JP2770527B2 - Information retrieval device - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an information retrieval apparatus for retrieving predetermined information from a database.

2. Description of the Related Art Conventionally, as a method of searching a database using ambiguous search conditions, for example, there has been a method using a similar word relationship as disclosed in JP-A-60-140443. FIG. 15 shows an example of a conventional information retrieval apparatus using a similar word relation.
In FIG. 15, 1501 is an external storage device that stores data, 1502 is a database management system that manages input and output of data in the external storage device, 1503 is a search condition input unit that receives search conditions, and 1504 is a search condition input unit. A similar word storage unit that stores similar word relationships, 1505 is a similar word conversion unit that changes input search conditions using the similar word relationships stored in the similar word storage unit, and 1506 is a search result output that outputs search results Department.

FIG. 16 shows a person database as an example of information stored in the external storage device 1501, and FIG. 17 shows an example of a synonym relation stored in the synonym storage unit. This synonym relationship indicates a synonym relationship between the industries, such as "soft" and "electric", "mechanical" and "precision" and "automobile", and
“Securities” and “trading companies” are registered as similar words.

The operation of the conventional information retrieval device configured as described above will be described below. Consider a case in which a search condition of “people whose type of business is close to machines” is input to the search condition input unit 1503. The synonym conversion unit 1505 extracts synonyms “precision” and “automobile” corresponding to the keyword “machine” from the synonym relation between the industries shown in FIG. 17 in the synonym storage unit 1504, and extracts the search condition. The data is converted into "people whose type is machine, precision, or automobile" and sent to the database management system 1502. The database management system 1502 searches the database using the converted search conditions. As a search result, the 18th
As shown in the figure, “E” and “F” are obtained and output by the search result output unit 1506.

In this conventional example, it is possible to broadly search for a person whose business type is not “machine” itself using a search condition of an ambiguous value “close to machine” for an attribute existing in the database such as “business type”. It is possible. However, if the attribute of the search condition does not exist in the database, for example, if you want to search the database using the search condition that "body type" is "skinny", since the attribute "body type" does not exist in the database The searcher must rewrite the search condition into a search condition having the same meaning using the attributes existing in the database. In other words, the conventional information search device can handle the ambiguity of the value, but can only search for the attribute using the attribute itself existing in the database.

On the other hand, as a conventional information search device that performs a search using an attribute that does not exist in the database, for example, a similar information search device described in IEICE Information Processing Research Group Study Group Material IP-89-2, pp.11-22 There is a system to search for figures. In this conventional example, first, feature quantities of a figure shape such as a density distribution, a schematic shape, a frequency distribution, a local correlation, and a local contrast are extracted from the figure data and stored in a database together with the figure data. Furthermore, a search method has been proposed in which similarity between figures is represented by a mathematical model using the feature amount, thereby searching for similar figures.

The example of the configuration diagram of the information search device that performs a search using an attribute that does not exist in the database using the mathematical expression model is shown in FIG.
Shown in the figure. In FIG. 19, reference numeral 1901 denotes an external storage device that stores data, 1902 denotes a database management system that manages input / output of data in the external storage device, 1903 denotes a search condition input unit that receives search conditions, and 1904 denotes a search condition input unit. A mathematical operation unit for converting a search condition using an attribute not to be used into a search condition represented by an attribute existing in the database using a mathematical model, and a search result output unit 1905 for outputting a search result.

However, in the conventional information retrieval device having such a configuration,
A qualitative and ambiguous search condition that cannot be described by a mathematical model could not be represented.

Problems to be Solved by the Invention As described above, the conventional information retrieval apparatus can handle the ambiguity of the data value, but only retrieves the attribute of the data using the attribute itself existing in the database. There was a problem that could not be done. Further, there is a problem that even an information search device that performs a search using an attribute that does not exist in a database cannot express a qualitative and ambiguous search condition that cannot be described by a mathematical model. Therefore, if the searcher has no knowledge of the contents or structure of the database, or of course has only search conditions that include attributes that do not exist in the database or qualitative and ambiguous search conditions, appropriate search Creating conditions was difficult. For this reason, the conventional information search apparatus has a problem that a skilled search engineer has to repeat trial and error of creating search conditions appropriate for the search intention and performing a search until an appropriate search result is obtained. Had a point.

