CN106372099A - Agricultural field ontology validity assessment method - Google Patents

Abstract

The invention discloses an agricultural field ontology validity assessment method. Agricultural field ontology concepts are assessed in combination with a recall ratio R, a precision ratio P and an F1 index; the consistency, accuracy and conciseness of a classification relationship among the agricultural field ontology concepts are assessed in combination with semantic similarity among the concepts; and an object attribute relationship among the agricultural field ontology concepts is assessed according to an assessment index. Therefore, three important problems, including validity assessment of the ontology concepts, validity assessment of the ontology classification relationship and validity assessment of the ontology non-classification relationship, existent in agricultural field ontology validity assessment are solved emphatically. An agricultural field ontology validity assessment method system proposed by the invention takes an agricultural field ontology as a research object, and can effectively assess the quality of the agricultural field ontology and improve the agricultural ontology construction efficiency.

Description

A kind of agriculture field body efficiency assessment method

Technical field

The present invention relates to computer information technology is in the application of agriculture field, more particularly, to a kind of agriculture field body is effective Property appraisal procedure.

Background technology

Body as a kind of shared ideas model that can advise knowledge in semantic level upper table, day by day become information management, The important component part of the multiple fields such as information retrieval and semantic web.Particularly since semantic web proposition, body is just gradually Become an important research direction in artificial intelligence and knowledge engineering field, in the acquisition of knowledge, expression, analysis and application etc. Aspect has far-reaching significance.Body, according to the level of detail and field dependency degree two indices, can be divided into 4 classifications: top Body, domain body, task ontology and applied ontology.Wherein, domain body is professional body, describes a certain specific Relation between the vocabulary of concept and concept in section field, or the most important theories of this ambit.Due to domain body Knowledge has significant domain feature, so domain body can be more reasonable and effectively carry out expressing for knowledge.

The structure of domain body makes the shared of domain knowledge and reuse be possibly realized.In numerous domain knowledges, agriculture Industry domain knowledge is a kind of high-quality production factors, and agricultural workforce and the efficiency of capital production can be made to increase substantially； Agriculture knowledge be also Agricultural Information application basis, as agricultural information system, agricultural decision making system, agriculture specialist system, Agriculture intelligence control system etc. all be unable to do without agricultural knowledge.Agriculture field body is one and comprises agriculture term, definition and art Between language, the system of normative connection explanation, is the formalization table of the mutual relation between concept, concept and concept in agriculture ambit Reach.It can be expressed as four-tuple, including body essential information, such as body title, founder, design time, modification time, mesh And the body such as Knowledge Source metadata information；The set of agricultural knowledge concept；Classification relation between agricultural knowledge concept and Non-categorical set of relationship；Axiomatic set theory present in body.

This physical ability of agriculture field describes on semantic level to the related data of agricultural, information and knowledge it is provided that disease The ontology services such as pest control, Making A Strategic Decision of The Agricultural Productions support, intelligent diagnostics and control.Carry within 2000 since FAO (Food and Agriculture Organization of the United Nation) Since going out Agricultural ontology service research plan, how effectively the agriculture field body quantity of different scales is being in just explosive growth, Assess the structure quality of these agriculture field bodies, if really meet domain knowledge situation, could really embody knowledge organization Function, effect of Agriculture knowledge service can be realized, automatically or semi-automatically whether Method for Ontology Learning scientific and reasonable, how to help Help user to select more suitably Agricultural ontology, become problem demanding prompt solution in current agricultural ontology research.Accordingly, it would be desirable to agriculture Industry domain body carries out efficiency assessment.

Body is assessed as an important means assessing this weight, by the method with science, it then follows unified Index system, assessment performance under specific application area or specific environment for the body and the suitability, are whether impact body can be An important factor in order of large-scale application in semantic net.Set up a set of scientific and effective body Assessment theory and assessment side Method, contributes to weighing whether body constructing method has effectiveness, is conducive to existing body is effectively managed, updates and ties up Shield, helps instruct ontological construction process and lifting body construction efficiency, thus forming more scientific rational domain body and more preferably Ground provides associated body service.

