JP2020154991A - Information management device and information management method - Google Patents

Abstract

To hierarchically manage information reflecting how it should be used.SOLUTION: An item extraction section 1 extracts an item from documents D1-D4.., and creates a hierarchical structure of a node assigned to the item. A node candidate creation section 2 standardizes names of items having the same concept that have been extracted from the documents D1-D4.., on the basis of morphological analysis and synonym analysis. A node extraction section 3 extracts a predetermined node from the hierarchical structure of nodes. A node integration section 4 integrates abstraction degrees of the items having the same concept that exhibit different abstraction degrees of a lower node associated with the predetermined node of the same concept. A classification section 5 classifies the predetermined nodes according to the items of the lower node associated with the predetermined nodes. A modeling section 6 estimates a model of information contents of the lower node, on the basis of elements of the information contents of the lower node associated with the predetermined node. A node division section 7 divides the items having the same concept of the lower node that are associated respectively with the predetermined nodes classified into different groups, into concrete items that are specific to each group.SELECTED DRAWING: Figure 1

Description

The present invention relates to an information management device and an information management method capable of managing information in a hierarchical manner.

In order to make it easier for users to obtain necessary information, a technique for classifying information hierarchically has been proposed.

For example, in Patent Document 1, the route pattern extraction unit specifies a route including a category including a concept selected by the comparison concept selection unit in the information classification hierarchy, and is higher than the concept of each category included in the route. Information on how it relates to the concept of the category is set, and the concept of each category is abstracted except for the user-specified concept input in the input reception unit to generate a route pattern and category. The generation unit generates candidate categories by replacing the concept of categories included in the route pattern so as to satisfy the above-set information, the control unit adds the candidate categories to the information classification hierarchy, and the output unit provides information. A technique for outputting a classification hierarchy is disclosed.

Japanese Unexamined Patent Publication No. 2012-43212

However, the conventional information classification hierarchy is constructed by considering only the fluctuation of the document notation, and the usage of items is not considered. For this reason, even if the items are the same, the description contents may differ depending on the document, and it may take time and effort to acquire the necessary information.

The present invention has been made in view of the above circumstances, and an object of the present invention is to provide an information management device and an information management method capable of hierarchically managing information reflecting how it is used.

In order to achieve the above object, the information management device according to the first aspect includes an extraction unit that extracts a predetermined node from the hierarchical structure of the nodes assigned to the conceptualized information, and the predetermined unit that is extracted by the extraction unit. It is provided with a classification unit that classifies the predetermined node extracted by the extraction unit based on the information of the lower node associated with the node.

According to the present invention, information that reflects how it is used can be managed hierarchically.

FIG. 1 is a block diagram showing a configuration example of the information management device according to the embodiment. FIG. 2 is a diagram showing an example of the document of FIG. FIG. 3 is a diagram showing another example of the document of FIG. FIG. 4 is a diagram showing still another example of the document of FIG. FIG. 5 is a diagram showing still another example of the document of FIG. FIG. 6A is a diagram showing a hierarchical structure of nodes to which the document items of FIG. 2 are assigned, and FIG. 6B is a diagram showing a hierarchical structure of nodes to which the document items of FIG. 3 are assigned. is there. FIG. 7 (a) is a diagram showing an integration example based on the semantic analysis of the concept of the hierarchical structure node of FIG. 6 (a), and FIG. 7 (b) is a diagram of the concept of the hierarchical structure node of FIG. 6 (b). FIG. 7 (c) is a diagram showing an integrated example of the concept based on the semantic analysis, and FIG. 7 (c) is a diagram showing an integrated example based on the semantic analysis of the concept of the node having a hierarchical structure based on the item of the document of FIG. 4 (b). 8 (a) is a diagram showing an example of extracting a predetermined node to be integrated or divided of lower nodes for the hierarchical structure of FIG. 6 (b), and FIG. 8 (b) is an item of the document of FIG. It is a figure which shows the extraction example of the predetermined node which is the target of integration or division of the lower node about the hierarchical structure based on. 9 (a) is a diagram showing an integration example based on the abstraction degree analysis of the concept of the hierarchical structure node of FIG. 7 (a), and FIG. 9 (b) is the concept of the hierarchical structure node of FIG. 8 (a). 9 (c) is a diagram showing an integration example of the concept based on the abstraction degree analysis of FIG. 7 (c), and FIG. 9 (d) is a diagram showing an integration example based on the abstraction degree analysis of the concept of the node of the hierarchical structure of FIG. 7 (c). Is a diagram showing an integration example based on the abstraction degree analysis of the concept of the node of the hierarchical structure shown in FIG. 8 (b). FIG. 10 is a diagram showing an example of a correspondence relationship between items and information contents extracted from the document of FIG. FIG. 11 (a) is a diagram showing a classification example of patterns of how to associate lower nodes associated with the habitat of each document, and FIG. 11 (b) is an example of a mathematical model of pattern P1 of FIG. 11 (a). 11 (c) is a diagram showing an example of a mathematical model of pattern P2 of FIG. 11 (a), and FIG. 11 (d) is a diagram showing an example of a mathematical model of pattern P3 of FIG. 11 (a). Is. FIG. 12A shows an example of dividing the lower node based on the pattern P1 of FIG. 11B, and FIG. 12B shows an example of dividing the lower node based on the pattern P2 of FIG. 11C. FIG. 12 (c) is a diagram showing an example of division of lower nodes based on the pattern P3 of FIG. 11 (d). 13 (a) is a diagram showing another extraction example of a predetermined node to be integrated or divided of lower nodes based on the hierarchical structure of FIG. 7 (a), and FIG. 13 (b) is FIG. 8 (b). It is a figure which shows the other extraction example of the predetermined node which is the target of integration or division of the lower node based on the hierarchical structure of. FIG. 14 is a flowchart showing pattern classification and mathematical modeling processing based on the substructure according to the embodiment. FIG. 15 is a flowchart showing a specific example of the process of S18 in FIG. FIG. 16 is a flowchart showing an example of node division processing based on the abstraction degree reset according to the embodiment. FIG. 17 is a flowchart showing another example of the node division process based on the abstraction degree reset according to the embodiment. FIG. 18 is a flowchart showing still another example of the node division process based on the abstraction degree reset according to the embodiment. FIG. 19 is a block diagram showing a hardware configuration example of the information management device of FIG.

