ , k ₂ ). DIST ^D (k ₁ , k ₂ ) is defined as follows.

In the example of the document 201 in FIG. 2, the distance between the first “keyword” and “appearance” is 3.

When the keyword set is S, S = {k ₁ , k ₂ ,..., K _n } (n ≧ 1). The appearance position of the keyword set S in the document D is defined as the appearance position of the keyword k _i (i = 1 to n) in S. When S ^D is a set of appearance positions in the document of the keyword set S, S ^D = k ₁ ^D ∪k ₂ ^D ∪... ∪k _n ^D. In the example of FIG. 2, when the keyword set is S = {“keyword”, “appearance”}, S ^D = {1, 9, 18, 35, 43} as shown by 203 in FIG. The smallest appearance value in the keyword set in the document is called the first appearance position of S. Also, the first appearance position in the document after the appearance position of the keyword set acquired immediately before is called the next appearance position of the keyword set. In the example of FIG. 2, the first appearance position of the keyword set S is 1. The next appearance position of the keyword set S is nine.
The distance between two keyword sets is defined by the distance between keywords included in the two keyword sets.

  The neighborhood condition is a search condition for determining authenticity based on a distance between two or more keyword sets.

FIG. 3 is a flowchart of search processing in the document search apparatus of the present invention.
First, when the search condition 101 is input to the search condition input unit 103 in step ST301, a search execution plan is generated by the search condition analysis unit 104 in step ST302. Next, the search processing execution unit 105 determines whether the documents stored in the database 110 meet the neighborhood condition one by one.

  First, in step ST303, it is determined whether one document meets the neighborhood condition. If the neighborhood condition is met (YES), the document is added to the search result (step ST304). The search result may hold only the document identifier or may hold additional information. When a document is added to the search result, the neighborhood condition determination in step ST303 is performed on the next document. If the document does not conform to the neighborhood condition in step ST303 (NO), the neighborhood condition is determined for the next document without adding the document to the search result (step ST303). In this way, when the neighborhood condition is determined once for all the documents, the search result output unit 106 outputs the search result 102 in step ST305.

FIG. 4 is a flowchart of the process (step ST303 in FIG. 3) of the search process execution unit 105 in FIG.
Here, it is assumed to determine the neighborhood condition of keyword set S ₁ to S _n. First, neighborhood condition judging unit 108, the keyword set collation unit 107, for each keyword set _S 1 to S _n, and first occurrence in a document to be determined, the keyword _k 1 to k _n at that appearance positions Is acquired (step ST401). The information acquired here may be the appearance position of the keyword set and the length of the keyword.

  The keyword set matching unit 107 calls the keyword matching unit 109 each time a keyword set appearance position acquisition request is made, and sequentially displays the keyword set appearance position and the keyword at that position, or the keyword length, one by one from the top of the document. Is output.

Next, the neighborhood condition determining unit 108 determines whether or not the position of the acquired keyword set matches the neighborhood condition in step ST402. In step ST402, if the document to be determined conforms to the neighborhood condition (YES), the process proceeds to step ST406, “conforms” is output, and the determination process ends. On the other hand, if the appearance position of the keyword set does not match the neighborhood condition in step ST402 (NO), the process proceeds to step ST403, in the document of the keyword set S _i (i = 1 to n) that does not satisfy the neighborhood condition. Get the next occurrence position at.

  In step ST404, it is determined whether or not the next appearance position of the keyword set has been acquired in step ST403. If it has been acquired (YES), the process returns to step ST402, and whether or not these appearance positions meet the neighborhood condition. Determine whether. If the next appearance position of the keyword set has not been acquired in step ST404 (NO), it means that there is no appearance position that matches the neighborhood condition in the document to be determined, so that “not compatible” is output in step ST405. And exit.

  The keyword matching unit 109 outputs the appearance position of the keyword in the document while referring to the document index 111 recorded in the database 110. Here, the document index 111 recorded in the database 110 may be an index that records a character string and an appearance position in the document, such as an n-gram index or a suffix array. The appearance position of the keyword in the document may be acquired by directly scanning the document. That is, as long as it has a function capable of acquiring the appearance position of a keyword in a document to be determined, any means for realizing it may be used.

  As described above, according to the first embodiment, the keyword set matching unit 107 that acquires the appearance positions of a plurality of keyword sets in a document in the order of the appearance positions of the keyword sets and the keyword set matching unit 107 are acquired. A proximity condition determination unit that determines whether or not the appearance positions of a plurality of keyword sets satisfy a predetermined neighborhood condition, and the appearance position acquisition processing in the keyword set matching unit 107 and the neighborhood condition determination unit in Since the search processing execution unit 105 that alternately executes the proximity condition determination process and outputs the determination result as a search result when the proximity condition is determined to be true, the number of keywords and the keyword in the document are included. Even if the appearance position increases, the amount of calculation required for the determination can be suppressed, which can contribute to shortening the search time in such a case.

  For example, a keyword set {“digital”, “digital”, “digital”} and {“camera”, “camera”, “camera”} in which the keywords “digital” and “camera” are developed synonyms and different notations are used. Even in the case of determining the neighborhood condition, in the present embodiment, these keywords can be directly determined as a keyword set, and therefore the search time can be shortened.

Embodiment 2. FIG.
In the second embodiment, the in-neighbor conditions with keyword group order designation are determined.

  Since the configuration of the second embodiment in the drawing is the same as that of the first embodiment shown in FIG. 1, description will be made with reference to FIG. The document search apparatus according to the second embodiment is configured such that the neighborhood condition determination unit 108 of the document search apparatus according to the first embodiment can determine the in-neighbor conditions with keyword group order designation. Further, since the entire flow of the search process is the same as that shown in FIG. 3, the description thereof is omitted here.

Here, the in-neighbor conditions with keyword group order specification will be described. The in-neighbor condition with keyword set order specification is a neighborhood condition for determining whether or not a keyword set appears in the specified order and whether the distance between the preceding and following keyword sets is equal to or less than the specified distance. Now, keyword set _S 1 to S _n are as specified in this order. Further, it is assumed that the distance d between the keyword sets is designated. At this time, the condition is determined as follows.

FIG. 5 is a flowchart of the process (step ST303 in FIG. 3) of the search process execution unit 105 in FIG. 1 in the second embodiment.
Search processing execution unit 105, a search condition analysis unit 104, the determination instruction for distance d and ordering there near the conditions between keyword set S ₁ to S _n and keyword set is given, the neighborhood condition determination unit 108, the keyword set _S 1 to S _n in step ST 501, the first occurrence in a document determination target, the keyword _k 1 to k _n in its appearance position is obtained from the keyword set collation unit 107. Next, in step ST502, the neighborhood condition determining unit 108 determines whether or not the appearance position of the acquired keyword set conforms to the in-neighbor conditions with order designation. That is, it is determined with respect to _{^{i = 1~n-1, DIST D}} (k i, k i + 1) whether to satisfy all ≦ d.

In step ST502, if the appearance position of the acquired keyword set for the document to be determined matches the in-neighbor condition with order specification (YES), the process proceeds to step ST506, and “adapted” is output. And exit. On the other hand, if it is determined in step ST502 that the order designation and in-neighbor conditions are not met (NO), the process proceeds to step ST503. In step ST503, which keyword appears first in the document with respect to the smallest i in the set of k _i and k _{i + 1} that did not satisfy DIST ^D (k _i , k _{i + 1} ) ≦ d. judge.

In step ST 503, in a document to be determined, if more of _{k i} is long appeared earlier than _{k i + 1} (YES), the next occurrence position of _{S i} in step ST 504, the keyword _{k i} at that appearance positions get. On the other hand, if k _{i + 1} appears earlier than k _{i in} the document to be determined in step ST503 (NO), the process proceeds to step ST505 and the next appearance position of S _{i + 1} and its appearance position. The keyword k _{i + 1 in} is acquired. When the next appearance position of the keyword set and the keyword at the appearance position are acquired in step ST504 or step ST505, the process returns to step ST502, and whether or not the acquired position of the keyword set conforms to the in-neighbor conditions with specified order. Determine whether.

  Although omitted in the flowchart of the processing of FIG. 5, if the appearance position of the keyword set cannot be acquired in step ST501, step ST504, or step ST505, the order is specified in the determination target document and the near-in condition is set. Since there is no matching appearance position, “not fit” is output and the process ends.

