CN107305490B - Metadata grouping method and device - Google Patents

Abstract

The invention discloses a metadata grouping method and device. The method includes: acquiring source information of metadata in each node of a database; determining the association between each node according to the source information of metadata in each node relationship and degree of association; a relationship network is constructed according to the nodes and the association relationship between the nodes, the edges in the relationship network are the association relationships between the nodes, and the edges in the relationship network are The weight is the degree of association between nodes with an association relationship; the relationship network is divided by particle swarm algorithm to obtain a plurality of sub-relationship networks, wherein each of the sub-relationship networks is a group. In the embodiment of the present invention, after the relationship network is constructed according to each node and the association relationship between the nodes, in the case of lack of metadata data, the metadata can still be grouped, which solves the problem of lack of metadata in the prior art. The problem that the metadata cannot improve the grouping category when processing the description information.

Description

Method and device for grouping metadata

technical field

The present invention relates to the technical field of mobile Internet and automatic information processing, and in particular, to a method and device for sending short messages.

Background technique

With the rapid development of information technology, software is usually designed in a model-driven manner. In data model-driven systems, such data models are often abstractions of a specific type, collectively referred to as metadata. In the era of explosive growth of data volume, the management of data model becomes more and more important. Metadata describes various characteristics of data, including content, status, quality and other dimensions. Therefore, in the era of big data, how to reasonably and efficiently group and manage metadata has become a prominent issue in data quality control.

At present, the widely used metadata grouping management mainly adopts the hierarchical structure of metadata management, and is managed and released by the ERP system software in a unified manner. conflict. The commonly used management methods also include management and control through a graphical interface. This method is implemented from the source of metadata input to the logical model, and the construction of the physical model is realized by a graphical interface, which ensures that the metadata grouping management can be intuitive. In addition, it also includes step-by-step clustering of metadata by building a classification tree, and unified management of the grouped results.

From the perspective of the existing technical solutions, there is a lack of automatic classification methods. Although the graphical interface can intuitively classify, in the big data fusion environment, the metadata also increases geometrically. Therefore, the manual method is completely unable to adapt to the metadata management and control in the existing context. And in the prior art, although hierarchical structure classification management is usually adopted, the hierarchical structure only reflects the similarity to a certain level, but the relationship between metadata cannot be described only by relying on the hierarchical relationship, so it cannot be described from more Considering it from the perspective of depth and breadth, it will inevitably affect the accuracy of subsequent classification management and control. Only using literal attributes to manage metadata lacks consideration of context. It is possible to classify metadata that are literally similar into the same category for management, but they should actually belong to different categories for management.

To sum up, the group management of metadata in the prior art lacks consideration of the context of the metadata, and the grouping category cannot be perfected when the metadata processing description information is missing.

SUMMARY OF THE INVENTION

The present invention provides a method and device for grouping metadata to solve the problem of lack of consideration of the context of metadata in the prior art for grouping management of metadata, and failure to complete the grouping category when metadata processing description information is missing.

The present invention provides a metadata grouping method, comprising:

Obtain the source information of metadata in each node of the database;

Determine the association relationship and association degree between the nodes according to the source information of the metadata in the nodes;

A relational network is constructed according to the nodes and the associations between the nodes, the edges in the relational network are the associations between the nodes, and the weights of the edges in the relational network are associations The degree of correlation between nodes;

The relationship network is divided by the particle swarm algorithm to obtain a plurality of sub-relationship networks, wherein each of the sub-relationship networks is a group.

In the embodiment of the present invention, the association relationship between the nodes formed by the metadata and the association degree between the nodes are determined according to the source information of the metadata in the database, and all nodes in the database are classified according to the association relationship between the node and other nodes. The relationship network is formed, and after determining the relationship network of the metadata, the particle swarm algorithm is used to divide the relationship network, and each sub-network can be obtained, and the sub-network can be regarded as a group. In the embodiment of the present invention, after the relationship network is constructed according to each node and the association relationship between the various nodes, the metadata can still be grouped in the case of lack of metadata data.

Further, determining the association relationship and association degree between the nodes according to the source information of the metadata in the nodes includes:

If a piece of metadata in the first node originates from the second node, determining that there is an association relationship between the first node and the second node;

The number of associations existing between the first node and the second node is determined as the association degree between the first node and the second node, wherein the first node and the second node are Two nodes are any two different tables in the database

In this embodiment of the present invention, if the metadata in one node is from another node, or the metadata in one node is the source of the metadata of another node, it is considered that there is an association relationship between the two nodes, and the metadata is used The relational relationship between the data constructs a relational network, which solves the problem that in the prior art, the relational relation is not considered, but only the description information of the metadata is used for grouping. And the number of associations between two nodes is used as the association degree, which reflects the specific relationship between the input and output of metadata between the two nodes.