In view of the foregoing, the present invention provides a searcher who does not know the contents and structure of a database, or a searcher who has only an abstract and ambiguous search condition, and has an ambiguous search condition including an abstract attribute that does not exist in the database. An object of the present invention is to provide an information retrieval apparatus capable of retrieving data appropriate for a retrieval intention from a database by using the information retrieval apparatus.

Means for Solving the Problems The present invention provides an external storage device that stores data, a database management system that manages input / output of data in the external storage device, and a search condition input unit that receives a search condition including a fuzzy number. An inference rule storage unit that stores a fuzzy inference rule, a fuzzy number storage unit that stores a fuzzy number, a fuzzy relation storage unit that stores a fuzzy relationship, and the inference rule and the fuzzy number. A fuzzy inference management unit that manages fuzzy inference to be input, a preconditioning unit calculating unit that obtains a degree of conformity between the search condition and the antecedent part of the inference rule according to an instruction of the fuzzy inference management unit, A consequent part operation unit for obtaining an inference result including a fuzzy number from the consequent part of the inference rule in accordance with the fuzzy number and the fuzzy number in the fuzzy number storage unit A fuzzy matching unit that uses the fuzzy relation in the relation storage unit to determine the degree of conformity between the inference result and the data in the external storage device; and a data in which the degree of conformity obtained by the fuzzy matching is equal to or greater than a threshold. And a search result output unit that outputs a search result as a search result.

The present invention also provides an external storage device for storing data,
A database management system for managing data input / output of the external storage device, a search condition input unit for receiving a search condition including a fuzzy number, an inference rule storage unit for storing a fuzzy inference rule, and a fuzzy number for storing a fuzzy number A storage unit, a fuzzy relation storage unit that stores a fuzzy relation, a fuzzy inference management unit that manages fuzzy inference using the inference rule, the fuzzy number, and the fuzzy relation and that receives the search condition as an input; An antecedent operation unit for obtaining a degree of conformity between the search condition and the antecedent of the inference rule according to an instruction of the inference management unit; and an inference including a fuzzy number from a consequent part of the inference rule according to the degree of adaptation. Using the consequent part calculating unit for obtaining the result and the fuzzy relation in the fuzzy number storage unit, the degree of conformity between the inference result and the data in the external storage device And fuzzy matching unit for obtaining,
A search result output unit that outputs, as a search result, data in which the degree of matching obtained by the fuzzy matching is equal to or greater than a certain threshold.

Further, the present invention provides an external storage device for storing data, a database management system for managing data input / output of the external storage device, a search condition input unit for receiving a search condition including a fuzzy number, and a fuzzy inference rule. An inference rule storage unit for storing, a fuzzy number storage unit for storing a fuzzy number, a fuzzy relation storage unit for storing a fuzzy relation, and the search condition is input using the inference rule, the fuzzy number, and the fuzzy relation. A fuzzy inference managing unit that manages fuzzy inference, a precondition part calculating unit that obtains a degree of conformity between the search power condition and the antecedent part of the inference rule according to an instruction of the fuzzy inference managing unit; And a consequent part operation unit for obtaining an inference result that is a real number from the consequent part of the inference rule, and outputting data matching the inference result as a search result. An information retrieval apparatus and a search result output unit.

The present invention also provides an external storage device for storing data,
A database management system that manages input / output of data in the external storage device, a fuzzy inference unit that receives the search condition as input, and infers an inference result that is a fuzzy number, and the inference result and data in the external storage device. Fuzzy matching unit for calculating the degree of conformity, a conversion table storage unit for storing a correspondence table between the degree of conformity determined by the fuzzy matching and the natural language, and adaptation for converting the degree of conformity to a natural language using the correspondence table An information search device including a degree conversion unit and a search result output unit that outputs data whose relevance is equal to or higher than a threshold value as a search result.

Effect of the Invention According to the configuration described above, when an ambiguous search condition including an abstract attribute that does not exist in the database is input, the fuzzy inference is performed using the search condition as an input, and the present invention is configured by the attribute existing in the database. Is converted to a search condition including a fuzzy number. Furthermore, by using the fuzzy numbers and fuzzy relations, the relevance between the converted search conditions and the data in the database is obtained, and data in which the relevance is above a certain threshold is output as a search result. Even searchers who do not know the content or structure of the data, or who have only abstract and ambiguous search conditions, can use the ambiguous search conditions that include abstract attributes that do not exist in the database to retrieve data that is appropriate for the search intent. Can be retrieved from the database.