FIG. 6 shows a determination path for the in-neighbor conditions with keyword group order designation according to the processing flow shown in FIG. 5 of the second embodiment.
Now, it is assumed that keyword sets S ₁ , S ₂ , S ₃ are given in this order as search conditions, and the distance designation is d. The appearance position of each keyword set in the document D to be determined is S ₁ ^D = {p ₁ , p ₂ , p ₃ }, S ₂ ^D = {p ₄ , p ₅ , p ₆ }, S ₃ ^D = {P ₇ , p ₈ , p ₉ }. The relationship between the appearance positions is as shown in FIG. Further, as the keyword sets S ₁ , S ₂ , S ₃ , for example, S ₁ is {digital, digital, digital}, S ₂ is {camera, camera, camera}, and each keyword set includes a plurality of keyword sets. Includes keywords.

  When such a search condition is given, the determination is executed as follows. The following processes (1) to (4) correspond to (1) to (4) in FIG.

(1) The first appearance positions p ₁ , p ₄ , and p _{7 in} the document D of the keyword sets S _{1 to} S ₃ are acquired in step ST501 of FIG. Next, in step ST502, it is determined whether or not the appearance position is in conformity with the in-neighbor conditions with the specified order. In the case of FIG. 6, since the keyword set S ₂ appears before S _1, it does not conform to the in-neighbor conditions with order designation. Therefore, the process proceeds to step ST503. In step ST503, the appearance positions of the keyword sets S ₁ and S _{2 in} the document D are determined. Since S ₂ appears earlier, the next appearance position p of the keyword set S _{2 in} step ST505. _{Get 5} .

(2) This time, in step ST502, the determination of the in-neighbor conditions with order designation is performed on the appearance positions p ₁ and p _{7 of} S ₁ and S ₃ and p ₅ acquired in step ST505. In this case, since the distance between S ₁ and S ₂ is larger than d, the in-neighbor condition is not satisfied with the order designation. Therefore, the process proceeds to step ST503. In step ST 503, towards _{S 1} it is because it appears in the document D earlier than _{S 2,} to obtain the following appearance position p2 of _{S 1} in step ST 504.

(3) in step ST 502, it determines the order has near the condition for the occurrence position _{p 2, p 5, p 7} . Now since the direction of _{S 3} have appeared previously in the document D than _{S 2,} to get the next occurrence position _{p 8} of _{S 3} in step ST505.

(4) In step ST502, the in-neighbor conditions with order designation are determined for the appearance positions p ₂ , p ₅ , and p ₇ , and since “matched” is output because “matched” is output in step ST506.

  The number of appearance positions of the keyword in the document can be regarded as being proportional to the number N of characters in the document. That is, the number of appearance positions of the keyword in the document can be expressed as O (N) (order N) in the order notation. Let M be the number of keywords included in the keyword set. Consider a case where an in-neighbor condition with order specification between K keyword sets is determined.

In the conventional method described in Patent Document 1, the in-neighbor condition with the order designation is determined by the round robin for the appearance position of the keyword. At this time, the amount of calculation required for determining the condition between K keywords was O (N ^K ). Further, the determination between keyword sets has to be performed by expanding the keyword sets and individually determining the conditions. Thus, the number of combinations will be ^{M K.} Therefore, the amount of calculation required to determine the in-neighbor condition with the order between the K keyword sets is O ((MN) ^K ).

  On the other hand, consider the case of determination of in-neighbor conditions with order designation in the second embodiment. The number of appearance positions in the document of the keyword set can be considered to be proportional to the total number N of characters in the document and the number of keywords M in the keyword set. That is, O (MN). In the determination method of the present embodiment, it is possible to determine the condition within the neighborhood with the specified keyword set order by the number of times of the sum of the appearance positions of all the keyword sets. Therefore, the amount of calculation required for determining the in-neighbor condition with order specification between K keywords is O (KMN).

  As described above, in the document search apparatus according to the second embodiment, it is possible to determine the in-neighbor conditions with the order designation between the keyword sets at a higher speed than in the conventional method.

  Here, the method of acquiring the appearance position of the keyword set in order from the beginning of the document and determining the in-neighbor conditions with the order designation has been shown, but even if the conditions are determined in order from the end to the beginning of the document, Similar effects can be obtained.

  As described above, according to the second embodiment, when all keyword sets appear in the document, the distance between the preceding and following keyword sets is specified as the vicinity condition of the determination process performed by the vicinity condition determination unit 108. Compared to the conventional method, the near-internal condition with specified order that is less than the distance can reduce the amount of calculation required to determine the in-neighbor condition with specified order, and shorten the search time in such search processing. Can do.

Embodiment 3 FIG.
In the third embodiment, the in-neighbor conditions without specifying the order of the keyword set are determined.

  Since the configuration of the third embodiment on the drawing is the same as that of the first embodiment shown in FIG. 1, description will be made with reference to FIG. The document search apparatus according to the third embodiment is configured such that the neighborhood condition determination unit 108 of the document search apparatus according to the first embodiment can determine the in-neighbor conditions without specifying the order of the keyword set. Further, since the entire flow of the search process is the same as that shown in FIG. 3, the description thereof is omitted here.

Here, the condition in the neighborhood without specifying the order of the keyword set will be described. The in-neighbor condition without specifying the order of keyword sets is a neighborhood condition for determining whether or not the distance between preceding and following keyword sets is equal to or less than a specified distance when all keyword sets appear in a document. Now, when the keyword set S ₁ to S _n and the distance d between the keyword set is given, without ordering near the condition is determined as follows.

FIG. 7 is a flowchart of processing (step ST303 in FIG. 3) of the search processing execution unit 105 in FIG. 1 when there are two keyword sets in the third embodiment.
When the distance d between the keyword sets S ₁ and S ₂ and the keyword set is given to the search processing execution unit 105, the neighborhood condition determination unit 108 determines the document to be determined of the keyword sets S ₁ and S ₂ in step ST701. The first appearance position and the keywords k ₁ and k ₂ at the appearance position are acquired from the keyword set collating unit 107. Next, in step ST702, the neighborhood condition determining unit 108 determines whether the appearance position of the acquired keyword set conforms to the in-neighbor conditions with order designation. That is, it is determined whether DIST ^D (k ₁ , k ₂ ) ≦ d or DIST ^D (k ₂ , k ₁ ) ≦ d is satisfied.

If, in step ST702, the appearance position of the acquired keyword set for the document to be determined matches the in-neighbor conditions without order designation (YES), the process proceeds to step ST706, and “matched” is output. And exit. On the other hand, if it is determined in step ST702 that the in-neighbor conditions with specified order are not met (NO), the process proceeds to step ST703. In step ST703, it is determined whether either of the keyword of _{k 1} and _{k 2} have emerged earlier in the document. In the document to be determined, more of _{k 1} is if appearing before the _{k 2} (YES), the next occurrence position of _{S 1} in step ST704, acquires the keyword _{k 1} in its appearance position.

In step ST 703, in a document to be determined, if more of the _{k 2} is long appeared earlier than _{k 1} (NO), the next occurrence position of _{S 2} in step ST 705, the keyword _{k 2} at that appearance positions get. When the next appearance position of the keyword set and the keyword at the appearance position are acquired in step ST704 or step ST705, the process returns to step ST702, and whether or not the position of the acquired keyword set meets the in-neighbor conditions with specified order. Determine whether.

  Although not shown in the flowchart of the processing in FIG. 7, if the appearance position of the keyword set cannot be acquired in step ST701, step ST704, or step ST705, the near-in-condition without order designation is set in the determination target document. Since there is no matching appearance position, “not fit” is output and the process ends.

FIG. 8 shows a path for determining the in-neighbor conditions without specifying the order of the two keyword sets, according to the processing flow shown in FIG. 7 of the third embodiment.
Assume that keyword sets S ₁ and S ₂ and a distance d are given as search conditions. The appearance position of each keyword set in the document D to be determined is S ₁ ^D = {p ₁ , p ₂ } and S ₂ ^D = {p ₃ , p ₄ , p ₅ }. The relationship between the appearance positions is as shown in FIG. When such a search condition is given, the determination is executed as follows. The following processes (1) to (3) correspond to (1) to (3) in FIG.

(1) In step ST701 in FIG. 7, the first appearance positions p ₁ and p _{3 in} the document D of S ₁ and S ₂ are acquired. Next, in step ST702, it is determined whether or not this appearance position meets the in-neighbor conditions without order designation. In the case of FIG. 8, since the distance between the keyword sets S ₂ and S ₁ is larger than d, it does not meet the in-neighbor condition without order designation. Therefore, the process proceeds to step ST703. In step ST703, the appearance positions of the keyword sets S ₁ and S _{2 in} the document D are determined. Since S ₂ appears earlier, the next appearance position p of the keyword set S _{2 in} step ST705. ₄ is acquired.