The embodiment will be described with reference to the drawings. It should be noted that the embodiments described below do not limit the invention according to the claims, and all of the elements and combinations thereof described in the embodiments are essential for the means for solving the invention. Not necessarily.

The information management device according to the embodiment classifies information based on how the notation of the document is used. At this time, the information management device extracts a predetermined node from the hierarchical structure of the nodes assigned to the conceptualized information, and classifies the predetermined node based on the information of the lower node associated with the predetermined node. Nodes are assigned, for example, document items. The node may be assigned a document heading or a document title. The node may be assigned an item name such as a form.

Hereinafter, the information management device according to the embodiment will be described by taking the case where the document items are assigned to the nodes as an example. In the following explanation, when the operation subject is described as "○○ part is", the processor reads the program ○○ part, loads it into the DRAM (Dynamic Random Access Memory), and then realizes the function of the ○○ part. It shall mean that.

FIG. 1 is a block diagram showing a configuration example of the information management device according to the embodiment.
In FIG. 1, the information management device includes an item extraction unit 1, a node candidate generation unit 2, a node extraction unit 3, a node integration unit 4, a classification unit 5, a modeling unit 6, a node division unit 7, a thesaurus dictionary 8, and a conceptual model. 9 is provided.

The item extraction unit 1 extracts items from documents D1 to D4, and generates a hierarchical structure of nodes to which the items are assigned. At this time, the item extraction unit 1 uses the description of the documents D1 to D4 ... as it is as the item name attached to the node. Therefore, even if the item names given to the nodes have the same concept, the notation may vary.

The node candidate generation unit 2 unifies the names of the items of the same concept extracted from the documents D1 to D4 ... Based on the morphological analysis and the synonym analysis. At this time, the node candidate generation unit 2 can refer to the thesaurus dictionary 8. Further, the node candidate generation unit 2 modifies the hierarchical structure of the nodes based on the inclusion relationship of the words extracted from the documents D1 to D4. For example, when the concept of the lower node associated with the predetermined node is a modifier that is not included in the concept of the predetermined node, the node candidate generation unit 2 can aggregate the lower node into the predetermined node.

The node extraction unit 3 extracts a predetermined node from the hierarchical structure of the nodes. For example, the node extraction unit 3 can extract nodes having one or less levels of lower node hierarchy as predetermined nodes. By extracting the nodes whose lower node hierarchy is one level or less as the predetermined node, it is possible to facilitate the pattern classification based on the item of the lower node associated with the predetermined node.

The node integration unit 4 integrates the degree of abstraction of the items of the lower node associated with the predetermined node. At this time, the node integration unit 4 can refer to the conceptual model 9. As a result, even if the items are in the same hierarchy, the item name described in the upper concept and the item name described in the lower concept can be matched.

The classification unit 5 classifies the predetermined node based on the item of the lower node associated with the predetermined node. At this time, the classification unit 5 can classify the predetermined node based on the combination of the concepts of the lower nodes associated with the predetermined node. For example, the classification unit 5 can classify the first node and the second node into different groups when the concept of the lower node associated with the first node cannot be the concept of the lower node associated with the second node. .. As a result, even when the item name of the first node and the item name of the second node are the same, it can be determined that the item of the first node and the item of the second node are used differently.

Here, it can be determined that the items assigned to the predetermined nodes classified into different groups are used differently even if the notation is the same, and the search for information reflecting the usage of the items is performed. Efficiency can be improved. For this reason, we support the creation of documents optimized for each type of report, design document, proposal, dissertation, internal and external, etc., and support the creation of dissertation with a good acceptance rate. Can be done.

The modeling unit 6 estimates a model of how the lower node is associated based on the information of the lower node associated with the predetermined node. At this time, the modeling unit 6 can generate a pattern of how the lower nodes are associated with the predetermined nodes classified in the same group by the classification unit 5. This pattern can indicate the degree of cohesion or the degree of variation of the lower nodes associated with the predetermined nodes assigned to the items extracted from the plurality of documents D1 to D4.

When estimating the model of how the lower node is associated, the modeling unit 6 can refer to the information content of the lower node associated with the predetermined node. For example, the modeling unit 6 can estimate a model of how the lower node is associated based on the element of the information content of the lower node associated with the predetermined node. The element of the information content of the lower node is, for example, a word included in the information content of the lower node. The model of how to associate the information of the lower node may be constructed based on the amount of information of the information content of the lower node, or may be constructed based on the similarity of the elements of the information content of the lower node.

The node dividing unit 7 divides the items of the lower nodes associated with the predetermined nodes classified into different groups into specific items peculiar to each group, and outputs the hierarchical structure of the nodes for each group. At this time, the node dividing unit 7 can divide the items of the lower node associated with the predetermined node based on the model estimated by the modeling unit 6. As a result, even when the items of the same concept are extracted from the documents D1 to D4, the item names can be different according to the difference in how these items are used, and the difference in how the items are used. Can be realized in the search that reflects.

The thesaurus dictionary 8 is a dictionary in which words are classified based on the similarity of meanings. The conceptual model 9 is a model showing the vertical relationship between concepts. At this time, the upper layer can have a higher degree of abstraction than the lower layer. As the conceptual model 9, for example, an ontology can be used.

Hereinafter, the processing of the information management device of FIG. 1 will be specifically described by taking an actual document as an example.
FIG. 2 is a diagram showing an example of the document of FIG.
In FIG. 2, document D1 is given the title of anemone fish ecology. Document D1 includes items such as habitat, breeding method, foraging method, gender and survival time. Habitat items include water quality, depth, temperature, symbiosis and habitat areas.

The item of water quality includes the information content of seawater. The item "water depth" includes the information content of 20-40 m. The item "temperature" includes the information content of 24 degrees. The item of symbiosis includes the information content of sea anemones. The item Habitat includes information about the Indo-Pacific and near the equator.

