JP4037001B2 - Database creation device and database search device - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a database creation device that automatically creates a database index, and a database search device that searches a created database while narrowing down categories.
[0002]
[Prior art]
FIG. 13 is a block diagram showing a conventional similarity search apparatus. This search device includes a database 17 that stores data represented by attribute-attribute value pairs, and a first index that narrows down the amount of data from the data in the database 17 using a first index before searching for similar data. A second search unit 15 calculates a similarity range from the similarity range between attribute values and the importance of the attribute, and performs a similarity search on the data searched by the first search unit 15 based on the second index. The search unit 16, the inference processing unit 14 including a first index changing unit that changes the first index according to the similarity range, and the similarity value of the first index are set, and based on the similarity value The first index generation unit 18 that determines the level of the first index, the input device 11, the output device 12, and the input / output control unit 13 are configured.
[0003]
This retrieval apparatus sets the similarity value of the first index, determines the level of the first index based on the similarity value, and calculates the similarity based on the similarity range between the attribute values and the second index Then, the similarity search is performed by changing the first index according to the calculated similarity range. Such a search device is disclosed in, for example, Japanese Patent Application Laid-Open No. 6-177602.
[0004]
FIG. 14 is a block diagram showing a conventional information retrieval apparatus. This search device includes an arithmetic unit 19 for inner producting an index keyword matrix from the keyword memory 24 and a search vector modified in accordance with the degree of correlation, a partial linear unit 20 for converting the result into partial linearity, and a partial linearity. A second calculator 21 that multiplies the output vector of the calculator 20 and the keyword matrix X, a normalizer 22 that normalizes each element of the product result to 0 and 1, and before the feedback to the calculator 19 once. A controller 23 for comparing with the output vector of the normalizer 22 of the device, a reader 26 for reading out the desired data from the database 25 based on the corresponding address of the position of the output vector of the modified partial linearizer 20, and the desired data And a display 27 for displaying.
[0005]
Specifically, this search device has the configuration shown in FIG. 15, and stores the keyword input from the keyword input unit 28 in the storage unit 33, converts it into a numerical vector by the computing unit 24, and stores the keyword correlation table 35. Referring to the keyword, a more relevant keyword corresponding to the degree of correlation is found, and the converter 29 converts the first input keyword into the found keyword. Then, the search device newly searches the database 36 using the keyword as a search condition by the searcher 30, reads it by the reader 31, and displays the result selected via the selector 37 on the display 32. Such a search device is disclosed in, for example, Japanese Patent Application Laid-Open No. 8-87508.
[0006]
[Problems to be solved by the invention]
However, in the above-described similarity search device, in order to obtain the similarity in units of words, there is a problem that data must be stored in the state of attribute / attribute value pairs, and the importance is used like the importance of the attribute. Therefore, there is a problem in that the result obtained by each user differs because a target or sample based on the user's intuition or experience or a target or sample reflecting the user's intention must be set.
[0007]
In addition, since the information retrieval apparatus described above converts an input keyword into another keyword using a keyword correlation table indicating a general degree of correlation calculated by some method, data in different fields is used. However, if the character string is the same word, the correlation is the same, and there is a problem that an appropriate result according to the field cannot be obtained. If the correlation table is corrected to avoid this, the entire keyword space is affected and the performance is not necessarily improved for all searches.
[0008]
The present invention has been made to solve the above problems, and automatically generates a degree of association between words included in data based on an index criterion given to a database. The purpose is to obtain a database creation device that automatically creates an index without affecting the index criteria by performing appropriate weighting, and a database search device that performs a search while narrowing down the category for the database. .
[0009]
[Means for Solving the Problems]
In order to achieve the above object, the present invention provides a data input device for inputting data into a database, an index standard reading device for inputting an index standard that defines a configuration that serves as a standard for the index, and input data For the word association, the frequency of words used in the index criterion is examined, the relevance level is calculated for the simultaneously appearing words, and the word relevance map is created based on the relevance level and the index criterion Obtained by dividing the number of words appearing in the first document of the input data by the number of words appearing in the second document which is the summary or heading of the first document. A word importance level assigning device that weights words in the second document by an amount corresponding to the value, and using the weight obtained by the word importance level assigning device, An index creation device that temporarily modifies the relevance of the word relevance map created by the word relevance map creation device, and creates an index for the input data using the modified word relevance map; It is characterized by comprising.