(2) This time, the appearance position _{p 1} of _{S 1,} to implement the determination of the ordering without near the condition in step ST702 with respect to _{p 4} acquired in step ST 705. Again, since the distance between S ₁ and S ₂ is greater than d, it does not meet the in-neighbor condition without order designation. Therefore, the process proceeds to step ST703. In step ST 703, towards _{S 1} it is because it appears in the document D earlier than _{S 2,} to get the next occurrence position _{p 2} of _{S 1} at step ST704.

(3) In Step ST702, the in-order neighborhood condition is determined for the appearance positions p ₂ and p ₄ , and the condition is satisfied. Therefore, the process proceeds to Step ST706, and “Applicable” is output.

Next, a search process for three or more keyword sets will be described.
FIG. 9 is a flowchart of processing (step ST303 in FIG. 3) of the search processing execution unit 105 for three or more keyword sets in the third embodiment.
Search processing execution unit 105, the keyword set _S 1 to S _n and the distance d between the keyword set is given, one determines the order of the keyword set _S 1 to S _n in step ST 901. Note that the initial order is arbitrary. Next, in step ST902, the neighborhood condition determination unit 108 determines the in-neighbor conditions with order designation in the order of the keyword set determined in step ST901. The determination process of the in-neighbor condition with order designation is performed in the same manner as in the second embodiment.

  In step ST902, if the document to be determined conforms to the in-neighbor conditions with order designation in the order of the keyword set determined in step ST901 (YES), the process proceeds to step ST905, and “conforms” is output. To do. On the other hand, if it is determined in step ST902 that the specified neighborhood condition is not satisfied (NO), it is determined in step ST903 whether there is a combination of the order of keyword sets that has not yet been determined in step ST902. In step ST903, if there is a combination of the order of keyword sets that have not been determined yet (YES), the order of keyword sets that have not been determined yet is determined in step ST901, and in step ST902, the neighborhood with the order specified by that order. Determine internal conditions. If it is determined in step ST903 that the determination of the in-neighbor condition with specified order has been performed for the combinations of the order of all keyword sets (NO), “not fit” is output in step ST904, and the process ends.

  In the determination process of the in-neighbor condition without order designation, the vicinity condition determination unit 108 may store information on the appearance position of the keyword set acquired from the keyword set matching unit 107 in a memory (not shown). By doing so, it is not necessary to call the keyword set matching unit 107 when acquiring the appearance position of the keyword set once acquired.

Let N be the number of characters in the document and M be the number of keywords included in the keyword set. First, let us consider a case where an in-neighbor condition without order specification between two keyword sets is determined. In the conventional method of Patent Document 1, in which the in-neighbor conditions without order designation are determined by brute force, the amount of calculation required for determining the condition between two keyword sets is O ((MN) ² ). On the other hand, in the case of the determination of the in-neighbor condition without order designation according to the third embodiment, the amount of calculation required for the determination of the in-neighbor condition with order designation between two keywords is O (2MN).

Next, consider a case in which an in-neighbor condition without order specification between K keyword sets is determined. Here, the number of appearance positions in the document of one keyword set is N. In the conventional method of Patent Document 1, the amount of calculation required for determining the in-neighbor condition without ordering between K keyword sets is O (N ^K ).

  The amount of calculation required for determining the in-neighbor condition with the order designation in the second embodiment is O (KN). In the case of an in-neighbor condition without order specification between K keywords, the in-neighbor condition with order designation is determined while changing the combination of the order of the keyword set. The combination of the order of K keyword sets is K! There is (K factorial). Therefore, the amount of calculation required for determining the in-neighbor condition without ordering between the K keyword sets is O (K! KN). This calculation amount is less than the conventional method if K is smaller than N.

  As described above, according to the third embodiment, when all keyword sets appear in the document, the distance between the preceding and following keyword sets is specified as the vicinity condition of the determination process performed by the vicinity condition determination unit 108. Because the neighborhood condition without order designation that is less than the distance is used, the amount of calculation required to determine the condition within neighborhood without order designation can be reduced compared to the conventional case, and the search time in such search processing can be shortened. Can do.

Embodiment 4 FIG.
In the fourth embodiment, conditions such as neighborhoods with keyword group order designation are determined.

  The configuration of the fourth embodiment on the drawing is the same as that of the first embodiment shown in FIG. 1, and will be described with reference to FIG. The document search apparatus according to the fourth embodiment is configured such that the proximity condition determination unit 108 of the document search apparatus according to the first embodiment can determine conditions such as the vicinity of keywords set in the specified order. Further, since the entire flow of the search process is the same as that shown in FIG. 3, the description thereof is omitted here.

Here, a condition such as a neighborhood with keyword group order specification will be described. A keyword set order neighborhood condition is a neighborhood condition that determines whether a keyword set appears in the specified order and whether the distance between the preceding and following keyword sets is a truly specified distance. . Now, keyword set _S 1 to S _n are as specified in this order. Further, it is assumed that the distance d between the keyword sets is designated. At this time, the condition is determined as follows.

FIG. 10 is a flowchart of the process (step ST303 in FIG. 3) of the search process execution unit 105 in the fourth embodiment.
Search processing execution unit 105, the keyword set _S 1 to S _n and the distance d between the keyword set is given, neighborhood condition judging unit 108, the keyword set _S 1 to S _n in step ST 1001, in the document to be determined first and appearance position, the keyword k ₁ to k _n in its appearance position for, acquires from the keyword set collation unit 107. Next, in step ST1002, the neighborhood condition determination unit 108 determines whether the appearance position of the acquired keyword set conforms to the conditions such as neighborhood with order designation. That is, it is determined with respect to _{^{i = 1~n-1, DIST D}} (k i, k i + 1) whether to satisfy all = d.

  In step ST1002, if the appearance position of the acquired keyword set conforms to the conditions such as neighborhood with order designation for the determination target document (YES), the process proceeds to step ST1003, and “conforms” is output. And exit. On the other hand, in step ST1002, if the appearance position of the acquired keyword set does not conform to the conditions such as neighborhood with order designation for the determination target document (NO), the process proceeds to step ST1004.

In Step ST1004, which keyword set appears earlier in the document with respect to the combination of k _i and k _{i + 1} that did not satisfy DIST ^D (k _i , k _{i + 1} ) = d and i is the smallest. Judge whether you are doing. If S _i appears first in the document (YES), it is determined in step ST1005 whether DIST ^D (k _i , k _{i + 1} ) is less than d. When DIST ^D (k _i , k _{i + 1} ) is less than d (YES), in step ST1007, the next appearance position of the keyword set S _{i + 1} and the keyword k _{i + 1} at the appearance position are acquired. If DIST ^D (k _i , k _{i + 1} ) is larger than d (NO), in step ST1006, the appearance position in the next document of the keyword set S _i and the keyword k _i at the appearance position are acquired.

In step ST1004, when S _{i + 1} appears _first in the document (NO), in step ST1007, the next appearance position of the keyword set S _{i + 1} and the keyword k _{i + 1} at the appearance position are acquired. When the next appearance position of the keyword set and the keyword at the appearance position are acquired in step ST1006 or step ST1007, the process returns to step ST1002, and it is determined whether or not the position of the acquired keyword set conforms to the conditions such as neighborhood with specified order judge.

  Although omitted in the flowchart of the processing of FIG. 10, if the appearance position of the keyword set in step ST1001, step ST1006, and step ST1007 cannot be obtained, the order is specified in the determination target document and the conditions such as neighborhood are met. Since there is no appearance position to be output, “not fit” is output and the process ends.

FIG. 11 shows a path for determining a condition such as a neighborhood with keyword group order designation according to the processing flow shown in FIG. 10 of the fourth embodiment.
Now, it is assumed that keyword sets S ₁ , S ₂ , S ₃ are given in this order as search conditions, and the distance designation is d. The appearance position of each keyword set in the document D to be determined is S ₁ ^D = {p ₁ , p ₂ , p ₃ }, S ₂ ^D = {p ₄ , p ₅ , p ₆ }, S ₃ ^D = {P ₇ , p ₈ , p ₉ }. The relationship between their appearance positions is as shown in FIG. When such a search condition is given, the determination is executed as follows. The following processes (1) to (5) correspond to (1) to (5) in FIG.

(1) In step ST1001 of FIG. 11, the first appearance positions p ₁ , p ₄ , and p _{7 in} the document D of S _{1 to} S ₃ are acquired. Next, in step ST1002, it is determined whether or not this appearance position is in conformity with a condition such as a neighborhood with order designation. For Figure 11, keyword set S ₂ is, because of the appearance before the S _1, does not conform to the order specified there near such conditions. Therefore, the process proceeds to step ST1004. In step ST1004, the appearance positions of the keyword sets S ₁ and S _{2 in} the document D are determined. Since S ₂ appears earlier, the next appearance position p of the keyword set S _{2 in} step ST1007. _{Get 5} .