FIG. 3 is a diagram showing another example of the document of FIG.
In FIG. 3, document D2 is given the title of Dobiuo ecology. Document D2 includes items such as habitat, breeding method, feeding method, gender and longevity. The item habitat includes items such as water quality, depth, temperature and habitat.

The item of water quality includes the information content of seawater. The item "water depth" includes the information content of 1 m. The habitat section includes information about the Pacific Ocean, Indian Ocean, and Atlantic Ocean.

FIG. 4 is a diagram showing still another example of the document of FIG.
In FIG. 4, document D3 is given the title of panda ecology. Document D3 includes items such as morphology, habitat, breeding method, foraging method and longevity. The item morphology includes the items size, hair and bark. The item of size includes the items of total length and weight. The item barking includes the items male and female. The item habitat includes the item country name and habitat. The item habitat includes the items temperate and bamboo grove.

The item of weight includes the information content of kg. The item "Osu" includes the information content "Meow Meow". The item "female" includes the information content "myanmyan". The item "country name" includes the information content "China"

FIG. 5 is a diagram showing still another example of the document of FIG.
In FIG. 5, document D4 is given the title of lion ecology. Document D4 includes items such as morphology, habitat, breeding method, foraging method, social system and longevity. The item morphology includes the items size, hair and bark. The item of size includes the items of total length and weight. The item habitat includes the item country name and habitat. The item Habitat includes the items Subtropical and Grassland. The item "country name" includes the information content of Africa.

FIG. 6A is a diagram showing a hierarchical structure of nodes to which the document items of FIG. 2 are assigned, and FIG. 6B is a diagram showing a hierarchical structure of nodes to which the document items of FIG. 3 are assigned. is there.
In FIG. 6A, the item extraction unit 1 extracts titles and items from the document D1 of FIG. Then, the item extraction unit 1 assigns the node N111 to the title of the ecology of anemone fish.

The item extraction unit 1 assigns nodes N121 to N125 to the items of habitat, breeding method, foraging method, sex, and survival time, respectively. The item extraction unit 1 assigns nodes N131 to N135 to the items of water quality, water depth, temperature, symbiosis, and habitat, respectively. The item extraction unit 1 associates the nodes N121 to N125 with the node N111, and associates the nodes N131 to N135 with the node N121.

In FIG. 6B, the item extraction unit 1 of FIG. 1 extracts a title and an item from the document D2 of FIG. Then, the item extraction unit 1 assigns the node N211 to the title of the ecology of flying fish.

The item extraction unit 1 assigns nodes N221 to N225 to the items of habitat, breeding method, feeding method, gender and longevity, respectively. The item extraction unit 1 assigns nodes N231 to N234 to the items of water quality, water depth, temperature, and habitat, respectively. The item extraction unit 1 associates the nodes N221 to N225 with the node N211 and associates the nodes N231 to N234 with the node N221.

Here, the item of the foraging method of the node N123 in FIG. 6A and the item of the inoculation method of the node N223 in FIG. 6B have the same concept, but the item extraction unit 1 describes the document D1. , D2 notation is used as it is. Further, the item of the survival time of the node N125 in FIG. 6A and the item of the lifespan of the node N225 in FIG. 6B have the same concept, but the item extraction unit 1 describes the documents D1 and D2. Is used as it is.

FIG. 7 (a) is a diagram showing an integration example based on the semantic analysis of the concept of the hierarchical structure node of FIG. 6 (a), and FIG. 7 (b) is a diagram of the concept of the hierarchical structure node of FIG. 6 (b). FIG. 7 (c) is a diagram showing an integrated example of the concept based on the semantic analysis, and FIG. 7 (c) is a diagram showing an integrated example based on the semantic analysis of the concept of the node having a hierarchical structure based on the item of the document of FIG. 4 (b).
In FIG. 7A, the node candidate generation unit 2 of FIG. 1 extracts an item of ecology from the title of anemone fish ecology of node N111 based on morphological analysis, and changes the name of node N111 to the item of ecology. .. Further, the node candidate generation unit 2 changes the item of the survival period of the node N125 to the item of the lifespan based on the synonym analysis.

In FIG. 7B, the node candidate generation unit 2 extracts the item "ecology" from the title of the flying fish ecology of the node N211 based on the morphological analysis, and changes the name of the node N211 to the item "ecology". Further, the node candidate generation unit 2 changes the item of the feeding method of the node N223 to the item of the breeding method based on the synonym analysis.

As a result, the node candidate generation unit 2 can integrate the notations of the items of the same concept even when the notations of the items of the same concept are different in the documents D1 and D2.

In FIG. 7 (c), the item extraction unit 1 of FIG. 1 extracts a title and an item from the document D3 of FIG. Then, the item extraction unit 1 assigns the node N311 to the title of panda ecology.

The item extraction unit 1 assigns nodes N321 to N325 to the items of habitat, morphology, foraging method, breeding method, and longevity, respectively. The item extraction unit 1 assigns nodes N331 to N335 to the items of country name, habitat, size, hair and bark, respectively. The item extraction unit 1 assigns nodes N341 to N346 to the items of temperate zone, bamboo grove, total length, weight, male and female, respectively. The item extraction unit 1 associates nodes N321 to N325 with node N311, associates nodes N331 and N322 with node N321, associates nodes N333 to N335 with node N322, and associates nodes N341 and N342 with node N332. Nodes N343 and N344 are associated with node N333, and nodes N345 and N346 are associated with node N335.

Here, the item extraction unit 1 sets a temporary item X1 of the upper concept of the temperate zone for the item of the temperate zone of the node N341, and sets a temporary item X2 of the higher concept of the bamboo forest for the item of the bamboo forest of the node N342. Can be set.

Next, the node candidate generation unit 2 extracts the item "ecology" from the title of the panda ecology of the node N311 based on the morphological analysis, and changes the name of the node N311 to the item "ecology". Further, the node candidate generation unit 2 determines whether or not the concept of the item of the node N345 and the concept of the item of the node N346 are included in the concept of the item of the bark of the node N335. Further, the node candidate generation unit 2 determines whether or not the information content of the node N345 as meow and the information content of the node N346 as myanmyan are included in the concept of the item of the bark of the node N335.

