CN103390058B - The domain knowledge browsing method of knowledge based map - Google Patents

Abstract

The present invention proposes a kind of domain knowledge browsing method of knowledge based map, the method can be navigated to the user being unfamiliar with domain knowledge.The method is divided into two stages, that is: Knowledge Map builds and important knowledge identification; Wherein have three data objects, that is: domain knowledge base, domain knowledge map and important knowledge, important knowledge is used for the starting point browsing knowledge base to knowledge user, it is knowledge user navigation that domain knowledge map is used in knowledge navigation process, and domain knowledge base provides the idiographic flow of knowledge; Method in the present invention does not need other information except domain knowledge base, can generate Knowledge Map and important knowledge from domain knowledge base; Adopt latent semantic analysis technique construction domain knowledge map can go up the noise data existed in removal field by a relatively large margin, the semantic connection features of outstanding domain knowledge map.

Description

The domain knowledge browsing method of knowledge based map

Technical field

The present invention relates to the method that knowledge is browsed, particularly relate to a kind of domain knowledge browsing method of knowledge based map.

Background technology

Information overload phenomenon refers to because too much information is supplied to information user, causes information user cannot carry out processing the phenomenon of (understanding content or find interested content).This phenomenon extensively exists in internet information is browsed.For improving information overload, adopt more method mainly two kinds at present, a kind of is adopt the method for information visualization to make presenting of information more directly perceived, and another kind adopts monitor user ' navigation patterns and the method for making a prediction.

Information visualization methods mainly by adopting text classification clustering technique content to be classified automatically, and adopts multidimensional scaling technology that classified information is mapped to two dimension or three dimensions, utilizes point in space to represent information.This method, with mode exhibition information intuitively, can reduce the generation of information overload phenomenon very well.But simultaneously along with the increase of quantity of information, adopt visualization method can bring visual overload (VisualOverload).Visual overload refers to and present the phenomenon that too much information causes information browse difficulty in certain space.

Detect the problem of information overload brought under user browsing behavior can be good at solving bulk information situation, adopt the monitoring technology of running background can not bring new operating load to information user.But application the method needs the behavioral data of long monitor user ', this behavioural information to be can not obtain or can not carry out collecting and processing due to safety problem under many circumstances.In addition simple making can only be applicable to information user basic understanding relevant information in this way, and the situation of browsed related content.

Existing assistant browsing technology is mainly used in internet information and browses auxiliary, can not support that domain knowledge is browsed well.Mainly due to the reason of following several aspect: (1) existing information visualization methods mainly supports the visual of internet information, the information such as the content of webpage, linking relationship each other and webpage unit language in visualization technique implementation procedure, can be utilized.And in domain knowledge unless the context outside, there is no other available information, therefore often can not realize effective information visualization effect; (2) domain knowledge does not often carry out unified management, but exists in the form of a file, and the collection therefore for user behavior is more difficult.The present invention for be the information user not understanding a certain field, therefore the behavioural information of this user seldom or not exists; (3) there is the known information of some priori (such as field keyword set, keyword distribution and relation etc.) in the knowledge in specific area, existing method does not utilize these information to provide and better browses householder method.In sum, the knowledge that current existing method can not be applicable to specific area is browsed.

Summary of the invention

The present invention proposes a kind of domain knowledge browsing method of knowledge based map, the method can be navigated to the user being unfamiliar with domain knowledge.

A domain knowledge browsing method for knowledge based map, is divided into two stages, that is: Knowledge Map builds and important knowledge identification; Wherein have three data objects, that is: domain knowledge base, domain knowledge map and important knowledge, important knowledge is used for the starting point browsing knowledge base to knowledge user, it is knowledge user navigation that domain knowledge map is used in knowledge navigation process, and domain knowledge base provides the idiographic flow of knowledge;

Knowledge Map builds and specifically comprises the following steps:

Step one, domain knowledge represent, the knowledge in domain knowledge base are expressed in computable mode;

Have employed the method for vector space model in step one to represent domain knowledge, each knowledge in knowledge base is represented as form as shown in Equation (1):

k＝{(t ₁，w ₁)，(t ₂，w ₂)，......，(t _M，w _M)}(1)

Wherein t _iwhat represent is the vocabulary be present in domain knowledge, w _iwhat represent is the significance level of this vocabulary to knowledge, and each knowledge in knowledge base can be expressed as the form of vector, the final co-occurrence matrix forming domain knowledge, and adopts D to represent co-occurrence matrix.