(2) In step ST1002, conditions such as neighborhood with order designation are determined for the appearance positions p ₁ , p ₅ , and p ₇ . This time, since the distance between S ₁ and S ₂ is larger than d, it does not conform to the conditions such as neighborhood with order designation. Therefore, the process proceeds to step ST1004. In step ST 1004, since the direction of _{S 1} is has appeared in the document D earlier than _{S 2,} it is determined whether the distance between the _{S 1} and _{S 2} is less than d in step ST1005. In the case of FIG. 11, since the distance between S ₁ and S ₂ is larger than d, the next appearance position p ₂ of S ₁ is acquired in step ST1006.

(3) In step ST1002, conditions such as neighborhood with order designation are determined for the appearance positions p ₂ , p ₅ , and p ₇ . Since S ₃ appears earlier in document D than S ₂ , the next appearance position p ₈ of S ₃ is acquired in step ST 1007.

(4) In step ST1002, conditions such as neighborhoods with specified order are determined for the appearance positions p ₂ , p ₅ , and p ₈ . Since the distance between S ₂ and S ₃ is less than d, it does not meet the conditions such as neighborhood with specified order. Therefore, the process proceeds to step ST1004. In step ST 1004, since the direction of _{S 2} have appeared in the document before the _{S 3,} the distance between _S 2, _{S 3} at step ST1005 determines whether less than d. Since the distance between S ₂ and S ₃ is less than d, the next appearance position p ₉ of S ₃ is acquired in step ST1007.

(5) In step ST1002, conditions such as neighborhood with order designation are determined for the appearance positions p2, p5, and p9, and in order to meet the conditions, “matched” is output in step ST1003 and the process ends.

Let N be the number of characters in the document and M be the number of keywords included in the keyword set. In the conventional method described in Patent Document 1, the amount of calculation required to determine the condition such as the neighborhood with order specification between K keyword sets is O ((MN) ^K ). On the other hand, in the case of determining the neighborhood condition without order designation in the fourth embodiment, the amount of calculation required for determining the neighborhood condition with order designation between the K keywords is O (KMN) as in the second embodiment. It is.

  As described above, in the document search apparatus according to the fourth embodiment, it is possible to determine the conditions such as the neighborhood with specified order between keyword sets at a higher speed than in the conventional method.

  In addition, here, the method of acquiring the appearance position of the keyword set in order from the top of the document and determining the conditions such as the neighborhood with the order designation is shown, but even if the conditions are determined in order from the end to the top of the document, Similar effects can be obtained.

  As described above, according to the fourth embodiment, as a neighborhood condition in the judgment process performed by the neighborhood condition judgment unit 108, keyword sets appear in the designated order, and the distance between the preceding and following keyword sets is designated. As a result, it is possible to reduce the amount of calculation required to determine the neighborhood condition with specified order and to shorten the search time in such a search process. Can do.

Embodiment 5 FIG.
In the fifth embodiment, a condition such as the neighborhood without specifying the order of keyword sets is determined.

  The configuration of the fifth embodiment on the drawing is the same as that of the first embodiment shown in FIG. 1, and therefore will be described with reference to FIG. The document search apparatus according to the fifth embodiment is configured such that the proximity condition determination unit 108 of the document search apparatus according to the first embodiment can determine conditions such as the vicinity with keyword group order designation. Further, since the entire flow of the search process is the same as that shown in FIG. 3, the description thereof is omitted here.

Here, a condition such as the neighborhood without specifying the order of the keyword set will be described. The keyword set order-free neighborhood condition is a neighborhood condition that determines whether or not the distance between the preceding and following keyword sets is a true specified distance when all keyword sets appear in the document. Now, the distance between the keyword set S ₁ to S _n and keyword set is relative d, unordered specifies vicinity such conditions is determined as follows.

  The flow of processing (step ST303 in FIG. 3) of the search processing execution unit 105 in the fifth embodiment is the same as that configured so as to determine conditions such as neighborhoods with order designation between keyword sets in step ST902 in FIG. Therefore, the description here is omitted.

Now, let N be the number of appearance positions in a document of one keyword set. In the conventional method described in Patent Document 1, the amount of calculation required for determining a condition such as a neighborhood without order designation between K keyword sets is O (N ^K ). On the other hand, in the case of the present embodiment, the amount of calculation required for determining the condition such as the neighborhood without order designation between the K keywords is O (K! KN) as in the third embodiment. This calculation amount is less than the conventional method if K is smaller than N.

  As described above, according to the fifth embodiment, when all keyword sets appear in the document, the distance between the previous and next keyword sets is specified as the vicinity condition of the determination process performed by the vicinity condition determination unit 108. Since the neighborhood condition with no order designation equal to the distance is used, the amount of calculation required to determine the condition with no order designation can be reduced compared to the conventional case, and the search time in such search processing can be shortened. it can.

Embodiment 6 FIG.
In the sixth embodiment, an out-of-neighbor condition with a specified order of keyword sets is determined.

  Since the configuration of the sixth embodiment on the drawing is the same as that of the first embodiment shown in FIG. 1, description will be made with reference to FIG. The document search apparatus according to the sixth embodiment is configured so that the proximity condition determination unit 108 of the document search apparatus according to the first embodiment can determine an out-of-proximity condition with a specified keyword set order. Further, since the entire flow of the search process is the same as that shown in FIG. 3, the description thereof is omitted here.

Here, an out-of-neighbor condition with a specified keyword set order will be described. The out-of-neighbor condition with keyword set order specification is a neighborhood condition for determining whether or not a keyword set appears in the specified order and whether the distance between the preceding and following keyword sets is equal to or greater than the specified distance. Now, keyword set _S 1 to S _n are as specified in this order. Further, it is assumed that the distance d between the keyword sets is designated. At this time, the condition is determined as follows.

FIG. 12 is a flowchart of the process (step ST303 in FIG. 3) of the search process execution unit 105 in the sixth embodiment.
Search processing execution unit 105, the keyword set _S 1 to S _n and the distance d between the keyword set is given, neighborhood condition judging unit 108, the first keyword set _{S 1,} in the document to be determined in step ST1201 And the keyword k ₁ at the appearance position are acquired from the keyword set collating unit 107. Thereafter, the following processing is repeated for i = 1 to n-1.

In step ST1202, the first appearance position in the determination target document of the keyword set S _{i + 1} and the keyword k _{i + 1} at the appearance position are acquired. Next, in step ST1203, it is determined whether or not the appearance positions of the acquired keyword sets S _i and S _{i + 1} match a condition such as neighborhood with order designation. That ^{is, it} is determined whether to satisfy _{DIST D (k i, k i} + 1) of ≧ d.

If the document does not meet the condition in step ST1203 (NO), the next appearance position of S _{i + 1} is acquired in step ST1205, and there is an order designation between the keyword sets acquired in step ST1203. Determine the conditions. If the condition is met in step ST1203 (YES), the first appearance position of the keyword set S _{i + 1} for the next i and the keyword k _{i + 1} at the appearance position are acquired for the next i in step ST1202. This process is repeated for i ranging from 1 to n−1, and if all i satisfy the conditions such as the neighborhood with order designation, “adapted” is output in step ST1204 and the process ends.

  Although not shown in the flowchart of the processing in FIG. 12, if the appearance position of the keyword set cannot be acquired in step ST1201 and step ST1205, the order is specified in the document to be judged, and the appearance conforms to the conditions such as the neighborhood. Since there is no position, it outputs “not fit” and ends.

FIG. 13 shows a determination path for the near-neighbor condition with keyword group order designation according to the processing flow shown in FIG. 12 of the sixth embodiment.
Now, it is assumed that keyword sets S ₁ , S ₂ , S ₃ are given in this order as search conditions, and the distance designation is d. The appearance position of each keyword set in the document D to be determined is S ₁ ^D = {p ₁ , p ₂ , p ₃ }, S ₂ ^D = {p ₄ , p ₅ , p ₆ }, S ₃ ^D = {P ₇ , p ₈ }. The relationship between the appearance positions is as shown in FIG. When such a search condition is given, the determination is executed as follows. The following processes (1) to (5) correspond to (1) to (5) in FIG.