Embodiment FIG. 1 shows a configuration diagram of an information retrieval apparatus according to an embodiment of the first invention. In FIG. 1, 101 is an external storage device for storing data, 102 is a database management system for managing data input / output of the external storage device, 103 is a search condition input unit for receiving a search condition including a fuzzy number, 10
4 is an inference rule storage unit that stores fuzzy inference rules, 105 is a fuzzy number storage unit that stores fuzzy numbers, 106 is a fuzzy relation storage unit that stores fuzzy relations, 107 uses the inference rules and the fuzzy numbers, A fuzzy inference management unit that manages fuzzy inference using the search condition as an input, 108
Is an antecedent operation unit that calculates the degree of conformity between the search condition and the antecedent part of the inference rule in accordance with the instruction of the fuzzy inference management unit. Consequent part operation part for finding inference results including numbers, 11
0 is a fuzzy matching unit that uses the fuzzy number in the fuzzy number storage unit and the fuzzy relation in the fuzzy relation storage unit to determine the degree of conformity between the inference result and the data in the external storage device. This is a search result output unit that outputs, as a search result, data in which the matching degree obtained by matching is equal to or greater than a certain threshold.

As an example of information stored in the external storage device 101,
The same person database of FIG. 16 as the conventional example is used.

FIG. 2 shows an example of inference rules stored in the inference rule storage unit 104. The if part and then part in FIG. 2 represent the antecedent and consequent parts of the fuzzy inference rule, respectively. In addition, the values of the variables in the inference rules are “thin,” “fat,” “large,”
“Slightly small” is a label indicating a fuzzy number.

FIG. 3 shows an example of the membership function stored in the fuzzy number storage unit 105. In FIG. 3, a, b, and c are fuzzy numbers for attributes “annual income”, “body shape”, and “height-weight”, “low, medium, high”, “thin, medium, fat, and fat”. "Small, slightly small, medium, large."

The fuzzy relation stored in the fuzzy relation storage unit 106 includes a similar relation and an inclusive relation. FIG. 4 shows an example of the similar relation. In FIG. 4, for example, a value of 0.2 at a portion where “soft” and “machine” intersect indicates a degree indicating a semantic similarity between the two.

The operation of the present embodiment will be described below using a search condition of “a person with a high annual income, a large body shape, and a business type similar to a machine” as an example.

When a search condition is input to the search condition input unit 103, the search condition is sent to the antecedent part operation unit 108 as it is. The antecedent part operation unit 108 and the consequent part operation unit 109 perform fuzzy inference using the above search conditions as input, and obtain inference results having attributes existing in the database. The inference result is used as a new search condition and sent to the fuzzy matching unit 110. The above search conditions are passed through fuzzy inference and fuzzy matching
The operation before being sent to 110 will be described below. First, the fuzzy inference management unit 107 sequentially retrieves the antecedent part of the inference rule from the inference rule storage unit 104, and sends it to the antecedent part operation unit. The antecedent operation unit 108 determines the degree of conformity between the search condition and the antecedent of each inference rule. Search conditions for which there is no matching inference rule are sent to the fuzzy matching unit 110 without performing fuzzy inference.

As for the search condition “annual income = high”, there is no inference rule whose attribute matches that of “annual income”, so that it is sent to the fuzzy matching unit 110 as it is.

The search condition “business type = close to machine” is sent to the fuzzy matching unit 110 as it is because there is no inference rule that matches the value “machine”.

With respect to the search condition “body type = fat”, the degree of conformity with the antecedent part of the inference rule is obtained using the fuzzy numbers shown in FIG. Here, the relevance ρ between the search condition and the antecedent part
Uses the largest grade of the intersection of both membership functions as follows:

x: domain of fuzzy number μ _k (x): membership function of search condition μ _l (x): membership function of antecedent part In this example, the search condition “body type = fat” is shown in FIG. The antecedent part of rule 2 "body type = slightly overweight" is satisfied, and the degree of conformity is 0.8 as shown in Fig. 3b. Incidentally, the antecedent part of the inference rule may not only be composed of one condition but also composed of a combination of a plurality of conditions. In this case, as the search condition and the antecedent part fitness ρ, the minimum value of the fitness of each condition is used as follows.