Step 2, Semantic Similarity Measurement, using co-occurrence matrix D as inputting and exporting similarity matrix S;

Adopt latent semantic analysis technology LSA to calculate the similarity between domain knowledge in step 2, by the singular value decomposition method in linear algebra, sparse co-occurrence matrix is mapped to a lower dimension spatially, mutually orthogonal between each dimension in this compression stroke, remove the noise data existed in original matrix; Finally by the included angle cosine value in low-dimensional orthogonal intersection space between compute vector, and using this value as the similarity between knowledge, calculate the similarity between all knowledge between two, final formation similar matrix S is as the output of this step.

Step 3, Knowledge Map build, and using similar matrix S as inputting and exporting Knowledge Map, contain the similarity between all knowledge in similar matrix S, from these similarities, which identifies be retained in final Knowledge Map, which is deleted.

Similarity analysis in step 3 adopts following method:

First, be divided into four classifications by ascending for all similarities and adopt S, M, L2 and L1 to represent, S represents minimum a part of similarity, refers to the similarity between the domain knowledge that there is not association, namely needs the similarity of deleting; M represents medium sized one group of similarity, refers to the association be prevalent between a certain domain knowledge knowledge, and the existence of this general similarity can cover the linked character of domain knowledge, is therefore defined as noise, needs deleted; L ₁represent one group of maximum similarity, associate domain knowledge closely for descriptive semantics, the semantic relation of descriptor field external knowledge, needs to be retained; L ₂represent one group of larger similarity, the incidence relation between sub-field is present in this group similarity, and therefore this group similarity needs to be retained.

Secondly, constantly similarity is joined Knowledge Map to little order according to similarity, until Knowledge Map associates completely.

After domain knowledge map has built, enter important knowledge cognitive phase, using the Knowledge Map built as inputting and exporting the important knowledge calculated, specifically comprise the following steps:

Step one, define important knowledge: be comprise one group of more knowledge of realm information by important knowledge definition; The degree on namely corresponding summit is comparatively large, therefore using the knowledge corresponding to larger for Knowledge Map moderate one group of summit as important knowledge;

Step 2, Knowledge Map structure analysis, determine whether there is important knowledge in constructed Knowledge Map, if existed, the important knowledge of carrying out next step calculates, if there is no then prove that the Knowledge Map built in the first stage is unreasonable, returning the first stage reanalyses domain knowledge, amendment Knowledge Map construction strategy;

Adopt the inner structure of the algorithm determination Knowledge Map in graph theory in step 2, comprise the criterion of small-world network and the criterion of Scale-free Network;

The feature of small-world network is less average path length and larger convergence factor, and average path length refers to the mean value of the path from arbitrary node to other nodes, utilizes formula (2) to calculate:

$L=\frac{2}{N(N-1)}\underset{0<i,j<N,i\&NotEqual;j}{\mathrm{\&Sigma;}}{l}_{\mathrm{ij}}---\left(2\right)$

Wherein N refers to the quantity of node, l _i,jrefer to the path of node i to node j; Convergence factor reflects the aggregation extent in Knowledge Map, is calculated by formula (3):

$C=\frac{1}{N}\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}\frac{2\&CenterDot;e_i}{d_i(d_i-1)}---\left(3\right)$

Wherein d _iwhat represent is the degree of node i, e _irepresent be node i adjacent node between association quantity;

The feature of Scale-free Network is that the degree of all nodes meets power distribution, concrete mode obtains after power distributes to judge whether to meet power distribution by the mode of power function fitting, discrimination standard R is obtained in fit procedure, the distribution of this value data of description meets the degree of power distribution, more higher close to 1 degree of agreement, otherwise on the contrary;

If Knowledge Map meets the feature of small-world network and Scale-free Network simultaneously, then think that Knowledge Map gathers different classifications, link comparatively tight in classification, there is link between classification, in declarative knowledge map, there is important knowledge.