(1) first occurrence _{p 1} in the document D in _{S 1} in step ST1201 of FIG. 12 is obtained. Next, to obtain the first occurrence _{p 4} in a document _{S 2} in step ST 1202. In step ST1203, it is determined whether or not the appearance position is in conformity with an out-of-neighbor condition with a specified order. In the case of FIG. 13, the keyword set S ₂ appears before S ₁ , so that it does not meet the out-of-neighbor condition with order designation. Therefore, the process proceeds to step ST1205. In step ST1205, acquires the next occurrence position p5 of the keyword set _{S 2.}

(2) In step ST1203, an out-of-neighbor condition with an order is determined for the appearance positions p ₁ and p ₅ . This time, since the distance between S ₁ and S ₂ is smaller than d, it is not suitable for the near-outside condition with order designation. Therefore, the process proceeds to step ST1205, acquires the next occurrence position _{p 6} of keyword set _{S 2.}

(3) In step ST1203, an out-of-neighbor condition is specified with respect to the appearance positions p ₁ and p ₆ . Now because they comply with the vicinity outside the conditions there specified order, to obtain the first occurrence p ₇ in a document keyword set S3 in step ST 1202.

(4) In step ST1203, an out-of-neighbor condition with an order is determined for the appearance positions p ₆ and p ₇ . Since the keyword set S ₃ appears before S ₂ , it does not meet the out-of-neighbor condition with the order specified. Therefore, to get the next occurrence position _{p 8} of keyword set _{S 3} at step ST1205.

(5) In step ST1203, an out-of-neighbor condition with order designation is determined for the appearance positions p ₆ and p ₈ . Here, since the conditions are met and all keyword sets have been determined, “matched” is output in step ST1206, and the process ends.

Let N be the number of characters in the document and M be the number of keywords included in the keyword set. In the conventional method for determining an out-of-neighbor condition with an order specification by brute force, the amount of calculation required for determining an out-of-neighbor condition with an order specification between K keyword sets is O ((MN) ^K ). On the other hand, in the case of determining the out-of-neighbor condition with the order designation according to the present embodiment, it is O (KMN) as in the second embodiment.

  Here, the method of acquiring the appearance position of the keyword set in order from the beginning of the document and determining the out-of-neighbor condition with the order designation has been shown, but even if the condition is determined in order from the end to the beginning of the document, Similar effects can be obtained.

  As described above, according to the sixth embodiment, as a neighborhood condition in the judgment process performed by the neighborhood condition judgment unit 108, keyword sets appear in the designated order, and a distance between preceding and following keyword sets is designated. Therefore, it is possible to reduce the amount of calculation required for determining the out-of-neighbor condition with specified order, and to shorten the search time in such search processing. be able to.

Embodiment 7 FIG.
In the seventh embodiment, an out-of-neighbor condition without specifying a keyword set order is determined.

  Since the configuration of the seventh embodiment on the drawing is the same as that of the first embodiment shown in FIG. 1, description will be made with reference to FIG. The document search apparatus according to the fifth embodiment is configured such that the neighborhood condition determination unit 108 of the document search apparatus according to the first embodiment can determine an out-of-proximity condition without specifying a keyword set order. Further, since the entire flow of the search process is the same as that shown in FIG. 3, the description thereof is omitted here.

Here, an out-of-neighbor condition without specifying the order of keyword sets will be described. The out-of-neighbor condition without specifying the order of keyword sets is a neighborhood condition for determining whether or not the distance between preceding and following keyword sets is greater than or equal to a specified distance when all keyword sets appear in the document. Now, the distance between the keyword set S ₁ to S _n and keyword set is relative d, unordered designated neighborhood outside conditions are determined as follows.

  In the seventh embodiment, the flow of the processing of the search processing execution unit 105 in FIG. 1 (step ST303 in FIG. 3) is configured to determine an out-of-order specified neighborhood condition between keyword sets in step ST902 in FIG. Since it is equivalent to a thing, description here is abbreviate | omitted.

Let N be the number of appearance positions in a document of one keyword set. In the conventional method described in Patent Document 1, in which the out-of-order unspecified out-of-order condition is determined by brute force of the keyword appearance position, the amount of calculation required for determining the out-of-order unspecified out-of-order condition between K keyword sets Becomes O (N ^K ). On the other hand, in the case of the present embodiment, the amount of calculation required for determining an out-of-order condition between the K keywords without specifying the order is O (K! KN) as in the third embodiment. This calculation amount is less than the conventional method if K is smaller than N.

  As described above, according to the seventh embodiment, when all the keyword sets appear in the document, the distance between the preceding and following keyword sets is specified as the vicinity condition of the determination process performed by the vicinity condition determining unit 108. Compared to the conventional method, the calculation amount required to determine the out-of-order non-neighbor condition can be reduced and the search time in such search processing can be shortened. Can do.

Embodiment 8 FIG.
In the eighth embodiment, the context condition of the keyword set is determined.

FIG. 14 is a block diagram showing the document search apparatus according to the eighth embodiment.
The illustrated document search apparatus is obtained by replacing the neighborhood condition determination unit 108 of the search processing execution unit 105 shown in FIG. 1 with a context condition determination unit 112. Since the other configuration is the same as that of FIG. 1, the same reference numerals are given to corresponding portions, and description thereof is omitted. The context condition determination unit 112 is provided in the search processing execution unit 105a and has a function of determining whether or not all keyword sets appear in the same document constituent unit when a plurality of keyword sets are given. Is.

  The entire flow in the search process is equivalent to the process of determining whether or not the neighborhood condition in step ST303 in FIG. 3 is matched with the context condition determination process, and thus description thereof is omitted here. .

Here, the context conditions of the keyword set will be described. The context condition of the keyword set is a condition for determining whether or not all of the keyword sets appear in the structural unit of the same document when two or more keyword sets are given. Here, the structural unit of a document is an element constituting the document such as a sentence, paragraph, chapter, section, page in the document. And keyword set _S 1 ~S _n is specified. At this time, the condition is determined as follows. Here, the constituent unit of the document D is comp ^D , the head position of the constituent unit is STR ^D (comp ^D ), and the tail position is END ^D (comp ^D ). It is assumed that the head position and the tail position of the structural unit are each represented by the number of characters from the head of the document.

FIG. 15 is a flowchart of processing of the search processing execution unit 105a in the eighth embodiment.
Search processing execution unit 105a, given keyword set S ₁ to S _n, firstly, acquires information of the start position and end positions of all the document constituent units of the document to be determined in step ST1501. Next, contextual condition judgment unit 112, the keyword set _S 1 to S _n in step ST 1502, the first occurrence in a document determination target, the keyword _k 1 to k _n in its appearance position, keyword set Obtained from the verification unit 107.

In step ST1503, it is determined whether or not the appearance position of the acquired keyword set matches the context condition. That is, STR ^D (comp ^D _j ) ≦ STR ^D (k _i ), END ^D (k _i ) with respect to all document constituent units comp ^D _j (j = 1 to m) where i = 1 to n. It is determined whether or not ≦ END ^D (comp ^D _j ) is satisfied. If the appearance position of the acquired keyword set conforms to the context condition for the document to be determined in step ST1503 (YES), the process proceeds to step ST1507, “conforms” is output, and the process ends. On the other hand, if the appearance position of the acquired keyword set does not match the context condition for the determination target document in step ST1503 (NO), the appearance position of the end of the keyword in the document is determined in step ST1504. rearmost k _i is determined or not span construction unit of two or more documents. That is, it is determined whether STR ^D (k _i ) ≦ END ^D (comp ^D _j ) ≦ END ^D (k _i ) for a certain _j .

In step ST1504, if it was over (YES), obtains the next occurrence position of _{S i} in step ST 1505, the process proceeds to step ST1506. In Step ST1504, _{if k i} is not across the structural unit of two or more documents (NO), there moves to step ST1506. In step ST1506, the next appearance position of all keyword sets that does not fall within the same document constituent unit as the last k _{i in} the last document of the keyword is acquired. If the next appearance position of a keyword set is acquired by step ST1506, it will return to step ST1503 and it will be determined whether the appearance position of the acquired keyword set is suitable for context conditions.

  Although omitted in the flowchart of the processing in FIG. 15, if the appearance position of the keyword set cannot be obtained in step ST1502, step ST1505, and step ST1506, the appearance position that matches the context condition in the determination target document. Since it means that there is no, output “not fit” and exit.

  In step ST1501, the start position and the end position of the document unit are acquired. However, the information on the start and end positions of the document unit is automatically acquired from the document when the document is registered in the database 110. It may be extracted and recorded in the database 110, or a document to be determined may be scanned and acquired at the time of search. In either case, when the structural unit of the document is a sentence, it can be automatically extracted from the document by setting the position of the character immediately after the phrase as the head position and the position of the next phrase as the end position. In the case where the structural unit of the document is a paragraph, a line feed character may be used as a delimiter for the structural unit instead of a punctuation mark.