Further, the building a relationship network according to the nodes and the associations between the nodes includes:

If there is an association relationship between the first node and the second node, it is determined that an edge exists between the first node and the second node in the relationship network, where the weight of the edge is The degree of association between the first node and the second node.

In the embodiment of the present invention, the association relationship existing between two associated nodes in the relationship network is simplified as an edge in the two nodes, and the weight of the edge is used to represent the relationship between the two nodes. Quantity, which better represents the relationship between two nodes in the relationship network.

Further, before the use of the particle swarm algorithm to divide the relationship network to obtain a plurality of sub-relationship networks, the method further includes:

If there is an invalid node in the relationship network, the invalid node is deleted, wherein the invalid node has a correlation with other nodes in the relationship network that is less than the merging threshold and is only used as the other node in the relationship network. The input source of metadata, or, the invalid node is the correlation with other nodes in the relationship network is less than the merging threshold and the invalid node is only used as the output source of metadata of other nodes in the relationship network, or all The invalid node is zero correlation with other nodes in the relationship network.

In the embodiment of the present invention, the nodes that only output metadata, the nodes that only have metadata input, or the nodes that have no relationship with other metadata are deleted in the relationship network, the relationship network is optimized, and the optimized relationship network is used for grouping, It can further improve the efficiency and accuracy of metadata grouping.

Further, before the use of the particle swarm algorithm to divide the relationship network to obtain a plurality of sub-relationship networks, the method further includes:

If there is a group of nodes in the relationship network that satisfies the following relationship, the group of nodes is merged into one node;

The relationship is that the number of nodes in the group of nodes is not less than 2, and the correlation between every two nodes with an output and input relationship is not less than the merging threshold.

In the embodiment of the present invention, when the nodes in the relationship network that are associated with each other are calculated, and the correlation between the nodes with the associated relationship is calculated to be not less than the merge threshold, it means that the metadata in one of the nodes comes entirely from With the metadata of another node, the eligible nodes need to be merged to further optimize the relationship network.

Further, after dividing the relationship network by using the particle swarm algorithm to obtain a plurality of sub-relationship networks, the method further includes:

The description information with the highest frequency in the sub-relationship network is used as the identification information of the sub-relationship network.

In the embodiment of the present invention, each sub-relationship network is marked with the frequency of occurrence of the description information of each node in each sub-relationship network after grouping, so that operations such as calling and modification can be better performed later.

The present invention also provides a metadata grouping device, comprising:

an acquisition unit, used to acquire source information of metadata in each node of the database;

a determining unit, configured to determine the association relationship and association degree between the nodes according to the source information of the metadata in the nodes;

A relationship network construction unit, configured to construct a relationship network according to each node and the relationship between the nodes, the edges in the relationship network are the relationship between the nodes, and the edges in the relationship network are the relationship between the nodes. The weight of the edge is the degree of association between nodes with an association relationship;

The grouping unit is used to divide the relationship network by using the particle swarm algorithm to obtain a plurality of sub-relationship networks, wherein each of the sub-relationship networks is a group.

In the embodiment of the present invention, the association relationship between the nodes formed by the metadata and the association degree between the nodes are determined according to the source information of the metadata in the database, and all nodes in the database are classified according to the association relationship between the node and other nodes. The relationship network is formed, and after determining the relationship network of the metadata, the particle swarm algorithm is used to divide the relationship network, and each sub-network can be obtained, and the sub-network can be regarded as a group. In the embodiment of the present invention, after the relationship network is constructed according to each node and the association relationship between the various nodes, the metadata can still be grouped in the case of lack of metadata data.

Further, the determining unit is specifically used for:

If a piece of metadata in the first node originates from the second node, determining that there is an association relationship between the first node and the second node;

The number of associations existing between the first node and the second node is determined as the association degree between the first node and the second node, wherein the first node and the second node are Two nodes are any two different nodes in the database.

Further, the determining unit is specifically used for:

If a piece of metadata in the first node originates from the second node, determining that there is an association relationship between the first node and the second node;

The number of associations existing between the first node and the second node is determined as the association degree between the first node and the second node, wherein the first node and the second node are Two nodes are any two different nodes in the database.

Further, the relationship network building unit is specifically used for:

If there is an association relationship between the first node and the second node, it is determined that an edge exists between the first node and the second node in the relationship network, where the weight of the edge is The degree of association between the first node and the second node.