[0010]
According to the present invention, when the data is input by the data input device and the index criterion is input by the index criterion reading device, the word association degree map creating device is used on the basis of the index for the input data. The frequency of appearance of a certain word is checked, the relevance level is calculated for the words that appear simultaneously, and a word relevance map is created based on the relevance level and index criteria. Further, the word importance assigning device divides the number of words appearing in the first document of the input data by the number of words appearing in the second document serving as the summary or heading of the first document. The word in the second document is weighted by the value obtained. Also, the index creation device temporarily modifies the degree of association of the word association degree map using the weighting, and creates an index for the input data using the modified word association degree map.
[0011]
Further, the present invention provides a data input device for inputting data to a database, an index standard reading device for inputting an index standard that defines a configuration that serves as a standard for the index, and the index for the input data. A word relevance map creating device that examines the frequency of occurrence of words used in the reference, calculates the relevance for simultaneously appearing words, and creates a word relevance map based on the relevance and the index criterion; The number of words appearing in the first document of the input data is divided by the number of words appearing in the second document that is the summary or heading of the first document, and the above-mentioned amount is obtained. Using the word importance level assigning device for weighting words in the second document and the weight obtained by the word importance level assigning device, the word relevance map creation An index creation device that temporarily modifies the relevance level of the word relevance map created by the position and creates an index for the input data using the modified word relevance map, and the index creation device And a result display device for displaying the search result. The data search device performs a search based on the index created by the above.
[0012]
According to the present invention, when the data is input by the data input device and the index criterion is input by the index criterion reading device, the word association degree map creating device is used on the basis of the index for the input data. The frequency of appearance of a certain word is checked, the relevance level is calculated for the words that appear simultaneously, and a word relevance map is created based on the relevance level and index criteria. Further, the word importance assigning device divides the number of words appearing in the first document of the input data by the number of words appearing in the second document serving as the summary or heading of the first document. The word in the second document is weighted by the value obtained. Also, the index creation device temporarily modifies the degree of association of the word association degree map using the weighting, and creates an index for the input data using the modified word association degree map. Then, the data search device performs a search based on the created index, and the result display device displays the search result.
[0013]
DETAILED DESCRIPTION OF THE INVENTION
Embodiments of a database creation device and a database search device according to the present invention will be described below in detail with reference to the accompanying drawings.
[0014]
Embodiment 1 FIG.
FIG. 1 is a block diagram showing an example of a database creation apparatus according to the present invention. This database creation device includes a database 3, a data input device 2 for inputting data 1 to the database 3 from the outside of the system, an index reference reading device 5 for inputting an index reference 4 from the outside of the system, and input data 1 Word relevance map creating device 7 that creates word relevance map 8 based on index criterion 4, Index creating device that generates index 9 based on input data 1, word relevance map 8, and index criterion 4 6 and a word importance level assigning device 10 for calculating the importance level of the word when creating the index.
[0015]
The database 3, the word association degree map 8, and the index 9 are stored in a storage device such as a hard disk. Further, the data input device 2, the index reference reading device 5, the index creation device 6, the word relevance map creation device 7 and the word importance assigning device 10 are respectively a computer system, for example, a data input program, an index reference reading program, This is realized by executing an index creation program, a word relevance map creation program, and a word importance assigning program.
[0016]
The index standard 4 indicates the structure of the index. As an example, the index standard 100 for a medical book database is shown in FIG. The medical book index standard 100 has, for example, the title “medicine” in the top layer, and three titles “basic medicine”, “internal science”, and “external science” in the lower layer. "Anatomy" and "Physiology" are one layer below "", and "Cardiology" and "Gastroenterology" are one layer below "Internal science", and one layer below "Surgery" The index is configured to form a tree structure such that there are “local surgery” and “orthopedic surgery”. Like the medical book index standard 100, the index standard 4 is also configured to have a tree structure.