FIG. 16 shows a determination path for keyword context context conditions according to the processing flow shown in FIG. 15 of the eighth embodiment.
Assume that keyword sets S ₁ , S ₂ , and S ₃ are given as search conditions. The appearance position of each keyword set in the document D to be determined is S ₁ ^D = {p ₁ , p ₂ , p ₃ }, S ₂ ^D = {p ₄ , p ₅ , p ₆ }, S ₃ ^D = {P ₇ , p ₈ , p ₉ }. The relationship between their appearance positions is as shown in FIG. When such a search condition is given, the determination is executed as follows. The following processes (1) to (3) correspond to (1) to (3) in FIG.

(1) In step ST1501 in FIG. 15, information on the start position and end position of all the structural units of the document D is acquired. Next, in step ST1502, the first appearance positions p ₁ , p ₄ , and p _{7 in} the document D of S _{1 to} S ₃ are acquired. In step ST1503, it is determined whether or not the acquired appearance position meets the context condition. In the case of FIG. 16, the keyword sets S ₁ and S ₃ are included in the structural unit 1602 and S ₂ is included in the structural unit 1601, and thus do not meet the context condition. Therefore, the process proceeds to step ST1506 via step ST1504. At step ST1506, _{since p 1} is in the rearmost position in the document, to get the next occurrence position _{p 5} of _{S 2} that are not included in the constituent unit 1602 includes keyword set _{S 1.}

(2) In step ST1503, the context condition is determined for the appearance positions p ₁ , p ₅ , and p ₇ . This time, S ₁ and S ₃ are included in the structural unit 1602 and S ₂ is included in the structural unit 1604, and thus do not meet the context condition. In Step ST1506, _{since p 5} is positioned at the rearmost in the document, to get the next occurrence position _p 2, _{p 8} of keyword set _{S 1} and _{S 2} which are not included in the constituent unit 1604.

(3) In step ST1503, context conditions are determined for the appearance positions p ₂ , p ₅ , and p ₈ . This time, S ₁ , S ₂ , and S ₃ are included in the same structural unit 1604, and thus satisfy the context condition. Therefore, “applicable” is output in step ST1507 and the process ends.

Let N be the number of characters in the document and M be the number of keywords included in the keyword set. In the conventional method described in Patent Document 1, the amount of calculation required to determine the context condition between K keyword sets is O ((MN) ^K ). On the other hand, in the case of the determination of the context condition of the present embodiment, it is O (KMN) as in the second embodiment.

  Here, the method of determining the context condition by acquiring the appearance position of the keyword set in order from the beginning of the document has been shown, but the same effect can be obtained by determining the condition in order from the end of the document toward the beginning. can get.

  As described above, according to the eighth embodiment, the keyword set matching unit 107 that acquires the appearance positions of a plurality of keyword sets in a document in the order of the appearance positions of the keyword sets and the keyword set matching unit 107 are acquired. A context condition determination unit 112 that determines whether or not a plurality of keyword sets meet a context condition indicating that they appear in a constituent unit of the same document; The context condition determination unit 112 alternately executes the context condition determination process and outputs the determination result as a search result when the context condition is determined to be true. Thus, it is possible to reduce the amount of calculation required for determining the context condition, and to shorten the search time in such a search process.

Embodiment 9 FIG.
In the ninth embodiment, the range condition of the keyword set is determined.

FIG. 17 is a configuration diagram illustrating the document search apparatus according to the ninth embodiment.
The illustrated document search apparatus is obtained by replacing the proximity condition determination unit 108 of the search processing execution unit 105 shown in FIG. 1 with a range condition determination unit 113. Since the other configuration is the same as that of FIG. 1, the same reference numerals are given to corresponding portions, and description thereof is omitted. The range condition determination unit 113 is provided in the search processing execution unit 105b. When a plurality of keyword sets and one document range are given, the range condition determination unit 113 determines whether or not all the keyword sets appear in the same document range. It has the function to do.

  The overall flow in the search process is the same as that obtained by replacing the determination process of whether or not the neighborhood condition in step ST303 in FIG. .

Here, the range condition of the keyword set will be described. The range condition of the keyword set is a condition for determining whether or not all keyword sets appear in the same document range when two or more keyword sets and one document range are given. Here, the document range is a range that constitutes a certain unit in the document, such as a summary, a preface, a postscript, and a body. And keyword set _S 1 ~S _n and document range is specified. At this time, the range condition is determined as follows. Here, it is assumed that the range in the document D is range ^D , the start position of the document range is STR ^D (range ^D ), and the end position is END ^D (range ^D ).

FIG. 18 is a flowchart of the process of the search process execution unit 105b in the ninth embodiment.
Search processing execution unit 105b, when the keyword set _S 1 to S _n and the document range range ^D is given, first, to acquire the information of the start position and end position of the range of the document to be determined in step ST1801. Next, the range condition determining unit 113, the keyword set _S 1 to S _n in step ST1802, the first occurrence in a document determination target, the keyword _k 1 to k _n in its appearance position, keyword set Obtained from the verification unit 107. In step ST1803, it is determined whether the appearance position of the acquired keyword set conforms to the range condition. That is, for all i = 1 to n and the document range range ^D , whether STR ^D (range ^D ) ≦ STR ^D (k _i ), END ^D (k _i ) ≦ END ^D (range ^D ) is satisfied. Determine whether.

  If the appearance position of the acquired keyword set conforms to the range condition for the document to be determined in step ST1803 (YES), the process proceeds to step ST1805, “conforms” is output, and the process ends. On the other hand, if the appearance position of the acquired keyword set does not match the range condition for the determination target document in step ST1803 (NO), in step ST1804, all keyword sets that do not satisfy the range condition are detected. Get the next occurrence position. When the next appearance position of the keyword set is acquired in step ST1804, it is determined in step ST1803 whether or not the appearance position of the acquired keyword set meets the range condition.

Although omitted in the flowchart of the processing in FIG. 18, if the appearance position of the keyword set cannot be acquired in steps ST1802 and ST1804, there is no appearance position that meets the range condition in the document to be determined. So, output “not fit” and exit. Also, even if there is a keyword to be in step ^{ST1803 END D (range D) <} END D (k i), to terminate with a "nonconforming".

  In step ST1801 of FIG. 18, the start position and end position of the document range are acquired, but information on the start and end positions of the document range is automatically extracted from the document when the document is registered in the database 110. It may be recorded in the database 110, or it may be obtained by scanning a document to be determined at the time of search. In either case, it is necessary to determine a document range extraction rule in advance.

FIG. 19 shows a determination path for keyword set range conditions according to the processing flow shown in FIG. 18 of the ninth embodiment.
Assume that keyword sets S ₁ , S ₂ , and S ₃ are given as search conditions. The appearance position of each keyword set in the document D to be determined is S ₁ ^D = {p ₁ , p ₂ , p ₃ }, S ₂ ^D = {p ₄ , p ₅ , p ₆ }, S ₃ ^D = {P ₇ , p ₈ , p ₉ }. The relationship between their appearance positions is as shown in FIG. When such a search condition is given, the determination is executed as follows. The following processes (1) to (3) correspond to (1) to (3) in FIG.

(1) In step ST1801 in FIG. 18, information on the start position and end position of the designated range of the document D is acquired. Next, in step ST1802, the first appearance positions p ₁ , p ₄ , and p _{7 in} the document D of S _{1 to} S ₃ are acquired. In step ST1803, it is determined whether or not the acquired appearance position meets the range condition. In the case of FIG. 19, any keyword set is not included in the specified range, and therefore does not meet the range condition. Therefore, in step _ST1804, acquires the S _1, S 2, the next occurrence position of _{S 3 p 2, p 5,} p 8.

(2) In step ST1803, the range condition is determined for the appearance positions p ₂ , p ₅ , and p ₈ . This time, since S ₁ and S ₂ are not included in the designated document range, the range condition is not met. In step ST1804, the next appearance positions p ₃ and p ₆ of S ₁ and S ₂ are acquired.

(3) In step ST1803, context conditions are determined for the appearance positions p ₃ , p ₆ , and p ₈ . This time, S ₁ , S ₂ , and S ₃ are included in the designated document range, and therefore meet the range condition. Therefore, “applicable” is output in step ST1805, and the process ends.

Let N be the number of characters in the document and M be the number of keywords included in the keyword set. In the conventional method described in Patent Document 1, the amount of calculation required to determine the range condition between K keyword sets is O (M ^K KN). On the other hand, in the case of the determination of the context condition of the present embodiment, it is O (KMN) as in the second embodiment.

  As described above, in the document search apparatus according to the ninth embodiment, it is possible to determine the range condition between keyword sets at a higher speed than in the conventional method.

  Here, the method of acquiring the appearance position of the keyword set in order from the beginning of the document and determining the range condition has been shown, but the same effect can be obtained by determining the condition in order from the end of the document toward the beginning. can get.