Further, the relationship network building unit is also used for:

If there is an invalid node in the relationship network, the invalid node is deleted, wherein the invalid node has a correlation with other nodes in the relationship network that is less than the merging threshold and is only used as the other node in the relationship network. The input source of metadata, or, the invalid node is the correlation with other nodes in the relationship network is less than the merging threshold and the invalid node is only used as the output source of metadata of other nodes in the relationship network, or all The invalid node is zero correlation with other nodes in the relationship network.

Further, the relationship network building unit is also used for:

If there is a group of nodes in the relationship network that satisfies the following relationship, the group of nodes is merged into one node;

The relationship is that the number of nodes in the group of nodes is not less than 2, and the correlation between every two nodes with an output and input relationship is not less than the merging threshold.

Further, the grouping unit is also used for:

The description information with the highest frequency in the sub-relationship network is used as the identification information of the sub-relationship network.

Description of drawings

In order to illustrate the technical solutions in the embodiments of the present invention more clearly, the following briefly introduces the accompanying drawings used in the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can also be obtained from these drawings without any creative effort.

1 is a schematic flowchart of a method for grouping metadata according to an embodiment of the present invention;

2 is a schematic diagram of nodes and metadata in an embodiment of the present invention;

3 is a schematic diagram of nodes and metadata in an embodiment of the present invention;

4 is a schematic diagram of an input and output relationship of metadata between nodes in a database in an embodiment of the present invention;

5 is a schematic diagram of an input and output relationship of metadata between nodes in a database in an embodiment of the present invention;

6 is a schematic diagram of a relationship network before and after optimization in an embodiment of the present invention;

7 is a schematic diagram of a relationship network after the optimization method is used to optimize the relationship network according to an embodiment of the present invention;

8 is a schematic diagram of an association relationship between nodes in a relationship network before grouping in an embodiment of the present invention;

9 is a schematic diagram of an association relationship between nodes when a relationship network is grouped in an embodiment of the present invention;

10 is a schematic diagram of the association relationship between nodes after the particle swarm algorithm is used to group the relationship network in an embodiment of the present invention;

FIG. 11 is a schematic structural diagram of an apparatus for grouping metadata according to an embodiment of the present invention.

Detailed ways

In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. . Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

An embodiment of the present invention provides a metadata grouping method, as shown in FIG. 1 , including:

Step 101, obtaining source information of metadata in each node of the database;

Step 102, determining the association relationship and association degree between the nodes according to the source information of the metadata in the nodes;

Step 103: Build a relational network according to the nodes and the associations between the nodes, the edges in the relational network are the associations between the nodes, and the weights of the edges in the relational network are: The degree of association between nodes with an association relationship;

Step 104 , using the particle swarm algorithm to divide the relationship network to obtain a plurality of sub-relationship networks, wherein each of the sub-relationship networks is a group.

In step 101, by acquiring the source information of the metadata in each node in the database, the input and output relationship of the metadata in each node can be known. In this embodiment of the present invention, it can be considered that a node is a table stored in the database, Metadata is a field in a table. For example, as shown in Figure 2, a represents a table in a database, and "Name", "Class", and "Address" in table a represent fields, that is, in the present invention In the embodiment, the a table is the node, and the "Name", "Class", and "Address" fields are the metadata in the a table.

In this embodiment of the present invention, the source information of the metadata refers to which node the metadata is output from, or from which layer of the database it is input into the database, and executes several languages for merging metadata, combining several Each metadata constitutes a node, etc., so the source information of metadata can be obtained in various ways, such as obtaining the source information of metadata by obtaining the establishment process of establishing a table in the database, and the present invention provides two kinds of obtaining. Source information for metadata.

method one

In the embodiment of the present invention, the source information of the metadata in the table can be obtained through the establishment process of the table in the database. As shown in FIG. 3, the table b in the database, that is, the node b Address”, “Age”, and “Sex” metadata are composed, as shown in Table 1. The source information of each metadata can be obtained from Table 1.

metadata source information Name node p output Class node p output Address node f output Age node m output Sex node v output

Table 1: Source information table of metadata in node b

Method Two

In the embodiment of the present invention, the source information of the metadata in the node is determined according to the entity relationship diagram between the nodes , metadata 2 comes from node a, metadata 3 comes from metadata e, metadata 4 and metadata 5 come from metadata b.

In the embodiment of the present invention, the above two methods are only used as the description of how to obtain the source information of the metadata, and other methods may also be used to obtain the source information of the metadata.

In the embodiment of the present invention, after obtaining the source information in the node, the association relationship and association degree between the nodes can be obtained. In the embodiment of the present invention, as shown in FIG. 4 and Table 1, node b and Node p, node f, node m, and node v are associated, and node c is associated with node a, node b, node d, and node e;

In this embodiment of the present invention, the degree of association is the number of metadata interactions between nodes that have an association relationship. Optionally, the input and output relationship of metadata between two nodes can be used, that is, the source information of metadata. Determine the degree of association of two nodes, as shown in Table 1, there are two metadata inputs and outputs between node c and node p, so the degree of association can be set to 2, and the relationship between node c and node f, node m, and node v is There is a metadata input and output between, so the correlation can be set to 1.