ρ _i : the degree of conformity between the i-th condition of the antecedent part and the search condition Next, the fuzzy inference management unit 107 sends the degree of conformity of the antecedent part of the inference rule and the consequent part to the consequent part operation unit 109. . Consequent arithmetic section 109 first determines a consequent output .mu.o _i for the i-th inference rules. Here, the output of the consequent part uses a fuzzy number in a form in which a grade exceeding the fitness of the antecedent part is cut off from the membership function of the consequent part as follows.

_{μo i (x) = min {} ρ, μ r (x)} x: domain of fuzzy number μ _r (x): membership functions of the consequent part Next, the inference rule attribute of the consequent is the same Since there are a plurality of inference rules, the inference result μo of the plurality of inference rules is obtained. Here, the union of the consequent output μo _i (x) of each inference rule is used as the overall inference result μo as follows.

x: fuzzy number domain μo _i (x): output of the consequent part of the i-th inference rule Since the relevance of the antecedent part of rule 2 to the search condition “body type = fat” is 0.8, From the membership function of the consequent part “height−weight = slightly small” as the output of the consequent part of rule 2, a fuzzy number in the form shown in FIG. You.
Since there is no inference rule other than the rule 2 that satisfies the search condition, the fuzzy number shown in FIG. 5 is used as it is as the overall inference result and sent to the fuzzy matching unit 110. Therefore, the fuzzy matching unit 110 determines that “annual income =
High, height-weight = (fuzzy number shown in FIG. 5), and business type = close to machine ".

The fuzzy matching unit 110 obtains the degree of matching between the search condition and the data in the database.
A fuzzy number is used for matching continuous values, and a fuzzy relationship is used for matching discrete values.

In this search example, for the search condition of “person with high annual income”, “annual income” is an attribute that takes a continuous value.
Using the fuzzy number “high” for “annual income” shown in a., the degree of conformity μ _HI (x) (x: annual income of each person) to the annual income data of each person in the database is obtained. For example,
Since Mr. B's annual income is 10 million yen, its fitness is μ _HI
(10 million) = 0.8. The relevance is obtained for all data, and a search result as shown in FIG. 6a is obtained. Similarly, for the search condition "height-weight = (fuzzy number shown in FIG. 5)", a result as shown in FIG. 6b is obtained.

On the other hand, for the search condition of “people whose business type is close to machines”, since “business type” is an attribute that takes a discrete value, the business type “machine” and the database Determine the degree of conformity with the business data of each person. The relevance is determined by the similarity of the fuzzy relation. For example, since the business type of "Mr. E" is "automobile", by examining the similarity between "machine" and "car" in the similarity relation table, the fitness of "Mr. E" becomes 0.8. The relevance is obtained for all data, and a search result as shown in FIG. 6c is obtained.

When the search condition is composed of a combination of a plurality of conditions, the minimum value of the fitness levels of the respective conditions is used as the fitness m of the entire search condition and each person in the database as follows.

x: each person in the database _mi (x): relevance between the i-th condition in the search condition and the data Therefore, the search condition "annual income = high and height-weight =
(The fuzzy number shown in Fig. 5) and the type of business = close to machine ", the person whose minimum value is 0.3 or more is taken from Fig. 6a, b, c, and the result of Fig. 6 d. Is obtained.

As described above, according to the present invention, a database can be searched with ambiguous search conditions including a fuzzy number by fuzzy inference and fuzzy matching. Also,
It is possible to search with ambiguous search conditions that include attributes that do not exist in the database.