Step 3, important knowledge calculate, and namely identify the quantity of important knowledge;

Calculated the quantity of important knowledge by computational grid efficiency, namely arrived the consumption of another one node by the arbitrary node in network, in Knowledge Map, then represent the easy degree navigating to an other domain knowledge from any domain knowledge; The formula (4) of network efficiency is as follows:

$E=\frac{1}{N(N-1)}\underset{i\&NotEqual;j}{\overset{N}{\mathrm{\&Sigma;}}}\frac{1}{d_{\mathrm{ij}}}---\left(4\right)$

Wherein, d _i,jwhat describe is bee-line between node i and node j; The node that constantly degree of deletion is higher from Knowledge Map, the simultaneously efficiency of computational grid, stop when network efficiency is reduced to 10% of starting efficiency, and determine the quantity of important knowledge.

Beneficial effect of the present invention:

Method in the present invention does not need other information except domain knowledge base, can generate Knowledge Map and important knowledge from domain knowledge base; Adopt latent semantic analysis technique construction domain knowledge map can go up the noise data existed in removal field by a relatively large margin, the semantic connection features of outstanding domain knowledge map.

Accompanying drawing explanation

Knowledge Map schematic diagram in Fig. 1 embodiment of the present invention;

Fig. 2 the present invention is based on the domain knowledge browsing method process flow diagram of Knowledge Map;

Fig. 3 the present invention is based on the fundamental analysis figure of all similarities of domain knowledge browsing method of Knowledge Map;

Fig. 4 Knowledge Map structure analysis of the present invention figure;

Embodiment

Below in conjunction with accompanying drawing, the present invention is further introduced.

The present invention proposes a kind of domain knowledge browsing method of knowledge based map, the method can be navigated to the user being unfamiliar with domain knowledge.Unique input required for method in the present invention is that (field here mainly refers to text to the domain knowledge compiled, can being converted into by additional textual information or labeling method of other types is described herein), export as the domain knowledge map for navigating and one group of important knowledge calculated.Here Knowledge Map refers to the association between one group of domain knowledge and these knowledge, adopts the non-directed graph in mathematics to express in the present invention, gives the example of a Knowledge Map in accompanying drawing 1.Include 10 domain knowledges in this example, between the knowledge of certain fields, there is semantic relation, as shown in the line in figure.Important knowledge refers to and comprises the more knowledge of this realm information, and this kind of knowledge can make user understand the related content in field faster.

Figure 2 shows the idiographic flow of method.The method is totally divided into two stages, that is: Knowledge Map builds and important knowledge identification.Three data objects are had, that is: domain knowledge base, domain knowledge map and important knowledge in method.Important knowledge is used for the starting point browsing knowledge base to knowledge user, and domain knowledge map is used for for knowledge user navigates in knowledge navigation process, field this be the idiographic flow that storehouse then can provide knowledge.Two stages main in method respectively comprise again three steps, set forth implementation method below by step.

Knowledge Map builds stage reception domain knowledge base as inputting and exporting the Knowledge Map built, and comprising three sub-steps, that is: domain knowledge represents, Semantic Similarity Measurement and Knowledge Map build.

(1) domain knowledge represents

The object that domain knowledge represents is expressed in computable mode the knowledge in domain knowledge base, present invention employs the method for vector space model (VSM) to represent domain knowledge.Each knowledge in knowledge base is represented as form as shown in Equation 1.

K={ (t ₁, w ₁), (t ₂, w ₂) ... .., (t _m, w _m) formula 1

T in formula 1 _iwhat represent is the vocabulary be present in domain knowledge, w _iwhat represent is the significance level of this vocabulary to knowledge, plays computing method and adopts a kind of widely used TF-IDF method (TF-IDF is weighing computation method of extensive employing, for reducing length, no longer describes in detail).Each knowledge in knowledge base can be expressed as the form of vector, and the final co-occurrence matrix forming domain knowledge, adopts D to represent co-occurrence matrix in the present invention.

(2) Semantic Similarity Measurement

Semantic Similarity Measurement using co-occurrence matrix D as inputting and exporting similarity matrix S.Latent semantic analysis technology (LSA) is adopted to calculate similarity between domain knowledge in the present invention, reason is that co-occurrence matrix D is generally large-scale sparse matrix, the direct stage can compare consumption of natural resource, and the keyword adopted in co-occurrence matrix computation process is in addition not separate.Sparse co-occurrence matrix is mapped to a lower dimension spatially by the singular value decomposition method in linear algebra by LSA technology.In this compression stroke separate between each dimension (orthogonal), eliminate the noise data existed in original matrix.Finally by the included angle cosine value in low-dimensional orthogonal intersection space between compute vector, and using this value as the similarity between knowledge.Calculate the similarity between all knowledge between two, final formation similar matrix S is as the output of this step.