Then, in the node candidate generation unit 2, the concept of the item of the node N345 and the concept of the item of the node N346 are not included in the concept of the item of the bark of the node N335, but the information content of the meow of the node N345 and the information content When the information content of node N346 is included in the concept of the item of barking of node N335, the item of node N345 and the item of node N346 are judged to be mere modifiers, and nodes N345 and N346 are noded. Consolidate to N335.

8 (a) is a diagram showing an example of extracting a predetermined node to be integrated or divided of lower nodes for the hierarchical structure of FIG. 6 (b), and FIG. 8 (b) is an item of the document of FIG. It is a figure which shows the extraction example of the predetermined node which is the target of integration or division of the lower node about the hierarchical structure based on.
In FIG. 8A, the node extraction unit 3 of FIG. 1 has a lower node N231 to which the lower layer structure is one stage when the hierarchical structure of the node of FIG. 7B is generated by the node candidate generation unit 2. The node N221 associated with N234 is extracted.

In FIG. 8B, the item extraction unit 1 of FIG. 1 extracts a title and an item from the document D4 of FIG. Then, the item extraction unit 1 assigns the node N411 to the title of lion's ecology.

The item extraction unit 1 assigns nodes N421 to N426 to the items of habitat, morphology, foraging method, breeding method, longevity and social system, respectively. The item extraction unit 1 assigns nodes N431 to N435 to the items of country name, habitat, size, hair, and bark, respectively. The item extraction unit 1 assigns nodes N441 to N444 to the items of subtropical, grassland, total length, and weight, respectively. The item extraction unit 1 associates nodes N421 to N426 with node N411, associates nodes N431 and N422 with node N421, associates nodes N433 to N435 with node N422, and associates nodes N441 and N442 with node N432. Nodes N443 and N444 are associated with node N433.

Here, the item extraction unit 1 sets a temporary item Y1 of the subtropical superordinate concept for the item of the subtropical node N441, and sets a temporary item Y2 of the superordinate concept of the grassland for the item of the grassland of the node N442. Can be set.

When the item extraction unit 1 generates the hierarchical structure of the nodes shown in FIG. 8B, the node extraction unit 3 can extract the nodes N432 associated with the lower nodes N441 and N442 having one lower layer structure. it can. Alternatively, the node extraction unit 3 may extract the node N433 associated with the lower nodes N443 and N444 whose lower layer structure is one stage.

9 (a) is a diagram showing an integration example based on the abstraction degree analysis of the concept of the hierarchical structure node of FIG. 7 (a), and FIG. 9 (b) is the concept of the hierarchical structure node of FIG. 8 (a). 9 (c) is a diagram showing an integration example of the concept based on the abstraction degree analysis of FIG. 7 (c), and FIG. 9 (d) is a diagram showing an integration example based on the abstraction degree analysis of the concept of the node of the hierarchical structure of FIG. 7 (c). Is a diagram showing an integration example based on the abstraction degree analysis of the concept of the node of the hierarchical structure shown in FIG. 8 (b).

In FIG. 9A, the node extraction unit 3 of FIG. 1 extracts the node N121 associated with the lower nodes N131 to N135 having the lower layer structure of one stage from the hierarchical structure of the nodes of FIG. 7A. To do. Further, in FIG. 9B, it is assumed that the node extraction unit 3 extracts the node N221 associated with the lower nodes N231 to N234 having the lower layer structure in one stage from the hierarchical structure of the nodes in FIG. 7B. ..

Then, the node integration unit 4 integrates the item of the habitat area of the lower node N135 in FIG. 9A into the item of the habitat area based on the abstraction degree analysis. As a result, the item name of the lower node N135 in FIG. 9A can be matched with the item name of the lower node N234 in FIG. 9B, and the fluctuation of the notation of the lower node can be eliminated.

Further, in FIG. 9C, it is assumed that the node extraction unit 3 extracts the node N332 associated with the lower nodes N341 to N342 having the lower layer structure in one stage from the hierarchical structure of the nodes in FIG. 7C. .. Further, in FIG. 9D, it is assumed that the node extraction unit 3 extracts the node N432 associated with the lower nodes N441 to N442 having the lower layer structure in one stage from the hierarchical structure of the nodes in FIG. 8B. ..

Then, the node integration unit 4 integrates the temporary item X1 of the lower node N341 into the item of climate and the temporary item X2 of the lower node N342 into the item of vegetation based on the abstraction degree analysis. Further, the node integration unit 4 integrates the temporary item Y1 of the lower node N441 into the item of climate and the temporary item Y2 of the lower node N442 into the item of vegetation based on the abstraction degree analysis. As a result, the item names of the lower nodes N341 and N342 in FIG. 9C can be matched with the item names of the lower nodes N441 and N442 in FIG. 9B, respectively, and the fluctuation of the notation of the lower node can be eliminated. can do.

FIG. 10 is a diagram showing an example of a correspondence relationship between items and information contents extracted from the document of FIG.
In FIG. 10, in the conceptual model 9, for example, the subordinate concepts of seawater, steam and freshwater are associated with the superordinate concept of water quality, and the subordinate concepts of the Indo-Pacific, the sea area near the equator, the Indian Ocean, the Pacific Ocean, and the East Asian rivers are referred to as habitats. Associate the superordinate concept, the temperate and subtropical subconcepts with the climate superordinate concept, and the grassland and bamboo forest subordinate concepts with the vegetation superordinate concept.

By referring to the concept model 9, the node integration unit 4 can integrate the item names of the concepts of the lower nodes having different degrees of abstraction. For example, document D1 of FIG. 2 describes the item of habitat for the information content of the Indo-Pacific and the vicinity of the equator. At this time, in the conceptual model 9 of FIG. 10, the item of habitat is associated with the information content of the Indo-Pacific and the sea area near the equator. Therefore, the node integration unit 4 can integrate the item of the habitat area of the lower node N135 of FIG. 9A into the item of the habitat area by referring to the conceptual model 9 of FIG.