[0017]
In the following description, the medical book index standard 100 will be specifically described as an example. However, it is needless to say that the present invention is not limited to the medical book database and the medical book index standard 100.
[0018]
The word association degree map 8 is a map in which the degree of association between words is given to the hierarchical relation of the index criterion 4. The word relevance map creation device 7 examines the frequency of appearance of words used in the medical book index standard 100, for example, with respect to the input data 1 and performs a predetermined calculation based on the frequency, as shown in FIG. A word relevance map 104 is obtained. A method of creating the word association degree map 8 by the word association degree map creating apparatus 7 will be specifically described with reference to the data 101 shown in FIG.
[0019]
For example, the data 101 is composed of three documents, the title of the document 1 is “Talk of the circulatory organ”, and the abstract is “... the most terrible circulatory disease is angina anatomically. I have heart failure ... " The title of Document 2 is "Complication of cardiovascular disease and hernia", and the abstract is "... anatomically, cardiovascular, ... hernia should be consulted by a surgeon. ... ". The title of Document 3 is “Gastrointestinal and Cardiovascular”, and the abstract is “… If you do not chew well, it may cause irritation to the digestive tract and vomiting. If you have a cardiovascular disease, etc ... "
[0020]
When only words are extracted from these documents 1 to 3, a word string 102 shown in FIG. 4 is obtained. That is, the word string 102 is “circulatory organ” for the title in document 1, “circulatory organ, anatomy, angina, heart failure” for the abstract, and “circulatory organ” for the title in document 2. , Hernia ", abstract," anatomy, circulatory organ, hernia, surgery ", document 3, title" digestive organ, cardiovascular ", abstract," chewy, digestive organs, vomiting, heart " , Angina, cardiovascular.
[0021]
Then, assuming that words appearing simultaneously in one document are related to each other, they are set as a co-occurrence word set 103, and processing is performed on all data. For the co-occurrence word set 103, the degree of association is defined using, for example, the ratio of co-occurrence to the total number of appearances as in the following equation (1). However, the total appearance frequency of a certain word (referred to as “KW1”) is N1, the total appearance frequency of another certain word (referred to as “KW2”) is N2, and “KW1” and “KW2” appear simultaneously. The co-occurrence frequency is N12, and the degree of association between “KW1” and “KW2” is μ12.
[0022]
μ12 = N12 / (N1 + N2-N12) (1)
[0023]
For example, referring to the document 1 described above, the co-occurrence word set 103 includes “circulatory organ, anatomy”, “circulatory organ, angina”, “circulatory organ, heart failure” as shown in FIG. , "Anatomy, angina", "anatomy, heart failure", "angina, heart failure", and so on. For example, for a set of co-occurrence of “cardiovascular and angina”, according to the above equation (1), (co-occurrence frequency of “cardiovascular” and “angina”) / {(“cardiovascular” ) + (Total appearance frequency of “angina”) − (co-occurrence frequency of “circulatory organ” and “angina”)}.
[0024]
Then, by assigning the value, that is, the degree of association 105 to the hierarchical relationship of the index criterion 100, the word association degree map 104 shown in FIG. 3 is obtained. In addition to the above formula (1), various calculation formulas such as calculating the co-occurrence ratio by weighting according to the number of times the word appears in one document can be applied.
[0025]
When the index creation target data includes, for example, a title, an abstract, and a text, the word importance assigning device 10 determines the value between a word that appears in the title, a word that appears in the abstract, and a word that appears in the text. Accordingly, weighting is appropriately performed. In other words, the abstract is generally a concise summary of the text, and the title is a further aggregation of the abstracts. In addition, a value is digitized and weighted for words appearing in the text. A method for determining the weighting by the word importance assigning apparatus 10 will be specifically described with reference to FIG. 6, taking the data 108 shown in FIG. 5 as an example.