In the embodiment of the present invention, according to the source information of metadata determined by the above two methods, the association relationship between two nodes with metadata input and output can be obtained. For example, as shown in FIG. 5 , there are nodes v1, There is an association relationship between node v2, node v3, and node v4. The association between node v1 and node v2 is defined as one edge, and the two association relationships between node v2 and node v3 are defined as two edges. The association between nodes v3 and v4 is defined as an edge.

In this embodiment of the present invention, optionally, a node in the database may be defined, that is, a table in the database is a quadruple, that is, a node includes source information of metadata in the node, a set of output metadata in the node, and a node The combination of input metadata and the textual description of the node. For example, node a is composed of four metadata: "Name", "Score1", "Score2", and "Score3". Node a also includes the "Name" field from The output of node m, the "Score1" field comes from the output of node n, "Score2" and "Score3" are fields that exist when node a is established, "Name" and "Score1" in node a are the set of input fields, " The fields "Score2" and "Score3" are sets of output fields, and the literal description of node a is "score table".

In this embodiment of the present invention, a weight can be used to represent the degree of association between two nodes. As shown in Table 1, the degree of association between node c and node p is 2, and the weight between node c and node p can be set as 2. The degree of association between node c and node f, node m, and node v is 1, and the weight between node c and node f, node m, and node v can be set to 1, or according to the database shown in Figure 5. In the entity relationship diagram between nodes, the weight between node v1 and node v2 can be set to 1, the weight between node v2 and node v3 can be set to 2, and the weight between node v3 and node v4 can be set to 1 .

In this embodiment of the present invention, the weight of an edge can also be determined by the correlation between two nodes. For example, in FIG. 5, node v3 has two metadata derived from node v2, that is, two of the three metadata of v2 The output is given to v3, that is, the correlation of this edge can be determined to be 2/3.

In this embodiment of the present invention, a relational network between nodes in the database can be constructed according to the edges between nodes and the weights between nodes and nodes. As shown in FIG. 5 , there is a relational network in the database. There is an association relationship between v1 and v2, and the weight of the edge between v1 and v2 is 1, and there is an association relationship between nodes v2 and v3, and the weight of the edge between v2 and v3 is 2, and node v3 and node v4 There is a relationship between them, and the weight of the edge between v3 and v4 is 1.

In the embodiment of the present invention, the nodes in the database are connected through edges and weights to form a relationship network with metadata input and output relationships between nodes, so as to avoid the lack of attribute descriptions through metadata in the prior art, Problems with limited grouping of metadata.

Further, after the relationship network between nodes in the database is established, the relationship network needs to be optimized. In the embodiment of the present invention, several methods for optimizing the relationship network are provided to optimize the relationship network.

method one

As shown in Figure 6, (a) in Figure 6 is the relationship network before optimization. There are two edges between node v2 and node v3. In order to simplify the number of edges in the relationship network, weights are added to the description of the edges. Indicates how many edges there are between two related nodes. For example, there are two edges between nodes v2 and v3, and the weight of the edge between nodes v2 and v3 is 2, then the simplified relationship network is shown in Figure 6. (b) As shown in the figure, there is only one edge between nodes v2 and v3, and the weight of this edge is 2, so there is only one edge between every two related nodes in the simplified relationship network.

Method Two

In the embodiment of the present invention, the optimization of the relationship network also needs to delete invalid nodes. In this embodiment of the present invention, the invalid node refers to that the correlation with other nodes in the relationship network is less than the merging threshold and is only used as an input source of metadata of other nodes in the relationship network, or the invalid node is The correlation with other nodes in the relationship network is less than the merge threshold and the invalid node is only used as an output source of metadata of other nodes in the relationship network, or the invalid node is related to other nodes in the relationship network. Correlation is zero.

In this embodiment of the present invention, the merging threshold refers to the possibility that two associated nodes can be merged into one node, and the correlation between the two nodes is not less than the merging threshold, indicating that the two nodes can be merged into one If the correlation between the two nodes is less than the merge threshold, it means that the two nodes cannot be merged into one node.

In this embodiment of the present invention, the correlation between two nodes is the value of the input-output relationship of metadata between the two nodes. For example, as shown in FIG. 7 , there is a set of nodes in the database, node vv2 and node vv3 , there is an edge between nodes vv2 and vv3, and the weight is 3, which means that all metadata in node vv3 comes from node vv2, or all metadata in node vv2 comes from node vv3, then the edge between nodes vv2 and vv3 The correlation is 2/3.