Further, in the conceptual model 9, the item of climate is associated with the information content of temperate zone and subtropical zone, and the item of vegetation is associated with the information content of grassland and bamboo forest. Therefore, the node integration unit 4 integrates the provisional items X1 and Y1 of the lower nodes N341 and N441 of FIGS. 9 (c) and 9 (d) into the item of climate by referring to the conceptual model 9 of FIG. Then, the provisional items X2 and Y2 of the lower nodes N342 and N442 can be integrated into the item called vegetation.

FIG. 11 (a) is a diagram showing a classification example of patterns of how to associate lower nodes associated with the habitat of each document, and FIG. 11 (b) is an example of a mathematical model of pattern P1 of FIG. 11 (a). 11 (c) is a diagram showing an example of a mathematical model of pattern P2 of FIG. 11 (a), and FIG. 11 (d) is a diagram showing an example of a mathematical model of pattern P3 of FIG. 11 (a). Is.

In FIG. 11A, for example, regarding a document relating to the ecology of anemone fish, flying fish, dolphin, sweetfish, medaka, panda and lion, the node extraction unit 3 of FIG. 1 assumes that the item of habitat is extracted as a predetermined node. ..

At this time, regarding the document on the ecology of bear flies, there are water quality, water depth, temperature, habitat and symbiosis as items of lower nodes linked to the item of habitat, and for the document on the ecology of Tobiuo, the item of habitat. The items of the subordinate nodes associated with it are water quality, water depth, temperature and habitat, and for the document on the ecology of dolphins, the items of the subordinate nodes associated with the item of habitat are water quality, water depth and temperature. ..

In addition, regarding the document on the ecology of medaka, there are water quality, habitat and velocity as items of the lower node linked to the item of habitat, and for the document on the ecology of medaka, the item of the lower node linked to the item of habitat. It is assumed that the items are water quality and flow velocity.

Furthermore, regarding the document on the ecology of pandas and the document on the ecology of lions, it is assumed that the items of the lower nodes associated with the item of habitat are climate and vegetation.

Then, the classification unit 5 of FIG. 1 classifies the item of habitat in each document on the ecology of anemone fish, flying fish, dolphin, sweetfish, medaka, panda and lion based on the item of the lower node associated with the item of habitat. To do. At this time, the classification unit 5 can use, for example, the distance between the vectors when the items of the lower nodes of each document are vectorized as an index for classifying the item of habitat in each document.

At this time, the classification unit 5 can generate a vector to which a component of 1 or 0 is added depending on the presence or absence of the item of the lower node. For example, the classification unit 5 generates a vector (1,1,1,1,1,0,0,0) for bear flies and (1,1,1,1,0,0) for tobiuo. , 0,0) is generated, for dolphins, the vector (1,1,1,0,0,0,0,0) is generated, and for sweetfish, (1,0,0,1) is generated. , 0,1,0,0), for medaka, (1,0,0,0,0,1,0,0), for pandas and lions, (0) , 0,0,0,0,0,1,1) is generated.

For anemone fish, flying fish and dolphins, the distance between the vectors is 1 or 2. For sweetfish and medaka, the distance between the vectors is 1. For pandas and lions, the distance between the vectors is zero. Clownfish, flying fish and dolphins are more than three distances from sweetfish and killifish. Clownfish, flying fish and dolphins are more than five distances from pandas and lions. Ayu and medaka are at least 4 distances from pandas and lions.

Therefore, by setting the threshold of the distance between the vectors of the lower nodes to 3, the classification unit 5 classifies the item of the habitat associated with the lower nodes whose distance between the vectors is smaller than 3 into the same group, and the vectors. The item of habitat associated with lower nodes with a distance of 3 or more can be classified into another group.

Note that the classification unit 5 classifies the first node and the second node into different groups when the concept of the lower node associated with the first node cannot be the concept of the lower node associated with the second node. May be good. For example, the item Climate and Vegetation for the ecology of pandas and lions cannot be the item for the ecology of anemone fish, flying fish, dolphins, sweetfish and medaka. Therefore, the item of habitat for pandas and lions can be classified into a different group from the item of habitat for clownfish, flying fish, dolphins, sweetfish and medaka. In addition, the item of flow velocity for the ecology of sweetfish and medaka cannot be the item for the ecology of anemone fish, flying fish and dolphin. Therefore, the item of habitat for sweetfish and medaka can be classified into a different group from the item of habitat for anemone fish, flying fish and dolphin.

Next, the modeling unit 6 has a habitat pattern P1 showing how to link to the item habitat for anemone fish, flying fish and dolphin, and a habitat pattern P2 showing how to tie to the item habitat for sweetfish and medaka. And generate a habitat pattern P3 that shows how to tie to the item habitat for pandas and lions.

At this time, the modeling unit 6 can estimate a mathematical model for each habitat pattern P1 to P3 based on the information of the lower nodes associated with each habitat pattern P1 to P3. In the mathematical model of each habitat pattern P1 to P3, for example, the existence probability of the subordinate items, the cohesiveness of the subordinate nodes of each habitat pattern P1 to P3, or the distribution model of the information associated with each subordinate item can be used. As the information associated with the subordinate items, items or information contents further subordinate to the subordinate items can be used. The degree of cohesion of the lower nodes can be calculated based on the variance of the existence probability of the lower items for each habitat pattern P1 to P3. The degree of cohesion of the lower nodes may be obtained based on the average distance from the representative vector of the vectors belonging to each habitat pattern P1 to P3.

For example, as shown in FIG. 11B, in the habitat pattern P1, the existence probabilities of the items of water quality, water depth, temperature, habitat and symbiosis are 1.0, 1.0, 1.0 and 0, respectively. 67, 0.33. As a result, the cohesiveness of the habitat pattern P1 is 0.45. In addition, in the item of habitat area of habitat pattern P1, it is possible to generate a distribution model in which the information content of the Pacific Ocean exists at a ratio of 0.5 and the information content of the Indian Ocean exists at a ratio of 0.3. ..