[0026]
The data shown in FIG. 5 has a title and an abstract. First, the number of words that appear for each title and abstract is counted (steps S1 and S2 in FIG. 6). For example, for document 1, there are four words included in the abstract: “Cardiovascular”, “Angina”, “Heart failure” and “Anatomy”. On the other hand, the word included in the title is one “circulatory organ”. Therefore, it is considered that the word included in the title has a value four times that of the word included in the abstract. Therefore, the title word is weighted four times with respect to the abstract word (step S3 in FIG. 6). This is performed for each data.
[0027]
For example, document 2 includes two words “Hernia” and “Cardiology” in the title, and four words “Anatomy”, “Cardiology”, “Hernia” and “Surgery” in the abstract. Thus, the title word is weighted twice. Document 3 also includes two words, “Mastication” and “Cardiovascular” in the title, and “Mastication”, “Gastroenterology”, “Vomiting”, “Anatomy”, “Physiology” and “Circulation” in the abstract. 6 words of “container” are included, so the word of the title is three times weighted. With such weighting, for example, in the example shown in FIG. 7, the degree of relevance of “circulator” is originally 0.2, 0.8 for document 1, 0.4 for document 2, and 0.6 for document 3. Therefore, the value of “circulatory organ” differs for each data, that is, document 1, document 2, and document 3 (step S4 in FIG. 6).
[0028]
The same applies to, for example, a case where weighting is performed between an abstract and a text, and a case where weighting is performed between items of other document data. FIG. 8 shows an example of weighting between the abstract and the text. In the data 110 shown in FIG. 8, for example, regarding the text, if the same word appears repeatedly, the number of appearances is taken into account. In addition, simply adding the number of appearances causes a difference in the amount of target documents and books, so normalization is desirable. That is, the amount of text in the body varies depending on the document or book, and more words generally appear as the amount of text increases. Therefore, normalization may be performed so that weighting is performed with respect to a fixed document amount or unit document amount, for example, per page or per 1000 characters.
[0029]
Next, the flow of index creation processing will be described. When the data 1 is input to the database 3 by the data input device 2 and the index reference 100 shown in FIG. 2 is input by the index reference reading device 5, for example, the word association degree map creating device 7 receives the input data 1 Based on the words appearing in the index criterion 4, for example, a word relevance map 8 as shown in FIG. 3 is created.
[0030]
Thereafter, the index creation device 6 creates the index 9 based on the index reference 4 and the word association degree map 8 for the index creation target data, for example, according to the flowchart shown in FIG. That is, first, nodes (words) of classification items included in each document are picked up and mapped to a word association degree map (step S11). As an example, FIG. 7 shows the appearance word marked in the word association degree map 8 for the document 3 of the data 106. In the illustrated example, the marking is made by expressing the corresponding words, ie “digestive organ”, “circulatory organ”, “chewing”, “vomiting”, “heart” and “anginal” in underlined bold letters. Indicated.
[0031]
Subsequently, weighting is performed by the word importance assigning device 10, and the word association degree map 8 is temporarily corrected (step S12). In the example shown in FIG. 7, in the case of document 3, two words “digestive organ” and “circulatory organ” appear in the title, and the number of appearance words in the abstract is six. Only during processing, for “digestive organs” and “circulatory organs”, the relevance of the word relevance map 8 is temporarily tripled to 0.6 (0.2 × 3).
[0032]
Subsequently, the word appearing in the index creation target data is checked as a terminal word, and a classification determination evaluation value is calculated so as to go back to the root node (“medicine”) from each terminal word (step S13). This is because each word mapped to the word relevance map 8 is calculated and evaluated according to a certain calculation procedure, so that the positioning in the entire classification system is not evaluated at the mapped position alone. This is to consider whether it is in
[0033]
That is, for example, in the example shown in FIG. 7, the classification item “heart” does not simply mean the word “heart”, but “heart” related to “cardiology” related to “internal medicine” related to “medicine”. Means the concept. In order to reflect this, for example, the mapped classification items are traced back in order from the terminal word node “heart” to the “medicine” of the root node, and the degree of relevance in the middle is added, and the obtained degree of relevance Is divided by the number of hierarchies obtained to obtain an average value, which is used as a classification determination evaluation value.