Optionally, in this embodiment of the present invention, since the correlation calculation between two nodes is determined by mathematical calculation, the value of the correlation calculation result is a positive number less than or equal to 1, and the merging threshold can be set as 0.8, that is, if the value of the modulus of the calculation result of the correlation between the two nodes is greater than or equal to 0.8, it is considered that the two nodes can be merged into one node.

In the embodiment of the present invention, there are three kinds of invalid nodes, and these invalid nodes will affect the division effect in the division of the relationship network, so these invalid nodes need to be deleted. One is that there is a node whose degree of association with other nodes in the database is less than the merging threshold, that is, the node cannot be merged with other nodes, and all metadata in this node is used for output to other nodes.

The other is that there is a node whose degree of association with other nodes in the database is less than the merging threshold, that is, the node cannot be merged with other nodes, and all the metadata in this node comes from the output of other nodes.

Another is that after the relationship network is established, there are still some nodes that are not associated with other nodes, that is, there is no metadata input and output relationship between the node and other nodes, and the degree of association between other nodes in the relationship network. If it is zero, the metadata cannot be grouped according to the input and output relationship of the metadata, so delete this type of node.

Method three

In the embodiment of the present invention, there is a set of nodes in the relationship network, the number of nodes is not less than n, the edge is not less than n-1, and the correlation between two nodes corresponding to each edge is not less than the merging threshold , all nodes in the group are merged into one node.

For example, as shown in Figure 7, if the merge threshold is 0.5, the edge correlation between node vv2 and node vv3 is 2/3, 2/30.5, so nodes vv2 and vv3 can be merged into one node.

Optionally, in this embodiment of the present invention, the correlation between nodes may also be calculated by the following formula:

Among them, m _out is the metadata output from the node to another node, and min is the metadata input by another node to the node. _In this embodiment of the present invention, the input-output relationship of the metadata in the node can be represented by a vector Indicates that, for example, in the two nodes shown in FIG. 7, the node vv2 outputs three metadata to the node vv3, and the node vv3 outputs three metadata to the node vv2. In this embodiment of the present invention, the metadata can be mapped to In the hash table, and then modulo the mapped hash value, a series of vectors calculated for each metadata can be obtained. For example, the field of the metadata name is calculated by the hash function, and then the calculation result is modulo. The modulo result will be mapped to a cell of the vector. For example, the result of the hash function calculation is divided by 1000 modulo, then there are 1000 cells. If the modulo result is in the first Within 900 cells, the value of the position of the cell is 1, and other positions are 0, so it will form (1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 0... ) vector.

Alternatively, in this embodiment of the present invention, a vector may be set to represent the input or output of metadata in the node, setting 0 means that the metadata in the node does not originate from other nodes, nor is it output to other nodes, setting 1 means that the node The metadata in is input to metadata in other nodes, or the metadata is output to other nodes.

In the embodiment of the present invention, the vector of the metadata input to the node vv3 is (1, 0, 1, 1), and the output vector of the node vv3 to the node vv2 is (1, 0, 1, 1). The formula calculates that the correlation between vv3 and vv2 is 1, then the nodes vv2 and vv3 can be merged.

In this embodiment of the present invention, before calculating the node correlation, it is also necessary to calculate whether the nodes need to be split. For example, the node x1 outputs four "a", "b", "c", and "d" to the node x3 Metadata, node x2 outputs two metadata "e" and "f" to node x3, and node x3 includes "a", "b", "c", "d", "e", "f" six Metadata, then node x3 can be split into node x31 and node x32. After splitting, all fields in node x31 come from the output of node x1, then merge node x1 and node x31. After splitting, node x32 All fields in are from the output of node x2, then merge nodes x2 and x32.

In the embodiment of the present invention, after optimizing the established relationship network, the relationship network should be grouped. The grouping avoids the problem of large time overhead caused by repeated fusion between nodes and groups in the traditional grouping aggregation process.

In the embodiment of the present invention, the process of grouping relational networks is as follows:

As shown in Figure 8, there is a relational network in the database, which consists of nodes A, B, C, and D. According to the relationship between the edges of the nodes in Figure 8, it can be determined that the initial position of the particle in the particle swarm algorithm is X1=(B , A, B, C), X2=(B, C, B, C), X3=(B, A, D, C), X4=(B, C, D, C), in the embodiment of the present invention , since there are only four nodes in one relational network, the initial position of the particle in the particle swarm algorithm is the connection relation between the four nodes. If the relational network in the database is the connection relation between multiple nodes, Then the number of initial positions of the particles also increases accordingly.