FIG. 7 is a configuration diagram of an information retrieval apparatus according to one embodiment of the second invention. In FIG. 7, reference numeral 701 denotes an external storage device for storing data, 702 a database management system for managing data input / output of the external storage device, 703 a search condition input unit for receiving a search condition including a fuzzy number,
Reference numeral 704 denotes an inference rule storage unit that stores fuzzy inference rules.
Is a fuzzy number storage unit that stores a fuzzy number, 706 is a fuzzy relationship storage unit that stores a fuzzy relationship, 707 uses the inference rule, the fuzzy number, and the fuzzy relationship,
A fuzzy inference management unit 708 that manages fuzzy inference using the search condition as input, an antecedent operation unit 708 for obtaining a degree of conformity between the search condition and the antecedent of the inference rule in accordance with an instruction of the fuzzy inference management unit. , 709 is a consequent part operation unit for obtaining an inference result including a fuzzy number from the consequent part of the inference rule in accordance with the degree of matching, and 710 is a fuzzy number in the fuzzy number storage unit and a fuzzy number in the fuzzy relation storage unit. Using a fuzzy relationship, a fuzzy matching unit that determines the degree of conformity between the inference result and the data in the external storage device, 711 uses, as a search result, data in which the degree of conformity determined by the fuzzy matching is equal to or greater than a threshold. This is a search result output unit to be output.

As an example of information stored in the external storage device 701,
The same person database of FIG. 16 as the conventional example is used. As an example of the inference rules stored in the inference rule storage unit 704,
The same one shown in FIG. 2 as in the first invention is used as an example of the membership function stored in the fuzzy number storage unit 705 as shown in FIG. 3 as in the first invention.

The operation of the embodiment of the second invention will be described below using the same search condition as that of the first invention, that is, "a person with a high annual income, a large body, and a business type close to a machine". I do.

When a search condition is input to the search condition input unit 703, the search condition is sent to the antecedent part calculation unit 708 as it is. The antecedent part operation unit 708 and the consequent part operation unit 709 perform fuzzy inference using the above search conditions as input, and obtain an inference result having an attribute existing in the database. The inference result is used as a new search condition and sent to the fuzzy matching unit 710. The above search conditions are passed through fuzzy inference and fuzzy matching
The operation up to the transmission to 710 will be described below.

The antecedent operation unit 708 obtains the degree of conformity of the inference rules with the antecedent of the inference rules, such as “Annual income = high”, “Body type = fat” or “industry = close to machine” in the search condition. In the first invention, only the fuzzy number is used to determine the degree of conformity of the antecedent part, but in the second invention, not only the fuzzy number but also the fuzzy relation is used. The operation for determining the degree of conformity will be described below.

In contrast to the search condition “business type = close to machine”, there was no inference rule matching “machine” in the first invention. On the other hand, in the second invention, by using the fuzzy relation of FIG. 4, not only “machine” but also inference rules applicable to similar industries are matched to search conditions. Here, the similarity is defined as the similarity in the fuzzy relation. Similar industries of "machinery" include "software (similarity 0.2)", "automobile (0.8)" and "electricity (0.
5) "and" Precision (0.8) "," Industry = Close to Machinery "matches the antecedent" Industry = Automobile "in Rule 3 of Fig. ”And“ automobile ”are 0.8.

When the antecedent part of the inference rule consists of a combination of a plurality of conditions, the method of obtaining the matching condition between the search condition and the antecedent part is the same as in the first invention.

The search conditions “annual income = high” and “body type = fat” are obtained in the same manner as in the first invention using the fuzzy numbers in FIG. 3, and therefore description thereof is omitted.

Next, the fuzzy inference management unit 707 sends the consequent part of the inference rule and its conformity to the consequent part calculation unit 709. The consequent part calculation unit 709 obtains the inference result in the same manner as in the embodiment of the first invention. In this search example, the antecedent of Rule 2 is 0.
8, and the conformity of the antecedent part of Rule 3 is 0.8, so that "height-weight = (fuzzy number shown in FIG. 5)" and "hobby = automobile (fitness 0.8)" are the inference results. This is inferred, and this is used as a search condition in the fuzzy matching unit 71.
Passed to 0. The other search condition “annual income = high” is passed to the fuzzy matching unit 710 as it is because there is no inference rule that matches. That is, the fuzzy matching unit 7
10 is a search condition of "Annual income = high and height-weight = (fuzzy number shown in Fig. 5) and (industry = close to machine, or hobby = car (fitness 0.8))" receive.