[0034]
In the example shown in FIG. 7, in the case of document 3, the degree of association between “heart” and its upper “circulatory organ” is 0.9, and “circulatory organ” and its uppermost “internal medicine” Is originally 0.2, but temporarily becomes 0.6 due to weighting. Furthermore, the degree of association between “internal science” and “medicine”, which is one higher level, is 0.3. is there. Therefore, the classification determination evaluation value of the word “heart” is 0.6 by adding 0.9, 0.6, and 0.3 and dividing it by 3. That is, the degree to which the document 3 is classified as “heart” is 0.6.
[0035]
However, in the document 3 shown in FIG. 7, the words “heart” and “circulatory organ” appear, but the words “internal science” and “medicine” in the middle of going back to the root node do not appear. In this way, when a word that has not been mapped appears until reaching the “medicine” of the root node, and is interrupted, the relevance of the word relevance map 8 is not added as it is, as in the next step S14. Process.
[0036]
That is, for example, referring to the example shown in FIG. 7, when the document 1 is traced from the terminal word node “angina” to the upper layer, the word “heart” does not appear in the document 1. Therefore, the average value of the relevance levels of the lower nodes of “heart” is obtained. Specifically, the average value of 0.5 ((0.5 + 0.5) / (relationship level) of 0.5 for “angina”, which is a lower node of “heart”, and 0.5 for “heart failure” 2). Then, the product of the average value and the value 0.9 of the relevance level of “heart” with respect to “circulatory organ” is obtained, and the value 0.45 (0.9 × 0.5) is added as the temporary relevance level. (Step S14).
[0037]
First, “the degree to which the document 3 is classified as“ heart ”is 0.6” is described. However, by performing the process of step S14, the classification determination table value of the word “heart” in the document 3 is Since the degree of relevance of “Internal Medicine” to “Medicine” is 0.18 ((0.6 + 0.6) /2×0.3), it is actually 0.56 ((0.9 + 0.6 + 0.18). ) / 3).
[0038]
If step S13 described above and the node is interrupted, step 14 is repeated for all terminal word nodes (step S15). For example, in the case of the data shown in FIG. 7, for Document 3, all of the “mastication”, “digestive organs”, “vomiting”, “heart”, “angina pectoris”, and “heart failure” are all on the way to the root node. Evaluate the node. The calculation formulas and calculation results of the classification judgment table values are shown for each of “Mastication”, “Gastrointestinal”, “Vomiting”, “Heart”, “Angina pectoris” and “Heart failure”. In the calculation formula, a value surrounded by ““ ”and“ ”” is a temporary relevance, and is added in order from the lower node to the upper node. Note that “physiology” and “basic medicine” are omitted.
[0039]
“Mastic”: (0.7+ “0.23” + “0.14” + “0.09”) / 4 = 0.29
“Vomiting”: (0.8+ “0.23” + “0.14” + “0.09”) / 4 = 0.32
“Digestion”: (“0.23” + “0.14” + “0.09”) / 3 = 0.15
“Angina pectoris”: (0.5 + 0.9 + 0.6 + “0.18”) / 4 = 0.55
“Heart failure”: (0.5 + 0.9 + 0.6 + “0.18”) / 4 = 0.55
“Heart”: (0.9 + 0.6 + “0.18”) / 3 = 0.56
“Circulator”: (0.6+ “0.18”) / 2 = 0.39
“Digestive organ”: (0.6+ “0.18”) / 2 = 0.39
[0040]
As described above, when all the evaluations of the classification items that are likely to be the classification destination of the index creation target data are completed, the item having the highest evaluation is determined as the classification destination and classified (step S16). In the example shown in FIG. 7, since the classification item of “heart” has the highest evaluation value (0.56), the classification destination is determined to be “heart”. Then, after the word association degree map 8 temporarily corrected by weighting is returned to the initial value (step S17), the same processing is repeated for all the documents for which the index is to be created (step S18).