In the embodiment of the present invention, after determining the initialization position of the particle, it is also necessary to determine the initialization speed of the particle. The initialization speed is related to the weight of each edge in the node. The initialization velocity of the particle is v1=(B,C,B,D)v2=(B,A,B,D)v3=(B,C,D,D)v4=(B,A,D,D).

After the initialization speed of the particle is determined, it is assumed that the A initialization of the particle X1 selects B, the B initialization selects A, the C initialization selects D, and the D initialization selects C. Then the particle x1 divides the following graph into two groups, as shown in Figure 9, and the weights on the node edges remain unchanged.

则ΔQ_v用经典BGLL算法进行计算，具体如下式所述：Then, according to the classical BGLL algorithm, the values of the individual fitness and the global fitness in the particle swarm are calculated. In this embodiment of the present invention, the calculation of the individual fitness value can be described as

Then ΔQ _v is calculated by the classical BGLL algorithm, as described in the following formula:

是关系网中群体的内部边的权重，例如，如图9所示，群1内部边的权重为1，群2内部边的权重为2，

是关系网中所有群体的权重，即群1与群2的权重和为3，W_i是所有节点i相关的边权重，例如，节点A相关的边权中为1，W_i，in是节点i与群相连接的所有边的权重，m表示关系网中所有边的权重之和，在图9中由于只有两个节点，所以节点A与群相连接的所有边的权重还是等于1。In the embodiment of the present invention, c represents the acceleration constant, v represents the population number,

is the weight of the internal edge of the group in the relationship network. For example, as shown in Figure 9, the weight of the internal edge of group 1 is 1, and the weight of the internal edge of group 2 is 2.

is the weight of all groups in the relationship network, that is, the sum of the weights of group 1 and group 2 is 3, and Wi is the edge weight related to all node _i , for example, the edge weight related to node A is 1, and Wi _{, in} are the nodes i is the weight of all edges connected to the group, and m represents the sum of the weights of all edges in the relational network. In Figure 9, since there are only two nodes, the weight of all edges connected to the group by node A is still equal to 1.

In the embodiment of the present invention, after calculating ΔQ _v , the value of individual fitness is calculated, and the value of global fitness is fit _g =max(p _best ).

After each iteration, the velocity of the particles in the particle swarm needs to be updated, and the update algorithm is described as follows:

即粒子在新的群中的个体适应度最大值除以种群中所有粒子的个体适应度的和，即在粒子的处于m位置时，所有可能的粒子最优适应度进行平均化，然后按照随机值的方式选择适合粒子的m位置的速度。在对粒子群中的粒子速度进行更新后，调整粒子的位置。in

That is, the maximum individual fitness of the particle in the new group is divided by the sum of the individual fitness of all particles in the group, that is, when the particle is at the m position, the optimal fitness of all possible particles is averaged, and then randomly The value of the method chooses the velocity appropriate for the m position of the particle. Adjust the position of the particles after updating the particle velocities in the particle swarm.

The above steps are repeated until fit _n -fit _n-1 <ε. In the embodiment of the present invention, ε is the convergence threshold. When fit _n -fit _n-1 <ε, ε is very small. Therefore, it is considered that the grouping is completed when the position of the particles in the population no longer changes greatly.

Further, in this embodiment of the present invention, in order to better manage the grouped metadata, it is also necessary to mark each group after the grouping, and the group name of each marked group is the mark of the metadata.

In this embodiment of the present invention, the description information with the highest occurrence frequency in each sub-relationship network obtained after grouping is used as the identification information of the sub-relationship network. For example, as shown in FIG. 10 , seven nodes in the relationship network are divided into two sub-relationship networks In the first sub-network, the description information of node A is the achievement, the description information of node B is the achievement, and the description information of node E is the math score, and the description information of the node F is the average score, so the score can be used as the identification information of the first subnet.

Optionally, in this embodiment of the present invention, the TF-IDF method may be used to identify the sub-relationship network. The main idea of the method is that if a word or phrase has a high TF in one article and rarely appears in other articles, it is considered that the word or phrase has a good ability to distinguish between categories and is suitable for classification. TF is an abbreviation of Term Frequency, which refers to the frequency with which a given word appears in the document. IDF is an acronym for Inverse Document Frequency, a measure of the general importance of a word. The IDF for a particular word can be obtained by dividing the total number of documents by the number of documents containing the word, and then taking the logarithm of the quotient obtained. In this embodiment of the present invention, TF can be represented by the following formula:

In this embodiment of the present invention, the numerator represents the number of occurrences of the word in the file, and the denominator represents the sum of the number of occurrences of all words in the file.