The fuzzy matching unit 710 obtains the matching degree between the search condition and each person in the database, as in the embodiment of the first invention. With respect to the search conditions “annual income = high” and “height / weight = (fuzzy number shown in FIG. 5)”, similar to the first invention, the result shown in FIG. 8 ab is obtained. As a result of the search condition “industry = close to machine or hobby = automotive (fitness 0.8)”, FIG. 8c in which the suitability of “Mr. C” in FIG. 6c is changed is obtained. Therefore, for the search condition "Annual income = high and height-weight = (fuzzy number shown in Fig. 5) and business type = close to machine", for each person, Fig. 8a, b, c Takes the minimum value of 0.3 or more, and the result of FIG. 8 d is obtained. This result includes:
Ms. C who was not obtained in the first invention is included.

As described above, according to the second invention, a flexible search in which search conditions are interpreted more broadly than in the first invention can be performed.

FIG. 9 is a configuration diagram of an information retrieval device according to one embodiment of the third invention. In FIG. 9, reference numeral 901 denotes an external storage device for storing data; 902, a database management system for managing data input / output of the external storage device; 903, a search condition input unit for receiving a search condition including a fuzzy number;
904 is an inference rule storage unit that stores fuzzy inference rules, 905
Is a fuzzy number storage unit that stores a fuzzy number, 906 is a fuzzy relationship storage unit that stores a fuzzy relationship, 907 uses the inference rule, the fuzzy number, and the fuzzy relationship,
A fuzzy inference management unit 908 that manages fuzzy inference using the search condition as an input, a preconditioner operation unit 908 for obtaining a degree of conformity between the search condition and the antecedent of the inference rule in accordance with an instruction of the fuzzy inference management unit , 909 is a consequent part operation unit for obtaining an inference result that is a real number from the consequent part of the inference rule according to the degree of matching, and 910 is a search result output that outputs a result of searching a database using the inference result as a search condition Department.

Examples of information stored in the external storage device 901 include:
The same person database of FIG. 16 as the conventional example is used. First
As in the case of the invention of FIG. 1, the one shown in FIG. 2 as an example of the inference rule stored in the inference rule storage unit 904 is shown in FIG. 3 as an example of the membership function stored in the fuzzy number storage unit 905. Used.

The operation of the embodiment of the third invention is the same as that of the first and second inventions in that "the annual income is high, the body shape is fat, and
This will be described below using an example of a search condition of “a person whose business type is close to a machine”.

After inputting the search conditions, fuzzy inference is performed, and as a result, “yearly income is high and height−weight = (fuzzy number shown in FIG. 5), and (industry = close to machine or hobby = Until an "automobile (degree of conformity: 0.8)" is obtained, the description is omitted because it is the same as that of the embodiment of the second invention.

Next, the consequent part operation unit 909 obtains an α level set of the fuzzy numbers in the condition. The α-level set is a non-fuzzy set defined as follows.

Aα = {x | μ _A (x) ≧ α｝, α [0,1] A: Fuzzy number Aα: α level set of A If α = 0.3, the fuzzy numbers in FIGS. From the fuzzy relation of FIG. 4, the α level set of the search condition is expressed as “annual income ≧
6.3 million yen, and height-weight ≤ 110 cm, and (industry = machinery, or industry = automobile, or industry = electricity,
Or, industry = precision, or hobby = automobile). The consequent part operation unit sends the condition, which is a real number, to the database management system 902 to obtain a search result as shown in FIG.

According to the third aspect of the present invention having the above-described configuration, the fuzzy matching section is not required as compared with the first and second aspects, so that the configuration is simplified and the calculation time is reduced.

FIG. 12 is a configuration diagram of an information search device according to an embodiment of the fourth invention. In FIG. 12, 1201 is an external storage device that stores data, 1202 is a database management system that manages input and output of data in the external storage device, 1203 is a search condition input unit that receives search conditions including a fuzzy number,
Reference numeral 1204 denotes a fuzzy inference unit that receives the search condition as input and infers an inference result that is a fuzzy number, 1205 denotes a fuzzy matching unit that obtains a degree of matching between the inference result and data in the external storage device, and 1206 denotes the fuzzy matching. A conversion table storage unit for storing a correspondence table between the degree of matching and the natural language obtained in step 1207, using the correspondence table, a matching degree conversion unit for converting the degree of matching into a natural language, and 1208. A search result output unit that outputs data equal to or more than a threshold value as a search result together with the relevance in natural language.

FIG. 13 shows an example of the correspondence table stored in the conversion table storage unit 1206.