Further, as shown in FIG. 11C, the existence probabilities of the items of water quality, habitat area and flow velocity in the habitat pattern P2 are 1.0, 0.5 and 1.0, respectively. As a result, the degree of cohesion for the habitat pattern P2 is 0.7. Also, in the item of water quality of habitat pattern P2, if the word freshwater appears but the words brackish water and seawater do not appear, (freshwater, brackish water, seawater) = (1.0, 0.0, 0.0). ) Can be generated. Furthermore, if the word East Asian river appears in the item of habitat pattern P2 habitat, but the word river does not appear other than that, a distribution model of (East Asian river) = (1.0) is generated. can do.

Further, as shown in FIG. 11D, the existence probabilities of the items of climate and vegetation in the habitat pattern P3 are 1.0 and 1.0, respectively. As a result, the degree of cohesion for the habitat pattern P3 is 1.0. In addition, when the words subtropical and temperate appear evenly in the item of climate of vegetation pattern P3, a distribution model of (subtropical, temperate) = (0.5,0.5) is generated, and vegetation pattern P3 If the words bamboo grove and grassland appear evenly in the item of vegetation, a distribution model of (bamboo grove, grassland) = (0.5, 0.5) can be generated.

FIG. 12A shows an example of dividing the lower node based on the pattern P1 of FIG. 11B, and FIG. 12B shows an example of dividing the lower node based on the pattern P2 of FIG. 11C. FIG. 12 (c) is a diagram showing an example of division of lower nodes based on the pattern P3 of FIG. 11 (d).

In FIG. 12A, the node dividing section 7 of FIG. 1 divides the items of the lower nodes associated with the habitat pattern P1 into specific items specific to the habitat pattern P1. For example, the node division 7 changes the item of habitat pattern P1 habitat to the item of sea area when only the information content representing the sea area such as the Pacific Ocean and the Indian Ocean appears in the item of habitat pattern P1 habitat. To do.

Further, in FIG. 12B, when the node division portion 7 shows the distribution model of (East Asian river) = (1.0) in the item of the habitat area of the habitat pattern P2, the habitat of the habitat pattern P2 Change the item "Region" to the item "River area".

Here, by dividing the items of the lower nodes associated with the habitat patterns P1 and P2 into specific items specific to each habitat pattern P1 and P2, for example, it is assumed that a paper is written about a freshwater fish called carp. , Ayu and Medaka ecology habitat patterns P2 can be referred to. For this reason, when writing a dissertation on freshwater fish called carp other than sweetfish and medaka, it is possible to prevent the item of flow velocity from falling out from the item of habitat, and the item of habitat is an extra item such as water depth. Can be prevented from being inserted, and the quality of the paper can be improved.

Further, in FIG. 12C, the node division 7 maintains the item of climate of habitat pattern P3 as it is in the item of climate of habitat pattern P3 when there is no bias in the information content embodying the climate. .. Further, the node dividing unit 7 maintains the item of vegetation of habitat pattern P3 as it is in the item of vegetation of habitat pattern P3 when there is no bias in the information content embodying the vegetation.

Here, if there is no bias in the information content of the item of the lower node, by maintaining the abstraction level of the item as it is, when writing a paper about animals living in the boreal zone or animals living in the forest, the habitat pattern P3 Can be referred to.

13 (a) is a diagram showing another extraction example of a predetermined node to be integrated or divided of lower nodes based on the hierarchical structure of FIG. 7 (a), and FIG. 13 (b) is FIG. 8 (b). It is a figure which shows the other extraction example of the predetermined node which is the target of integration or division of the lower node based on the hierarchical structure of.

In FIG. 13A, the node extraction unit 3 of FIG. 1 extracts a predetermined node from the hierarchical structure of the nodes reflecting the processing result of the classification unit 5 of FIG. For example, the node extraction unit 3 sets the lower nodes N131 to N135 in the item of the habitat of the node N121 in FIG. 7A as the pattern PA as the information content of the node N121. At this time, the hierarchy of the lower nodes N121 to N125 of the node N111 to which the item of ecology is assigned becomes one stage. Therefore, the node extraction unit 3 can extract the node N111 to which the item of ecology is assigned as a predetermined node by extracting the nodes whose lower node hierarchy is one stage or less. As a result, the item of ecology that was not extracted from the hierarchical structure of FIG. 7A can also be subject to pattern classification by the classification unit 5.

Further, in FIG. 13B, the node extraction unit 3 sets the lower nodes N441 and N442 of the item of the habitat of the node N432 in FIG. 8B as the pattern PB as the information content of the node N432. At this time, the hierarchy of the lower nodes N431 and N432 of the node N421 to which the item of habitat is assigned becomes one stage. Therefore, the node extraction unit 3 can extract the node N421 to which the item of habitat is assigned as a predetermined node by extracting the nodes whose lower node hierarchy is one stage or less. As a result, the item of habitat that was not extracted from the hierarchical structure of FIG. 8B can also be subject to pattern classification by the classification unit 5.

FIG. 14 is a flowchart showing pattern classification and mathematical modeling processing based on the substructure according to the embodiment.
In FIG. 14, the node name N to be analyzed and the subordinate node name list of the node name N are acquired (S11).

Then extracted node node name N from all documents, the lower node vector of the node i as v _i, extracting corresponding respectively to all nodes that calculates a lower node vector v _i (S12). The lower node vector v _i is a vector that quantifies the correspondence between the lower node group actually associated with the node i having the node name N and the node information described in the lower node name list M of the node name N. ..

Next, clustering descendants vector v _i, node K of the node name N extracted from all documents (K is a positive integer) are classified into pieces of a group (S13). Any clustering method can be used for clustering. For example, the number of classifications can be determined in advance, classified by the K-means method, and the threshold value regarding the similarity between vectors can be arbitrarily set to perform hierarchical clustering.

Next, k = 1 is set (S14).

Next, the nodes belonging to the k-th cluster group as k pattern node of the node name N, imparts ^{k N} is the group id (S15).