In this embodiment of the present invention, the IDF can be represented by the following formula:

In the embodiment of the present invention, TF represents the frequency of the entry in any sub-network. For example, as shown in FIG. 10 , TF represents the frequency that the description information of node A is the score in the first sub-network. In the embodiment of the present invention, the TF of the score is 2/4=0.5, and the IDF represents the frequency that the description information of the entry in all nodes A appears in all the sub-relationship networks, and then takes the logarithm of the frequency. In the embodiment of the present invention , after grouping the nodes in the database, 5000 sub-relationship networks are obtained, and the description information of grades has appeared in 50 sub-relationship networks, then the IDF of the grades is log(5000/50)=2, then according to TF- The TF-IDF value of the description information of the grade calculated by the IDF method is 0.5*2=1.

In the embodiment of the present invention, the description information of the nodes is sorted according to the calculated TF-IDF value of the description information of each node. The higher the TF-IDF value is, the higher the sorting position is, and the sorting position is the first after sorting. The description information is used as the identification information of the sub-network.

Optionally, in this embodiment of the present invention, binary and ternary word segmentation may be performed on the description information of the nodes in the sub-relationship network first, and then the TF-IDF values of the binary and ternary word segmentations may be calculated. And the TF-IDF value of the trigram is sorted. The higher the sorting position is, the first binary or trigram after the sorting is used as the identification information of the sub-relationship network.

For example, as shown in Figure 10, the description information of node F is the average grade, then the description information of F is the average grade, and the binary word segmentation is "average" and "grade", and the description information of F is the average grade. The participles are "average achievement" and "achievement", and the TF-IDF values of "average", "grade", "average achievement" and "grade" are calculated respectively. Sort by the TF-IDF value of "Grade", the higher the sorting position is, the first "Average", "Grade", "Average Grade" or "Grade" in the sorting position will be used as the identifier of the sub-relationship network information.

The present invention also provides a metadata grouping device, as shown in Figure 11, comprising:

Obtaining unit S1001, for obtaining source information of metadata in each node of the database;

Determining unit S1002, configured to determine the association relationship and association degree between the nodes according to the source information of the metadata in the nodes;

A relationship network construction unit S1003, configured to construct a relationship network according to the nodes and the association relationship between the nodes, the edges in the relationship network are the association relationships between the nodes, and the edges in the relationship network The weight of the edge is the degree of association between nodes with an association relationship;

The grouping unit S1004 is configured to use the particle swarm algorithm to divide the relationship network to obtain a plurality of sub-relationship networks, wherein each of the sub-relationship networks is a group.

Further, the determining unit S1002 is specifically used for:

If a piece of metadata in the first node originates from the second node, determining that there is an association relationship between the first node and the second node;

The number of associations existing between the first node and the second node is determined as the association degree between the first node and the second node, wherein the first node and the second node are Two nodes are any two different nodes in the database.

Further, the relationship network construction unit S1003 is specifically used for:

If there is an association relationship between the first node and the second node, it is determined that an edge exists between the first node and the second node in the relationship network, where the weight of the edge is The degree of association between the first node and the second node.

Further, the relationship network construction unit S1003 is also used for:

If there is an invalid node in the relationship network, the invalid node is deleted, wherein the invalid node has a correlation with other nodes in the relationship network that is less than the merging threshold and is only used as the other node in the relationship network. The input source of metadata, or, the invalid node is the correlation with other nodes in the relationship network is less than the merging threshold and the invalid node is only used as the output source of metadata of other nodes in the relationship network, or all The invalid node is zero correlation with other nodes in the relationship network.

Further, the relationship network construction unit S1003 is also used for:

If there is a group of nodes in the relationship network that satisfies the following relationship, the group of nodes is merged into one node;

The relationship is that the number of nodes in the group of nodes is not less than 2, and the correlation between every two nodes with an output and input relationship is not less than the merging threshold.

Further, the grouping unit S1004 is also used for:

The description information with the highest frequency in the sub-relationship network is used as the identification information of the sub-relationship network.

The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block in the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to the processor of a general purpose computer, special purpose computer, embedded processor or other programmable data processing device to produce a machine such that the instructions executed by the processor of the computer or other programmable data processing device produce Means for implementing the functions specified in a flow or flow of a flowchart and/or a block or blocks of a block diagram.

These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory result in an article of manufacture comprising instruction means, the instructions The apparatus implements the functions specified in the flow or flow of the flowcharts and/or the block or blocks of the block diagrams.

These computer program instructions can also be loaded on a computer or other programmable data processing device to cause a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process such that The instructions provide steps for implementing the functions specified in the flow or blocks of the flowcharts and/or the block or blocks of the block diagrams.

Although preferred embodiments of the present invention have been described, additional changes and modifications to these embodiments may occur to those skilled in the art once the basic inventive concepts are known. Therefore, the appended claims are intended to be construed to include the preferred embodiment and all changes and modifications that fall within the scope of the present invention.