  As described above, according to the ninth embodiment, the keyword set matching unit 107 that acquires the appearance positions of a plurality of keyword sets in a document in the order of the appearance positions of the keyword sets and the keyword set matching unit 107 are acquired. A range condition determination unit 113 that determines whether or not a range condition indicating that a plurality of keyword sets appear in a specific document range is satisfied, and an appearance position acquisition process in the keyword set matching unit 107; Since it includes a search processing execution unit that alternately executes the range condition determination processing in the range condition determination unit 113 and outputs the determination result as a search result when the range condition is determined to be true, The amount of calculation required for determining the range condition can be reduced, and the search time in such a search process can be shortened.

Embodiment 10 FIG.
In the tenth embodiment, a composite condition of a keyword set is determined.

FIG. 20 is a configuration diagram illustrating the document search apparatus according to the tenth embodiment.
The illustrated document search apparatus is obtained by replacing the proximity condition determination unit 108 of the search processing execution unit 105 shown in FIG. Since the other configuration is the same as that of FIG. 1, the same reference numerals are given to corresponding portions, and description thereof is omitted. The composite condition determination unit 114 is provided in the search processing execution unit 105c, and when two or more keyword sets are given, all the keyword sets are in the vicinity of any one of the second to seventh embodiments. It has a function of determining whether or not a condition, a context condition of the eighth embodiment, a range condition of the ninth embodiment, and a logical condition obtained by combining them with a logical operation are satisfied.

  Since the entire flow in the search process is equivalent to the process of determining whether or not the neighborhood condition in step ST303 in FIG. 3 is adapted to the complex condition determination process, the description here is omitted. .

  Here, the compound condition of the keyword set will be described. The compound condition of a keyword set is that when two or more keyword sets are given, all of the keyword sets are neighborhood conditions, context conditions, range conditions, and logical operations (AND / OR / NOT etc.). This is a condition for determining whether or not the combined logical condition is met. When the composite condition includes determination of a range condition, a single document range is also given as a search condition.

FIG. 21 is a flowchart of processing of the search processing execution unit 105c in the tenth embodiment.
Now, it is assumed that a compound condition obtained by combining a condition within a neighborhood with a specified keyword set order, a context condition, and a range condition with a logical condition AND is provided as a search condition. When the keyword set, the distance between the keyword sets, and one document range are designated, the search processing execution unit 105c determines, in step ST2101, all the structural units of the document to be determined by the keyword set matching unit 107. , Get the start and end position of the specified document range. Next, in step ST2102, the keyword set matching unit 107 acquires the first appearance position of all keyword sets. In step ST2103, the composite condition determination unit 114 determines whether the appearance position of the acquired keyword set meets the range condition.

  In step ST2103, when the range condition is met (YES), the process proceeds to step ST2104, and it is determined whether the appearance position of the acquired keyword set matches the context condition. If the context condition is satisfied in step ST2104 (YES), it is determined in step ST2105 whether the appearance position of the acquired keyword set is compatible with the neighborhood condition. If it is determined in step ST2105 that the neighborhood condition is met, “matched” is output in step ST2106, and the process ends.

  On the other hand, when the condition is not satisfied in each condition determination in step ST2103, step ST2104, and step ST2105 (NO), the next appearance position of the keyword set is acquired in step ST2107. In step ST2107, the keyword set from which the appearance position is acquired differs depending on conditions that are not matched in the determination. If the range condition is not met, the appearance position of the keyword set is acquired according to the ninth embodiment. If the context condition is not met, the appearance position of the keyword set is acquired according to the eighth embodiment. If the neighborhood condition is not met, the appearance position of the keyword set is acquired according to any one of the second to seventh embodiments. If the appearance position of a keyword set is acquired in step ST2107, it will return to step ST2103 and will determine a range condition again.

  Although omitted in the flowchart of the processing of FIG. 21, when the appearance position of the keyword set cannot be acquired in steps ST2102 and ST2107, it means that there is no appearance position that satisfies the range condition in the document to be determined. , "Not fit" is output and the process ends. Also, in step ST2103, if the appearance position of the keyword set in the document is behind the document range, “not fit” is output and the process ends.

FIG. 22 shows a determination path for a keyword set compound condition according to the processing flow shown in FIG. 21 of the tenth embodiment.
Assume that keyword sets S ₁ , S ₂ , and S ₃ are given as search conditions. The appearance position of each keyword set in the document D to be determined is represented by S ₁ ^D = {p ₁ , p ₂ , p ₃ }, S ₂ ^D = {p ₄ , p ₅ , p _6, p ₇ }, Let S ₃ ^D = {p ₈ , p ₉ , p ₁₀ }. The relationship between the appearance positions is as shown in FIG. Note that specific operations that satisfy the compound conditions of the keyword sets S ₁ , S ₂ , and S ₃ are the range conditions in the ninth embodiment, the context conditions in the eighth embodiment, and the second to seventh embodiments. Since this is an operation that satisfies all the conditions of the proximity condition by any of the above, description here is omitted.

  The region in the document that meets the range condition is local compared to the neighborhood condition or context condition. Therefore, the complex condition combined with the logical condition AND can be narrowed down the area that matches the condition earlier by determining from the range condition. In the case of a compound condition combined with a logical condition OR, a determination result can be acquired earlier by determining a wide area that meets the condition first.

  Here, the method of determining the compound condition by acquiring the appearance position of the keyword set in order from the beginning of the document has been shown, but the same effect can be obtained by determining the condition in order from the end to the beginning of the document. can get.

  As described above, according to the tenth embodiment, the keyword set matching unit 107 that acquires the appearance positions of a plurality of keyword sets in a document in the order of the appearance positions of the keyword sets and the keyword set matching unit 107 are acquired. A composite that combines a predetermined neighborhood condition, a context condition indicating that it appears in a structural unit of the same document, a range condition indicating that it appears in a specific document range, and a logical condition of these conditions A composite condition determination unit 114 that determines whether or not a condition is satisfied, and alternately performs an appearance position acquisition process in the keyword set matching unit 107 and a composite condition determination process in the composite condition determination unit 114, Since the search processing execution unit 105c that outputs the result of the determination as a search result when the condition is determined to be true is provided, compared to the conventional case, there is a more It is possible to reduce the calculation amount required for the determination of the conditions, it is possible to shorten the search time in such a search process.

Embodiment 11 FIG.
In the eleventh embodiment, in the first to tenth embodiments, the keyword set collating unit 107 acquires the appearance position of the keyword set in the document at high speed.

  Since the configuration of the eleventh embodiment on the drawing is the same as that of any of the first to eleventh embodiments, the description of the configuration other than the illustration and the keyword set matching unit 107 is omitted. The keyword set matching unit 107 of the eleventh embodiment acquires only the next appearance position of the keyword at the appearance position output immediately before as a keyword set, and compares the acquired appearance position with the appearance position of another keyword. The smallest one is output as the next appearance position of the keyword set.

Next, the operation of the keyword set matching unit 107 in Embodiment 11 will be described.
When there is a first appearance position acquisition request in the keyword set document, the keyword set matching unit 107 acquires the first appearance position from the keyword matching unit 109 for all keywords in the keyword set. The keyword matching unit 109 outputs the keyword appearance positions one by one in order from the top of the document each time a keyword appearance position acquisition request is made. For example, in the example of FIG. 2, in the keyword set {keyword, appearance}, 1 that is the first appearance position of “keyword” and 9 that is the first appearance position of “appearance” are output.

  If the appearance position of any keyword in the keyword set cannot be acquired, the keyword set matching unit 107 outputs “no hit” and ends. On the other hand, if the appearance position of one or more keywords in the keyword set can be acquired, the one that appears earliest in the document among the obtained appearance positions is output and the process ends. In the example of FIG. 2, since the first appearance position of each keyword is [1, 9], 1 is output as the appearance position of the keyword set. At this time, if necessary, the keyword at the appearance position and the keyword length may also be output. Then, the keyword set matching unit 107 outputs the keyword at the appearance position output at the end (“keyword” in the example of FIG. 2) and the information on the appearance positions acquired for all the keywords ([1, 9 in the example of FIG. 2). ]) Is stored internally.

FIG. 23 is a flow of the process for acquiring the appearance position of the keyword set for the second and subsequent times.
When there is a request for obtaining the appearance position of the keyword set for the second and subsequent times, the keyword set collating unit 107 obtains the next appearance position from the keyword collating unit 109 for the keyword at the appearance position output last time in step ST2301. . In step ST2302, it is determined whether or not the appearance position of the keyword has been acquired.