Then, ^k the mean vector of the lower node vectors _{v i} of ^N groups of nodes, ^{k N} each lower node to a node of the group is calculated as the cord attached probability vector _{P kN} (S16).

Furthermore, as ^{k N} group node variation of the index sigma _kN of calculating the variance of the lower node vectors _{v i} of ^{k N} group of nodes (S17).

Next, to estimate the mathematical model of the information content that is actually stored in each lower node k ^N group of nodes (S18).

Next, k = k + 1 is set (S19).

Next, it is determined whether k ≦ K (S20). If k ≦ K, the process returns to S15, and if k ≦ K, the process ends.

FIG. 15 is a flowchart showing a specific example of the process of S18 in FIG.
In Figure 15, ^{k N} lower node name number tied to a node of the group M (M is a positive integer) to acquire (S31).

Next, m = 1 is set (S32).

Next, the node s of ^{k N} group (s = 1, ..., S (S is a positive integer)) with respect to the information content _y ^s m stored in the lower node m (s = 1, ..., S) Is extracted (S33). At this time, if the node s is not associated with the lower node m, y _s ^m = 0.

Next, it is determined whether or not the base mathematical model Y _m exists (S34). For example, the base model of a node to which the item height is assigned can be a normal distribution.

Then, if the mathematical model _{Y m} as a base is present, the information content _y ^s parameters of the mathematical model _{Y m} calculated on the basis of ^m, mathematical models for ^k information content lower nodes m of ^N group _y ^{s m} _{Y m} Calculate ^kN (S35) and proceed to S40.

On the other hand, if the mathematical model _{Y m} as a base is absent, it calculates the presence probability _p ^m kN (z) in the information content _y ^{s m} of each element z, which may be stored in the lower node m (S36).

Next, the information content _y ^{s m} a plurality of elements z determines whether there simultaneously (S37).

When a plurality of elements z exist at the same time in the information content y _s ^m , the mathematical model Y _m ^kN is calculated by taking the sum of z * p _m ^kN (z) for those elements z (S38), and the process proceeds to S40. ..

On the other hand, when a plurality of elements z do not exist at the same time in the information content y _s ^m , the vector P _m ^kN relating to all the elements z of the existence probability p _m ^kN (z) is stored in the mathematical model Y _m ^kN (S39).

Next, m = m + 1 is set (S40).

Next, it is determined whether m ≦ M (S41). If m ≦ M, the process returns to S32, and if m ≦ M, the process ends.

FIG. 16 is a flowchart showing an example of node division processing based on the abstraction degree reset according to the embodiment.
In FIG. 16, in a plurality of groups k ^N having different k, lower nodes u (u = 1, ..., U (U is a positive integer)) each having D (D is a positive integer) or more data are extracted. (S51). D is an arbitrarily set threshold value. For example, when D = 1, in the example of FIG. 11C, the nodes of water quality and habitat are extracted.

Next, u = 1 is set (S52).

Next, the elements that can be stored in the lower node u (elements with p> 0) are compared between the groups, and the elements of the target group are explained, but the target is the concept name of the maximum abstraction that does not include the elements of other groups. The node name of the node u related to the group is reset (S53).

Next, u = u + 1 is set (S54).

Next, it is determined whether u ≦ U (S55). If u ≦ U, the process returns to S52, and if u ≦ U, the process ends.

FIG. 17 is a flowchart showing another example of the node division process based on the abstraction degree reset according to the embodiment.
In FIG. 17, m = 1 is set (S61).

Next, for the lower node m of the node name N, the node name list L given before the abstraction is acquired (S62).

Next, k = 1 is set (S63).

Next, the node name in the list L, to compare the information content y _s ^m which is stored in the lower node m in the group k ^N, the node name lowest abstract of information content y _s ^m encompassing the node name Reset to the node name of the lower node m (S64).

Next, k = k + 1 is set (S65).

Next, it is determined whether k ≦ K (S66). If k ≦ K, the process returns to S63, and if k ≦ K, the process proceeds to S67.

Next, m = m + 1 is set (S67).

Next, it is determined whether m ≦ M (S68). If m ≦ M, the process returns to S61, and if m ≦ M, the process ends.

FIG. 18 is a flowchart showing still another example of the node division process based on the abstraction degree reset according to the embodiment.
In FIG. 18, m = 1 is set (S71).

Next, k = 1 is set (S72).

Next, the number X of the elements included in the data in which the information content y _s ^m stored in the lower node m in the group k ^N is not 0 is calculated (S73). In the number X, if a plurality of elements exist in the information content y _s ^m , all of them are added.

Next, it is determined whether or not the base mathematical model _Yo exists for the node o that is conceptually equivalent to or lower than the node m (S74).

If the underlying mathematical model Y _o is present, the number of elements X data set to obtain a threshold value for determining whether belonging to each mathematical model Y _o (S75).

Then, for each mathematical model Y _o, a group k ^N calculates the probability that information content y _s ^m which is stored in the lower node m belongs in the lowest level of abstraction node of mathematical models Y _o below the threshold Reset o to the node name of the lower node m (S76), and proceed to the process of S79.

On the other hand, when the base mathematical model _Yo does not exist, the threshold value of the number of element types of the target concept as a reference for determining whether the number of elements belongs to a certain concept is acquired (S77). ..

Next, among the nodes including the information content y _s ^m stored in the lower node m in the group k ^N, the node o which is the lowest node among the nodes having the number of element types below the threshold value is the node of the lower node m. Reset to the first name (S78).

Next, k = k + 1 is set (S79).

Next, it is determined whether k ≦ K (S80). If k ≦ K, the process returns to S72, and if k ≦ K, the process proceeds to S81.

Next, m = m + 1 is set (S81).

Next, it is determined whether m ≦ M (S82). If m ≦ M, the process returns to S71, and if m ≦ M, the process ends.

FIG. 19 is a block diagram showing a hardware configuration example of the information management device of FIG.
In FIG. 19, the information management device 101 includes a processor 11, a communication control device 12, a communication interface 13, a main storage device 14, and an external storage device 15. The processor 11, the communication control device 12, the communication interface 13, the main storage device 14, and the external storage device 15 are connected to each other via the internal bus 16. The main storage device 14 and the external storage device 15 are accessible from the processor 11.