[0041]
As an evaluation method for determining a document classification destination, a method of normalizing and adding according to the number of layers or the size of a map can be applied. For example, in consideration of the hierarchy in the word relevance map 8 of words that appear in the booklet, the relevance may be weighted uniformly for deep layers (that is, lower layers). Then, even if the number of levels of the classification system is very large, the arithmetic mean value of the relevance level is excessively low, and it is possible to avoid accidental appearance and classification into upper level word items. it can.
[0042]
According to the first embodiment, the degree of association between words included in the data is automatically generated based on the index criterion 4, and further appropriate weighting is performed on the degree of association. Even if the index creator does not particularly specify a sample or a typical example, an index for each word can be automatically created without affecting the index standard 4.
[0043]
The data for which the index is created is not limited to a document, and any data can be used as long as it is data stored in a database and can recognize words. For example, the index creation target data may be WEB page data on the Internet including a tag corresponding to the control code.
[0044]
Embodiment 2. FIG.
FIG. 10 is a block diagram showing an example of the database search apparatus according to the present invention. This database search device includes a data search device 38 that performs a search based on the index created by the index creation device 6 in the database creation device of the first embodiment shown in FIG. 1, and a result display device that displays the search result. 39 is added. Accordingly, the data input device 2, the database 3, the index reference reading device 5, the index creation device 6, the word association degree map creation device 7 and the word importance assigning device 10, and the index criterion 4 and the word association degree constituting the database creation device. Since the map 8 is the same as that of the first embodiment, the description thereof is omitted.
[0045]
The data search device 38 is realized in a computer system by executing, for example, a data search program. For example, the data search device 38 displays an index menu 111 created based on the index standard 100 as shown in FIG. 2 on the result display device 39, and an appropriate item from the menu. For example, a mouse cursor 112 is provided. Therefore, although not shown, the data search device 38 is connected with a pointing device such as a mouse or a keyboard as an input device. The result display device is, for example, a cathode ray tube or a liquid crystal display device which is a display device of a computer system.
[0046]
When searching for an index created by the index creation device 6, the searcher moves the mouse cursor 112 to the menu displayed on the result display device 39 to indicate an appropriate item and select it. By doing so, it is possible to search for an index and to search for a target book.
[0047]
As shown in FIG. 12, the index reference is divided for each node, the “internal medicine” is instructed from the “medicine” node 114 with the mouse cursor 112, and the “internal medicine” node 115 is opened. The target book may be searched by pointing the “circulatory organ” with the cursor 112 to open the “circulatory organ” node 116 and finally selecting the “lymph gland” with the mouse cursor 112. Searches can be effectively performed by a menu-like interface that narrows down the classification items one after another with respect to the index criteria having such a tree structure.
[0048]
【The invention's effect】
As described above, according to the present invention, when data is input by the data input device and the index reference is input by the index reference reading device, the word association degree map creating device The frequency of occurrence of words used in the index criterion is checked, the relevance level is calculated for the words that appear at the same time, and a word relevance map is created based on the relevance level and the index criterion. Further, the word importance assigning device divides the number of words appearing in the first document of the input data by the number of words appearing in the second document serving as the summary or heading of the first document. The word in the second document is weighted by the value obtained. Also, the index creation device temporarily modifies the degree of association of the word association degree map using the weighting, and creates an index for the input data using the modified word association degree map. Therefore, even if the index creator does not particularly specify a sample or a typical example, an index for each word can be automatically created without affecting the index criteria.
[0049]
According to the next invention, when the data is input by the data input device and the index reference is input by the index reference reading device, the word association degree map creating device is used for the input data with the index reference. The frequency of occurrence of a word is checked, a relevance level is calculated for words that appear simultaneously, and a word relevance map is created based on the relevance level and an index criterion. Further, the word importance assigning device divides the number of words appearing in the first document of the input data by the number of words appearing in the second document serving as the summary or heading of the first document. The word in the second document is weighted by the value obtained. Also, the index creation device temporarily modifies the degree of association of the word association degree map using the weighting, and creates an index for the input data using the modified word association degree map. Then, the data search device performs a search based on the created index, and the result display device displays the search result. Therefore, the index can be searched efficiently and the target data can be searched.