It will be apparent to those skilled in the art that various modifications and variations can be made in the present invention without departing from the spirit and scope of the invention. Thus, provided that these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims (10)

1. A metadata grouping method, wherein the method comprises: Obtain the source information of metadata in each node of the database; Determine the association relationship and association degree between the nodes according to the source information of the metadata in the nodes; A relational network is constructed according to the nodes and the associations between the nodes, the edges in the relational network are the associations between the nodes, and the weights of the edges in the relational network are associations The degree of correlation between nodes; The relationship network is divided by particle swarm algorithm to obtain a plurality of sub-relationship networks, wherein each of the sub-relationship networks is a group; The determining the association relationship and association degree between the nodes according to the source information of the metadata in the nodes includes: If a piece of metadata in the first node originates from the second node, determining that there is an association relationship between the first node and the second node; The number of associations existing between the first node and the second node is determined as the association degree between the first node and the second node, wherein the first node and the second node are Two nodes are any two different nodes in the database. 2. The method according to claim 1, wherein the constructing a relationship network according to the nodes and the associations between the nodes comprises: If there is an association relationship between the first node and the second node, it is determined that an edge exists between the first node and the second node in the relationship network, where the weight of the edge is The degree of association between the first node and the second node. 3. The method according to claim 1, characterized in that before the use of particle swarm algorithm to divide the relationship network to obtain a plurality of sub-relationship networks, the method further comprises: If there is an invalid node in the relationship network, the invalid node is deleted, wherein the invalid node has a correlation with other nodes in the relationship network that is less than the merging threshold and is only used as the other node in the relationship network. The input source of metadata, or, the invalid node is the correlation with other nodes in the relationship network is less than the merging threshold and the invalid node is only used as the output source of metadata of other nodes in the relationship network, or all The invalid node has zero correlation with other nodes in the relationship network. 4 . The method according to claim 1 , wherein, before dividing the relationship network by using particle swarm algorithm to obtain a plurality of sub-relationship networks, the method further comprises: 5 . If there is a group of nodes in the relationship network that satisfies the following relationship, the group of nodes is merged into one node; The relationship is that the number of nodes in the group of nodes is not less than 2, and the correlation between every two nodes with an output and input relationship is not less than the merging threshold. 5 . The method according to claim 1 , wherein after dividing the relationship network by using particle swarm algorithm to obtain a plurality of sub-relationship networks, the method further comprises: 6 . The description information with the highest frequency in the sub-relationship network is used as the identification information of the sub-relationship network. 6. A device for grouping metadata, comprising: an acquisition unit, used to acquire source information of metadata in each node of the database; a determining unit, configured to determine the association relationship and association degree between the nodes according to the source information of the metadata in the nodes; A relationship network construction unit, configured to construct a relationship network according to each node and the relationship between the nodes, the edges in the relationship network are the relationship between the nodes, and the edges in the relationship network are the relationship between the nodes. The weight of the edge is the degree of association between nodes with an association relationship; a grouping unit, configured to divide the relationship network by using the particle swarm algorithm to obtain a plurality of sub-relationship networks, wherein each of the sub-relationship networks is a group; The determining unit is specifically used for: If a piece of metadata in the first node originates from the second node, determining that there is an association relationship between the first node and the second node; The number of associations existing between the first node and the second node is determined as the association degree between the first node and the second node, wherein the first node and the second node are Two nodes are any two different nodes in the database. 7. The device according to claim 6, wherein the relationship network building unit is specifically used for: If there is an association relationship between the first node and the second node, it is determined that an edge exists between the first node and the second node in the relationship network, where the weight of the edge is The degree of association between the first node and the second node. 8. The device according to claim 6, wherein the relationship network construction unit is further used for: If there is an invalid node in the relationship network, the invalid node is deleted, wherein the invalid node has a correlation with other nodes in the relationship network that is less than the merging threshold and is only used as the other node in the relationship network. The input source of metadata, or, the invalid node is the correlation with other nodes in the relationship network is less than the merging threshold and the invalid node is only used as the output source of metadata of other nodes in the relationship network, or all The invalid node is zero correlation with other nodes in the relationship network. 9. The apparatus according to claim 6, wherein the relationship network construction unit is further used for: If there is a group of nodes in the relationship network that satisfies the following relationship, the group of nodes is merged into one node; The relationship is that the number of nodes in the group of nodes is not less than 2, and the correlation between every two nodes with an output and input relationship is not less than the merging threshold. 10. The apparatus according to claim 6, wherein the grouping unit is further used for: The description information with the highest frequency in the sub-relationship network is used as the identification information of the sub-relationship network.

Images