The operation of the information retrieval apparatus of the present embodiment configured as described above will be described below. The operation from the input of the search condition to the fuzzy matching unit is the same as that of the first or second invention, and the description is omitted. Consider a case where a search condition of “people whose type of business is close to machines” is input. In the same manner as in the first invention, the fourteenth
Assuming that the result of FIG. A is obtained, the fitness conversion unit 1207 calculates
Using the correspondence table in the conversion table storage unit 1208, the relevance of the search result is converted into the corresponding natural language. The result of Fig. 14a is converted to the result of Fig. 14b, and the search result 1208
Output by

According to the fourth aspect described above, the relevance of the search result can be easily perceived by human beings, so that the searcher can easily obtain a more satisfactory search result.

In the fuzzy inference in the first, second, and third embodiments, the min operation is used for calculating the fitness when the antecedent part of the inference rule includes a plurality of conditions. Good. Further, the min operation is used to obtain the output of the consequent part from the conformity of the antecedent part of the inference rule, but another operation may be used. In addition, although the union operation is used to obtain the entire inference result from the inference results of a plurality of inference rules, another operation may be used. In the third invention, it goes without saying that instead of using the α-level set in fuzzy inference, an inference rule of the consequent part real value may be used.

Effect of the Invention As described above, according to the present invention, even a searcher who does not know the contents and structure of a database or a searcher who has only an abstract and ambiguous search condition can obtain an abstract attribute which does not exist in the database. By using ambiguous search conditions including "", data suitable for the search intention can be searched from the database, and the practical effect is great.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of an information retrieval apparatus according to an embodiment of the first invention, FIG. 2 is a diagram showing an example of an inference rule, FIG. 3 is a diagram showing an example of a membership function, and FIG. FIG. 5 is a diagram showing an example of a relation, FIG. 5 is a diagram showing an example of an inference result in the first invention, FIG. 6 is a diagram showing a result of a search example of the first invention, and FIG. FIG. 8 is a view showing a result of a search example of the second invention, FIG. 9 is a view showing a configuration of the information search apparatus in the embodiment of the third invention, and FIG. FIG. 11 is a diagram showing a state of obtaining an α-level set in the third invention, FIG. 11 is a diagram showing a result of a search example of the third invention, and FIG. 12 is a configuration diagram of an information search device in an embodiment of the fourth invention , FIG. 13 is a diagram showing an example of a correspondence table between relevance and a natural language, FIG. 14 is a diagram showing a result of a search example of the fourth invention, and FIG. Configuration diagram of a first example of search device, FIG. 16 shows an example of data in the external storage device, 17
FIG. 18 is a diagram showing an example of a similar word relation used in a conventional information search device. FIG. 18 is a diagram showing a result of a search example of the first example of the conventional information search device. FIG. 6 is a configuration diagram of a second example of FIG. 101 ... external storage device, 102 ... database management system, 103 ... search condition input unit, 104 ... inference rule storage unit,
105: Fuzzy number storage unit, 106: Fuzzy relation storage unit, 107: Fuzzy inference management unit, 108: Antecedent operation unit, 109: Consequence operation unit, 110: Fuzzy matching unit, 111 …… Search result output unit, 701 …… External storage device, 70
2 Database management system 703 Search condition input unit 704 Inference rule storage unit 705 Fuzzy number storage unit 706 Fuzzy relation storage unit 707 Fuzzy inference management unit 708 Consequent part operation part, 709 …… Consequent part operation part,
710: Fuzzy matching unit, 711: Search result output unit, 901: External storage device, 902: Database management system, 903: Search condition input unit, 904: Inference rule storage unit, 905: Fuzzy number Storage unit, 906: Fuzzy relation storage unit, 907: Fuzzy inference management unit, 908: Antecedent operation unit, 909: Consequence operation unit, 910: Search result output unit, 12
01 ... external storage device, 1202 ... database management system, 1203 ... search condition input unit, 1204 ... fuzzy inference unit, 1205 ... fuzzy matching unit, 1206 ... conversion table storage unit, 1207 ... conversion Unit 1208 search result output unit 1501 external storage device 1502 database management system 1503 search condition input unit 1504 similar word storage unit 1505 similar word conversion unit 1506 ... Search result output unit, 1901 external storage device, 1902 database management system, 1903 search condition input unit, 1904 mathematical formula operation unit, 1905 search result output unit.