(3) Knowledge Map builds

Knowledge Map builds similar matrix S as inputting and exporting Knowledge Map.Knowledge Map as previously mentioned in the present invention refers to the association between one group of domain knowledge and these knowledge.Contain the similarity between all knowledge in similar matrix S, the groundwork of this step is exactly from these similarities, which identifies be retained in final Knowledge Map, and which is deleted.Fig. 3 gives the fundamental analysis to all similarities for this reason.

As shown in Figure 3, the present invention is divided into four classifications by ascending for all similarities and adopt S, M, L2 and L1 to represent.In figure+representing that corresponding similarity needs to retain, the similarity of-sign correspondence needs to delete.S represents minimum a part of similarity, refers to similarity between the domain knowledge that there is not association (K in such as accompanying drawing 2 ₆and K ₂between there is not association), therefore need to delete this group similarity.M represents medium sized one group of similarity, refers to the association be prevalent between a certain domain knowledge knowledge.The inevitable common sparing vocabulary of knowledge in same field, the therefore general similarity of knowledge in a field.The existence of this general similarity can cover the linked character of domain knowledge, is therefore defined as noise in the present invention, and this group noise needs deleted equally.L ₁represent one group of maximum similarity, associate domain knowledge closely for descriptive semantics.Knowledge in a field generally can be divided into different sub-fields, and the inner semantic relation in sub-field is tightr.The semantic relation of this group similarity main descriptor field external knowledge, therefore this group similarity needs to be retained.L ₂represent one group of larger similarity, the incidence relation between sub-field is present in this group similarity, and therefore this group similarity needs to be retained equally.

Based on above analysis, the present invention adopts a kind of simple method to decide which similarity to be retained in final Knowledge Map, that is: constantly similarity is joined Knowledge Map to little order according to similarity, until Knowledge Map associates completely.Here association completely refers to the association of figure in graph theory, and method of discrimination has ripe method, repeats no more.Special instruction be the size having eliminated similarity in final Knowledge Map, adopt Boolean to carry out representing (0 represents onrelevant, and 1 represents relevant).

According to the flow process shown in accompanying drawing 2, after domain knowledge map has built, enter important knowledge cognitive phase.This stage using the Knowledge Map built as inputting and exporting the important knowledge calculated.This stage is divided into three steps, that is: define important knowledge, Knowledge Map structure analysis and important knowledge and calculate.

(4) important knowledge is defined

Important knowledge can be defined from different perspectives according to different demands.The present invention comprises one group of more knowledge of realm information.The correlation comparison comprising one group of more knowledge of realm information and other knowledge is many, larger from the degree the angle of mathematics graph theory being exactly corresponding summit.Therefore the present invention using the knowledge corresponding to larger for Knowledge Map moderate one group of summit as important knowledge.

(5) Knowledge Map structure analysis

The object of Knowledge Map structure analysis to determine whether there is important knowledge in constructed Knowledge Map.Do not comprise important knowledge in Knowledge Map in some cases, when such as, in Knowledge Map all degree of vertexs are similar, think that all domain knowledges have similar significance level, therefore there is not important knowledge.What accompanying drawing 4 was shown is the Knowledge Map example that there is important knowledge.Knowledge accumulating in this Knowledge Map becomes three sub-fields, known K ₁, K ₁₃and K ₈degree comparatively large, the domain knowledge therefore corresponding to it can be confirmed as important knowledge.Think in the present invention if the structure shown in the inner structure of domain knowledge map to accompanying drawing 4 is similar, then there is important knowledge in Knowledge Map.

The present invention is by adopting the inner structure of the algorithm determination Knowledge Map in graph theory.Small-world network is the type of a kind of mathematics figure, and in this figure, most node is not with adjacent to each other, but most of node just can be able to arrive from other nodes through several step.If the point in a small-world network is represented a people, and tie line represents person to person's understanding, then this small-world network can reflect the Small-World Phenomena that stranger is linked by the people of common understanding each other.The feature of small-world network is less average path length and larger convergence factor.Average path length refers to the mean value of the path from arbitrary node to other nodes, can calculate with formula 2.