  In step ST2302, if it has been acquired (YES), in step ST2303, it is the most in-document among the keyword appearance position acquired in step ST2301 and the keyword appearance position acquired in the previous keyword set matching process. Output what appears earlier. For example, in the example of FIG. 2, since the keyword at the appearance position output last time is “keyword”, in step ST2302, 18 that is the next appearance position of “keyword” is acquired. Next, in step ST2303, the appearance position 9 of the other keyword “appearance” is compared, and 9 is output as the next appearance position of the keyword set {keyword, appearance} as the smallest one.

  If the appearance position cannot be acquired in step ST2302, the process proceeds to step ST2304. In step ST2304, it is determined whether or not there is information on the appearance position of the keyword acquired in the keyword set matching process before the previous time. If there is information on the appearance position (YES), the process proceeds to step ST2303. When there is no appearance position information in step ST2304 (NO), in step ST2305, “no hit” is output and the process ends. For example, in the example of FIG. 2, when 35 is output as the appearance position of the keyword set and the next appearance position is requested, the next appearance position of the “keyword” at the appearance position 35 cannot be acquired. Is output as the next appearance position of the keyword set. After that, in step ST2304, since there is no candidate for the next appearance position, “no hit” is output.

  Although the details of the processing of the keyword matching unit 109 were not mentioned, the keyword matching unit 109 returns information indicating what character the keyword appears from the top of the document in response to the keyword appearance position acquisition request. As long as it is possible, it may be realized by any method. That is, information on a set of characters and character strings and appearance positions in the document may be recorded as an index in the storage device, or the document may be directly scanned when an appearance position acquisition request is made, You may acquire the appearance position of a keyword.

  As described above, since the keyword set collating unit 107 is configured, the appearance position of the keyword set can be output only by acquiring the appearance position for one keyword each time there is a request for obtaining the appearance position of the keyword set. . Thus, the appearance position of the keyword set can be acquired by executing the minimum necessary keyword matching without acquiring all the appearance positions for all the keywords in the keyword set.

  Here, the method for acquiring the appearance position of the keyword set in order from the beginning of the document has been described. However, depending on the condition determination method in the first to tenth embodiments, the appearance position of the keyword set in order from the end of the document. It is also possible to obtain. In that case, the keyword matching unit 109 is configured to output one by one in order from the end of the document. In step ST2303, it is only necessary to output the most recent appearance in the document among the appearance positions of the keywords.

  As described above, according to the eleventh embodiment, the keyword set matching unit 107 acquires only the next appearance position of the keyword at the appearance position output immediately before as the keyword set, and the acquired appearance position The appearance position of the keyword set in the document can be acquired at a high speed because the smallest position is output as the next occurrence position of the keyword set.

It is a block diagram which shows the document search device by Embodiment 1 of this invention. It is explanatory drawing which shows the relationship between a document and a keyword. 3 is a flowchart of search processing in the document search apparatus according to the first embodiment. 4 is a flowchart of processing of a search processing execution unit in the document search device according to the first embodiment. 10 is a flowchart of processing of a search processing execution unit in the second embodiment. FIG. 10 is an explanatory diagram showing a determination path for a neighborhood condition with a specified keyword set order according to the second embodiment; 14 is a flowchart of processing of a search processing execution unit when there are two keyword sets in the third embodiment. FIG. 10 is an explanatory diagram illustrating a determination path for a condition in the neighborhood without specifying the order of two keyword sets according to the third embodiment. 14 is a flowchart of processing of a search processing execution unit for three or more keyword sets in the third embodiment. 15 is a flowchart of processing of a search processing execution unit in the fourth embodiment. FIG. 20 is an explanatory diagram showing a determination path for conditions such as neighborhoods with specified keyword set order according to the fourth embodiment; 18 is a flowchart of processing of a search processing execution unit in the sixth embodiment. FIG. 20 is an explanatory diagram illustrating a determination path for an out-of-neighbor condition with a specified order of keyword sets according to the sixth embodiment. FIG. 10 is a configuration diagram illustrating a document search device according to an eighth embodiment. 20 is a flowchart of processing of a search processing execution unit in the eighth embodiment. FIG. 20 is an explanatory diagram illustrating a determination path for a context condition of a keyword set according to an eighth embodiment. FIG. 20 is a configuration diagram illustrating a document search device according to a ninth embodiment. 20 is a flowchart of processing of a search processing execution unit in the ninth embodiment. FIG. 20 is an explanatory diagram illustrating a determination path for a keyword set range condition according to the ninth embodiment; FIG. 20 is a configuration diagram illustrating a document search device according to an embodiment 10; 22 is a flowchart of processing of a search processing execution unit in the tenth embodiment. FIG. 38 is an explanatory diagram illustrating a determination path for a composite condition of a keyword set according to the tenth embodiment. This is a flow of processing for acquiring the appearance position of the keyword set for the second and subsequent times in the eleventh embodiment.

Explanation of symbols

  101 Search conditions, 105, 105a, 105b, 105c Search processing execution unit, 107 Keyword set collation unit, 108 Neighborhood condition determination unit, 110 Database, 112 Context condition determination unit, 113 Range condition determination unit, 114 Compound condition determination unit

Claims (11)

A keyword set matching unit that obtains the appearance positions of a plurality of keyword sets in a document in the order of the appearance positions of the keyword sets, and the appearance positions of the plurality of keyword sets obtained by the keyword set matching unit satisfy a predetermined neighborhood condition. A neighborhood condition determination unit that determines whether or not to satisfy,
A search that alternately executes an appearance position acquisition process in the keyword set matching unit and a neighborhood condition determination process in the neighborhood condition determination unit, and outputs the determination result as a search result when the neighborhood condition is determined to be true. A document search apparatus comprising a processing execution unit.   The neighborhood condition is an in-neighbor condition with an order specification in which a keyword set appears in a specified order and a distance between preceding and following keyword sets is equal to or less than a specified distance. Document retrieval device.   2. The document according to claim 1, wherein the neighborhood condition is an unordered neighborhood condition in which a distance between preceding and following keyword sets is equal to or less than a specified distance when all keyword sets appear in the document. Search device.   2. The document according to claim 1, wherein the neighborhood condition is a neighborhood condition such that a keyword set appears in a specified order and a distance between preceding and following keyword sets is equal to a specified distance. Search device.   2. The document search according to claim 1, wherein when all keyword sets appear in the document, the neighborhood condition is a condition such as an unordered neighborhood condition in which a distance between preceding and following keyword sets is equal to a specified distance. apparatus.   The neighborhood condition is an out-of-neighbor condition with an order specification in which a keyword set appears in a specified order and a distance between preceding and following keyword sets is equal to or greater than a specified distance. Document retrieval device.   2. The document according to claim 1, wherein the neighborhood condition is an out-of-order non-neighbor condition where a distance between preceding and following keyword sets is equal to or greater than a specified distance when all keyword sets appear in the document. Search device. A keyword set collation unit that acquires the appearance positions of a plurality of keyword sets in a document in the order of the appearance positions of the keyword sets, and a plurality of keyword sets obtained by the keyword set collation unit appear in a constituent unit of the same document A context condition determination unit that determines whether or not a context condition indicating that
Retrieval in which appearance position acquisition processing in the keyword set matching unit and context condition determination processing in the context condition determination unit are alternately executed, and when the context condition is determined to be true, the determination result is output as a search result A document search apparatus comprising a processing execution unit. A keyword set matching unit that obtains the appearance positions of a plurality of keyword sets in a document in the order of the appearance positions of the keyword sets, and a plurality of keyword sets obtained by the keyword set matching unit appear in a specific document range. A range condition determination unit that determines whether or not a range condition indicating that,
Retrieval in which appearance position acquisition processing in the keyword set matching unit and range condition determination processing in the range condition determination unit are alternately executed, and when the range condition is determined to be true, the determination result is output as a search result A document search apparatus comprising a processing execution unit. A keyword set matching unit that obtains appearance positions of a plurality of keyword sets in a document in the order of the appearance positions of the keyword sets, and a plurality of keyword sets obtained by the keyword set matching unit include a predetermined neighborhood condition, A composite condition determining unit that determines whether or not a composite condition that combines a context condition that indicates occurrence in a structural unit, a range condition that indicates that it appears in a specific document range, and a logical condition of these conditions is combined; Have
A search that alternately executes an appearance position acquisition process in the keyword set matching unit and a complex condition determination process in the complex condition determination unit, and outputs the determination result as a search result when the complex condition is determined to be true. A document search apparatus comprising a processing execution unit.   The keyword set matching unit acquires only the next appearance position of the keyword at the appearance position output immediately before as a keyword set, compares the obtained appearance position with the appearance position of another keyword, and determines the smallest one. The document search apparatus according to claim 1, wherein the document search apparatus outputs the next occurrence position of the keyword set.

Images