────────────────────────────────────────────────── ─── Continuation of the front page (56) References JP-A-1-259419 (JP, A) Vandenberghe, R.A. ct. al. "Some practical aspects of fuzzy Database Techniques: AnExample" Information Systems Vol. 14, No. 6 (1989) pp. 465-472 Motohide Umano "Current Status and Trends of Fuzzy Databases", proc Advanced Database Symposium vol. 1989 (1989) pp. 207-214. Tomitoshi Sugimoto, "Prototype of Fuzzy Database", Research Report of Faculty of Engineering, Toyo University. 23 (1987) pp. 79-89 (58) Field surveyed (Int. Cl. ⁶ , DB name) G06F 17/30 JOIS (JICST)

Claims (4)

(57) [Claims] 1. An external storage device for storing data, a database management system for managing data input / output of the external storage device, a search condition input unit for receiving a search condition including a fuzzy number, and storing a fuzzy inference rule. An inference rule storage unit, a fuzzy number storage unit for storing a fuzzy number, a fuzzy relation storage unit for storing a fuzzy relation, and a fuzzy inference using the inference rule and the fuzzy number and inputting the search condition. A fuzzy inference managing unit to be managed, a precondition part calculating unit that obtains a degree of conformity between the search condition and the antecedent part of the inference rule according to an instruction of the fuzzy inference managing unit; and the inference rule according to the degree of conformity. A consequent part calculating part for obtaining an inference result including a fuzzy number from the consequent part; a fuzzy number in the fuzzy number storage part and a fuzzy number in the fuzzy relation storage part. And a fuzzy matching unit for determining the degree of conformity between the inference result and the data in the external storage device, and outputting, as a search result, data in which the degree of conformity determined by the fuzzy matching is equal to or greater than a threshold. An information search device comprising: a search result output unit. 2. An external storage device for storing data, a database management system for managing input / output of data in the external storage device, a search condition input unit for receiving a search condition including a fuzzy number, and storing a fuzzy inference rule. An inference rule storage unit, a fuzzy number storage unit for storing a fuzzy number, a fuzzy relation storage unit for storing a fuzzy relation, and inputting the search condition using the inference rule, the fuzzy number, and the fuzzy relation. A fuzzy inference management unit that manages the fuzzy inference to be performed, a precondition part calculation unit that calculates the degree of conformity between the search condition and the antecedent part of the inference rule according to an instruction of the fuzzy inference management unit, A consequent part operation unit for obtaining an inference result including a fuzzy number from a consequent part of the inference rule, and a fuzzy number in the fuzzy number storage unit and the fuzzy relation. Using the fuzzy relationship in 憶部,
A fuzzy matching unit for determining the degree of conformity between the inference result and the data in the external storage device, and a search result output unit for outputting, as a search result, data in which the degree of conformity determined by the fuzzy matching is greater than or equal to a threshold. An information retrieval device, comprising: 3. An external storage device for storing data, a database management system for managing input / output of data in the external storage device, a search condition input unit for receiving a search condition including a fuzzy number, and storing a fuzzy inference rule. An inference rule storage unit, a fuzzy number storage unit for storing a fuzzy number, a fuzzy relation storage unit for storing a fuzzy relation, and inputting the search condition using the inference rule, the fuzzy number, and the fuzzy relation. A fuzzy inference management unit that manages the fuzzy inference to be performed, a precondition part calculation unit that calculates the degree of conformity between the search condition and the antecedent part of the inference rule according to an instruction of the fuzzy inference management unit, A consequent part operation unit for obtaining a real inference result from the consequent part of the inference rule, and a search result for outputting data matching the inference result as a search result. Information retrieval apparatus characterized by an output unit. 4. An external storage device for storing data, a database management system for managing data input / output of the external storage device, a search condition input unit for receiving a search condition including a fuzzy number, and inputting the search condition. And a fuzzy inference unit that infers an inference result that is a fuzzy number, a fuzzy matching unit that determines the degree of fitness between the inference result and data in the external storage device, and a degree of fitness and natural language that is determined by the fuzzy matching. A conversion table storage unit that stores a correspondence table of the above, a fitness conversion unit that converts the fitness into a natural language using the correspondence table, and outputs data having a fitness equal to or greater than a threshold value as a search result An information search device comprising: a search result output unit.