$L=\frac{2}{N(N-1)}\underset{0<i,j<N,i\&NotEqual;j}{\mathrm{\&Sigma;}}{l}_{\mathrm{ij}}$ Formula 2

Wherein N refers to the quantity of node, l _i,jrefer to the path of node i to node j.Convergence factor reflects the aggregation extent in Knowledge Map, can be calculated by formula 3.In social network analysis, this coefficient is for describing the degree of knowing each other between someone friend.If Knowledge Map meets the feature of small-world network, the knowledge accumulating in declarative knowledge map is in different sub-fields, and sub-field is inner crucial more.

$C=\frac{1}{N}\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}\frac{2\&CenterDot;e_i}{d_i(d_i-1)}$ Formula 3

Wherein d _iwhat represent is the degree of node i, e _irepresent be node i adjacent node between association quantity.

Scale-free Network is the complex network with a class feature, its characteristic feature is that most of node in a network is only connected with little node (degree of node is very little), and having few node to be connected (degree of node is very high) with very many nodes, the distribution also with regard to node degrees all in network meets power-law distribution.If Knowledge Map meets the feature of Scale-free Network, then think and have little a part of node to have higher degree in network, be regarded as the gathering center of Knowledge Map.

If Knowledge Map meets the feature of small-world network and Scale-free Network in the present invention simultaneously, then think that the Knowledge Map shown in Knowledge Map to accompanying drawing 3 has similar inner structure, also there is important knowledge in declarative knowledge map.

(6) important knowledge calculates

The object that important knowledge calculates is the quantity identifying important knowledge.Domain knowledge corresponding to the node that knowledge degree of being defined as important in important knowledge the present invention is larger, therefore identifies that total total how many important knowledge has just become subject matter.For determining the quantity of important knowledge, the concept introducing network efficiency can be calculated by formula 4.What network efficiency described is the consumption being arrived another one node by the arbitrary node in network, then represents the easy degree navigating to an other domain knowledge from any domain knowledge in Knowledge Map.

$E=\frac{1}{N(N-1)}\underset{i\&NotEqual;j}{\overset{N}{\mathrm{\&Sigma;}}}\frac{1}{d_{\mathrm{ij}}}$ Formula 4

Determine that the thinking of important knowledge is the node that constantly degree of deletion is higher from Knowledge Map, simultaneously the efficiency of computational grid, stop when network efficiency is reduced to 10% of starting efficiency, and determine the quantity of important knowledge.

Claims (5)

1. a domain knowledge browsing method for knowledge based map, is divided into two stages, that is: Knowledge Map builds and important knowledge identification; Wherein have three data objects, that is: domain knowledge base, domain knowledge map and important knowledge, important knowledge is used for the starting point browsing knowledge base to knowledge user, it is knowledge user navigation that domain knowledge map is used in knowledge navigation process, and domain knowledge base provides the idiographic flow of knowledge; It is characterized in that:

Knowledge Map builds and specifically comprises the following steps:

Step one, domain knowledge represent, the knowledge in domain knowledge base are expressed in computable mode;

Step 2, Semantic Similarity Measurement, using co-occurrence matrix D as inputting and exporting similarity matrix S;

Step 3, Knowledge Map build, and using similar matrix S as inputting and exporting Knowledge Map, contain the similarity between all knowledge in similar matrix S, from these similarities, which identifies be retained in final Knowledge Map, which is deleted;

After domain knowledge map has built, enter important knowledge cognitive phase, using the Knowledge Map built as inputting and exporting the important knowledge calculated, specifically comprise the following steps:

Step one, define important knowledge: be comprise one group of more knowledge of realm information by important knowledge definition; The degree on namely corresponding summit is comparatively large, therefore using the knowledge corresponding to larger for Knowledge Map moderate one group of summit as important knowledge;

Step 2, Knowledge Map structure analysis, determine whether there is important knowledge in constructed Knowledge Map, if existed, the important knowledge of carrying out next step calculates, if there is no then prove that the Knowledge Map built in the first stage is unreasonable, returning the first stage reanalyses domain knowledge, amendment Knowledge Map construction strategy;

Step 3, important knowledge calculate, and namely identify the quantity of important knowledge;

Calculated the quantity of important knowledge by computational grid efficiency, namely arrived the consumption of another one node by the arbitrary node in network, in Knowledge Map, then represent the easy degree navigating to an other domain knowledge from any domain knowledge; The formula (4) of network efficiency is as follows:

$E=\frac{1}{N(N-1)}\underset{i\&NotEqual;j}{\overset{N}{\&Sigma;}}\frac{1}{d_{ij}}---\left(4\right)$

Wherein, N refers to the quantity of node, d _ijwhat describe is bee-line between node i and node j; The node that constantly degree of deletion is higher from Knowledge Map, the simultaneously efficiency of computational grid, stop when network efficiency is reduced to 10% of starting efficiency, and determine the quantity of important knowledge.