Further, an input device 20 and an output device 21 are provided outside the information management device 101. The input device 20 and the output device 21 are connected to the internal bus 16 via the input / output interface 17. The input device 20 is, for example, a keyboard, a mouse, a touch panel, a card reader, a voice input device, or the like. The output device 21 is, for example, a screen display device (liquid crystal monitor, organic EL (Electro Luminescence) display, graphic card, etc.), an audio output device (speaker, etc.), a printing device, and the like.

The processor 11 is hardware that controls the operation of the entire information management device 101. The processor 11 may be a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit). The processor 11 may be a single core losser or a multi-core losser. The processor 11 may include a hardware circuit (for example, an FPGA (Field-Programmable Gate Array) or an ASIC (Application Specific Integrated Circuit)) that performs a part or all of the processing. The processor 11 may include a neural network.

The main storage device 14 can be composed of, for example, a semiconductor memory such as SRAM or DRAM. The main storage device 14 can store a program being executed by the processor 11 or provide a work area for the processor 11 to execute the program.

The external storage device 15 is a storage device having a large storage capacity, and is, for example, a hard disk device or an SSD (Solid State Drive). The external storage device 15 can hold an executable file of various programs and data used for executing the program. The information management program 15A can be stored in the external storage device 15. The information management program 15A may be software that can be installed in the information management device 101, or may be incorporated as firmware in the information management device 101.

The communication control device 12 is hardware having a function of controlling communication with the outside. The communication control device 12 is connected to the network 19 via the communication interface 13. The network 19 may be a WAN (Wide Area Network) such as the Internet, a LAN (Local Area Network) such as WiFi or Ethernet (registered trademark), or a mixture of WAN and LAN. May be good.

The input / output interface 17 converts the data input from the input device 20 into a data format that can be processed by the processor 11, and converts the data output from the processor 11 into a data format that can be processed by the output device 21. ..

The processor 11 reads the information management program 15A into the main storage device 14 and executes the information management program 15A to extract a predetermined node from the hierarchical structure of the nodes assigned to the conceptualized information and associate it with the predetermined node. Predetermined nodes can be classified based on the information of the subordinate nodes. At this time, the processor 11 realizes the functions of the item extraction unit 1, the node candidate generation unit 2, the node extraction unit 3, the node integration unit 4, the classification unit 5, the modeling unit 6, and the node division unit 7 in FIG. Can be done.

The execution of the information management program 15A may be shared by a plurality of processors and computers. Alternatively, the processor 11 may instruct a cloud computer or the like to execute all or a part of the information management program 15A via the network 19 and receive the execution result.

The present invention is not limited to the above-described embodiment, and includes various modifications. For example, the above-described embodiment has been described in detail in order to explain the present invention in an easy-to-understand manner, and is not necessarily limited to those having all the described configurations. Further, it is possible to replace a part of the configuration of one embodiment with the configuration of another embodiment, and it is also possible to add the configuration of another embodiment to the configuration of one embodiment. Further, it is possible to add / delete / replace a part of the configuration of each embodiment with another configuration. Further, each of the above configurations, functions, processing units, processing means and the like may be realized by hardware by designing a part or all of them by, for example, an integrated circuit.

1 item extraction unit, 2 node candidate generation unit, 3 node extraction unit, 4 node integration unit, 5 classification unit, 6 modeling unit, 7 node division unit, 8 thesaurus dictionary, 9 conceptual model

Claims (15)

An extraction unit that extracts a predetermined node from the hierarchical structure of the nodes assigned to the conceptualized information,
An information management device including a classification unit that classifies the predetermined node extracted by the extraction unit based on the information of a lower node associated with the predetermined node extracted by the extraction unit. The information management device according to claim 1, wherein the node is assigned a document item. The information management device according to claim 1, wherein the predetermined node is a node in which the hierarchy of the lower node is one stage or less. The information management device according to claim 3, wherein the extraction unit considers the predetermined node classified into the same group in the classification unit and a lower node associated with the predetermined node as a single node. The information management device according to claim 1, wherein the classification unit classifies the predetermined node based on a combination of concepts of the lower node associated with the predetermined node. The extraction unit extracts the first node and the second node as the predetermined node,
The classification unit requests that the first node be classified into a group different from the second node when the concept of the lower node associated with the first node cannot be the concept of the lower node associated with the second node. Item 1. The information management device according to item 1. The information management device according to claim 1, further comprising a modeling unit that models how the subordinate nodes are associated with each other based on the information of the subordinate nodes that are associated with the predetermined node classified by the classification unit. The information management device according to claim 7, wherein the modeling unit generates a pattern of how the lower nodes are associated with the predetermined node classified into the same group by the classification unit. The information management device according to claim 1, further comprising a division unit that divides the concept of a lower node associated with the predetermined node based on the classification result of the predetermined node. The information management device according to claim 9, wherein the division unit divides the concept of a subordinate node associated with the predetermined node classified into different groups into a concrete concept peculiar to each group. The information management device according to claim 1, further comprising an integration unit that integrates the degree of abstraction of the concept of a lower node associated with the predetermined node. The information management device according to claim 11, wherein the integration unit integrates the concept of the lower node based on the information content of the lower node. The information management device according to claim 2, further comprising a node candidate generation unit that unifies the names of items of the same concept extracted from the document based on morphological analysis and synonym analysis. Information management method executed by the processor
The processor
Extract the specified node from the hierarchical structure of the node to which the item is assigned,
An information management method for classifying the predetermined node based on the item of the lower node associated with the predetermined node. The processor
Based on morphological analysis and synonym analysis, unify the names of items of the same concept extracted from the document,
Integrate the abstraction level of the items of the lower node associated with the predetermined node,
The predetermined node is classified based on the item of the lower node in which the abstraction level is integrated.
Based on the information of the lower node associated with the predetermined node, a model of how the lower node is associated is estimated.
The information management method according to claim 14, wherein the items of the lower node associated with the predetermined node are divided based on the model of the lower node association method.

Images