[Brief description of the drawings]
FIG. 1 is a block diagram showing an example of a database creation device according to the present invention.
FIG. 2 is a system diagram showing an example of the configuration of index criteria used in the database creation device.
FIG. 3 is a schematic diagram showing an example of a word association degree map created by the database creation device.
FIG. 4 is an explanatory diagram for explaining a method of creating a word association degree map.
FIG. 5 is an explanatory diagram for explaining a method of weighting a word association degree map.
FIG. 6 is a flowchart illustrating an example of a weighting determination method.
FIG. 7 is a schematic diagram showing an example of a word association degree map that is weighted.
FIG. 8 is a schematic diagram showing an example of weighting between an abstract and a text.
FIG. 9 is a flowchart illustrating an example of an index creation method.
FIG. 10 is a block diagram showing an example of a database search device according to the present invention.
FIG. 11 is a schematic diagram showing an example of a search menu used in the database search device.
FIG. 12 is a schematic diagram showing another example of a search menu used in the database search device.
FIG. 13 is a block diagram showing a conventional database search apparatus.
FIG. 14 is a block diagram showing a conventional database search apparatus.
FIG. 15 is a block diagram showing a conventional database search device.
[Explanation of symbols]
1 data, 2 data input device, 3 database, 4 index standard, 5 index standard reading device, 6 index creation device, 7 word relevance map creation device, 8 word relevance map, 9 index, 10 word importance assigning device.

Claims (2)

A data input device for entering data into the database;
An index standard reading device for defining an index standard configuration and inputting an index standard configured so that the index has a tree structure;
For the input data, the frequency of occurrence of words used in the index criterion is examined, the degree of association is calculated for words that appear at the same time, and based on the degree of association and the index criterion, the index criterion A word relevance map creating device for creating a word relevance map, which is a map in which the relevance between the words is given to a hierarchical relationship;
The number of words appearing in the first document of the input data is divided by the number of words appearing in the second document that is the summary or heading of the first document, and the above-mentioned amount is obtained. A word importance assigning device for weighting words in the second document;
Each word that appears in the input data by temporarily correcting the relevance of the word relevance map created by the word relevance map creating device using the weight obtained by the word importance assigning device Ri dates back said modified word relevance map as a word end to the root node, adding the relevance of the course traced back to the root node, the average is divided by the number of hierarchical predated the total of the obtained relevance A classification determination evaluation value that is an evaluation value for classifying the input data by obtaining a value is calculated, and the terminal word having the highest classification determination evaluation value is used as a classification destination to index the input data An index creation device for creating
A database creation device comprising: A data input device for entering data into the database;
An index standard reading device for defining an index standard configuration and inputting an index standard configured so that the index has a tree structure;
For the input data, the frequency of occurrence of words used in the index criterion is examined, the degree of association is calculated for words that appear at the same time, and based on the degree of association and the index criterion, the index criterion A word relevance map creating device for creating a word relevance map that is a map in which the relevance between the words is given to a hierarchical relationship;
The number of words appearing in the first document of the input data is divided by the number of words appearing in the second document that is the summary or heading of the first document, and the above-mentioned amount is obtained. A word importance assigning device for weighting words in the second document;
Each word that appears in the input data by temporarily correcting the relevance of the word relevance map created by the word relevance map creating device using the weight obtained by the word importance assigning device Ri dates back said modified word relevance map as a word end to the root node, adding the relevance of the course traced back to the root node, the average is divided by the number of hierarchical predated the total of the obtained relevance A classification determination evaluation value that is an evaluation value for classifying the input data by obtaining a value is calculated, and the terminal word having the highest classification determination evaluation value is used as a classification destination to index the input data An index creation device for creating
A data search device for performing a search based on the index created by the index creation device;
A result display device for displaying the search results;
A database search apparatus comprising:

Images