2. the domain knowledge browsing method of a kind of knowledge based map as claimed in claim 1, it is characterized in that: have employed the method for vector space model to represent domain knowledge in the step one that Knowledge Map builds, each knowledge in knowledge base is represented as the form as shown in formula (1):

k＝{(t ₁,w ₁),(t ₂,w ₂),……，(t _M,w _M)}(1)

Wherein t _iwhat represent is the vocabulary be present in domain knowledge, w _iwhat represent is the significance level of this vocabulary to knowledge, and each knowledge in knowledge base can be expressed as the form of vector, the final co-occurrence matrix forming domain knowledge, and adopts D to represent co-occurrence matrix.

3. the domain knowledge browsing method of a kind of knowledge based map as claimed in claim 1 or 2, it is characterized in that: in the step 2 that Knowledge Map builds, adopt latent semantic analysis technology LSA to calculate the similarity between domain knowledge, by the singular value decomposition method in linear algebra, sparse co-occurrence matrix is mapped to a lower dimension spatially, mutually orthogonal between each dimension in this space, remove the noise data existed in original matrix; Finally by the included angle cosine value in low-dimensional orthogonal intersection space between compute vector, and using this value as the similarity between knowledge, calculate the similarity between all knowledge between two, final formation similar matrix S is as the output of this step.

4. the domain knowledge browsing method of a kind of knowledge based map as claimed in claim 1 or 2, is characterized in that: the similarity analysis in the step 3 that Knowledge Map builds adopts following method:

First, be divided into four classifications by ascending for all similarities and adopt S, M, L ₂and L ₁represent, S represents minimum a part of similarity, refers to the similarity between the domain knowledge that there is not association, namely needs the similarity of deleting; M represents medium sized one group of similarity, refers to the association be prevalent between a certain domain knowledge knowledge, and the existence of this general similarity can cover the linked character of domain knowledge, is therefore defined as noise, needs deleted; L ₁represent one group of maximum similarity, associate domain knowledge closely for descriptive semantics, the semantic relation of descriptor field external knowledge, needs to be retained; L ₂represent one group of larger similarity, the incidence relation between sub-field is present in this group similarity, and therefore this group similarity needs to be retained; Secondly, constantly similarity is joined Knowledge Map to little order according to similarity, until Knowledge Map associates completely.

5. the domain knowledge browsing method of a kind of knowledge based map as claimed in claim 1, it is characterized in that: the inner structure adopting the algorithm determination Knowledge Map in graph theory in the step 2 of important knowledge cognitive phase, comprises the criterion of small-world network and the criterion of Scale-free Network;

The feature of small-world network is less average path length and larger convergence factor, and average path length refers to the mean value of the path from arbitrary node to other nodes, utilizes formula (2) to calculate:

$L=\frac{2}{N(N-1)}\underset{0<i,j<N,i\&NotEqual;j}{\&Sigma;}{l}_{ij}---\left(2\right)$

Wherein N refers to the quantity of node, l _ijrefer to the path of node i to node j; Convergence factor reflects the aggregation extent in Knowledge Map, is calculated by formula (3):

$C=\frac{1}{N}\underset{i=1}{\overset{N}{\&Sigma;}}\frac{2\&CenterDot;e_i}{d_i\left(d_i-1\right)}---\left(3\right)$

Wherein d _iwhat represent is the degree of node i, e _irepresent be node i adjacent node between association quantity;

The feature of Scale-free Network is that the degree of all nodes meets power distribution, concrete mode obtains after power distributes to judge whether to meet power distribution by the mode of power function fitting, discrimination standard R is obtained in fit procedure, the distribution of this value data of description meets the degree of power distribution, more higher close to 1 degree of agreement, otherwise on the contrary;

If Knowledge Map meets the feature of small-world network and Scale-free Network simultaneously, then think that Knowledge Map gathers different classifications, link comparatively tight in classification, there is link between classification, in declarative knowledge map, there is important knowledge.