Associated specialist scholar has done substantial amounts of research in terms of body assessment both at home and abroad, and proposes a lot of body assessment sides Method.On the whole, body appraisal procedure mainly has seven kinds, and they are respectively: appraisal procedure based on user, the commenting of task based access control Estimate method, the appraisal procedure based on principle, the appraisal procedure Ji Yu " gold criterion ", the appraisal procedure based on application, be based on language The material appraisal procedure in storehouse and the appraisal procedure based on compound standard.With the constantly improve of body appraisal procedure, body automatic Change assessment also to be developed rapidly, and have developed some body assessment tools, wherein odeval, ontoqa, core, Ontomanager, aeon etc. are wherein to apply more extensive body assessment tool.

Although existing body appraisal procedure and assessment tool achieve some achievements, in terms of some, body can be entered Row assessment, but for Chinese body and the efficiency assessment method being applied to complicated ambit body such as agriculture field body Or few, and existing body appraisal procedure compares one-sided, does not form a whole set of scientific comprehensive body assessment Theory and method system.

Content of the invention

The technical problem being existed based on background technology, the present invention proposes a kind of agriculture field body efficiency assessment side Method.

A kind of agriculture field body efficiency assessment method proposed by the present invention, comprises the following steps:

S1, assessment agriculture field Ontological concept；

Classification relation between s2, assessment agriculture field Ontological concept；

Non-categorical relation between s3, assessment agriculture field Ontological concept；

Step s1 specifically include following step by step:

S11, agriculture field body to be assessed is parsed according to concept, obtained concept set；

S12, setting Concept Evaluation index, Concept Evaluation index includes recall ratio r, precision ratio p and f1 index, wherein,

It is defined on recall ratio and in the case that precision ratio occurs contradiction, this recall ratio and precision ratio are weighted be in harmonious proportion The average value obtaining is f index,

$f=\frac{({\mathrm{\α}}^2+1)p\×r}{{\mathrm{\α}}^2(p+r)}$

F1 index is f index during α=1,

$f1=\frac{2\×p\×r}{p+r}$

S13, the ambit scope according to body to be assessed, select or build corresponding domain lexicon；

S14, it is that recall ratio, precision ratio and f1 exponent pair concept are estimated according to Concept Evaluation index, obtain correctly general Read number；

S15, the assessment result of output recall ratio, precision ratio and f1 index；

Step s2 specifically include following step by step:

S21, by agriculture field body to be assessed according to concept between subclassof relation parsed, and protected Save as data structure be figure g=<v, e>data file, v be Ontological concept set, e be classification relation set；

S22, three evaluation indexes of setting classification relation assessment, are concordance, accuracy and terseness respectively；

Semantic similarity between s23, calculating concept, and optimal threshold ε is set；

S24, set up Concept Semantic Similarity set s (s₁, s₂..., s_n) and concordance, accuracy and terseness judgement knot Fruit set t (t₁, t₂..., t_n), s_i=sim (v_i, v_j), i, j are natural numbers, and concept vi is concept vj hypernym, and v_i≠ v_j；

S25, judge that whether concept vi hypernym number is more than 1, be, then judge the classification error of concept vi and be stored in judgement and tie Fruit set t (t₁, t₂..., t_n)；

S26, by sim (v_i, v_j) compare with optimal threshold ε, and judge whether concept vi and concept vj accord with according to comparative result Close classification relation, meet, then by s_iIt is stored in Concept Semantic Similarity set s (s₁, s₂..., s_n)；Do not meet and then judge concept vi Classification error simultaneously will determine that result is stored in judged result set t (t₁, t₂..., t_n)；

S27, judge whether sim (v_i, v_j)=sim (v_i, v_k), if it is present judging concept v_jWith concept v_kSend out Raw redundancy error is simultaneously stored in judged result set t (t₁, t₂..., t_n)；

S28, repeat step s25 to s27, until all calculating and having judged Ontological concept set v and classification relation set e Finish；

S29, output Similarity Measure results set s and concordance, accuracy and terseness judged result set t；

Step s3 specifically include following step by step:

S31, parsed according to object properties relation pair agriculture field body, the attributed relational graph that obtains of parsing is preserved For data structure it isData file, and represented with adjacency matrix a；V is Ontological concept set, e is attribute Set of relationship,For relation on attributes concept pair；

The evaluation index of object properties relation between s32, setting agriculture field Ontological concept, and sentenced according to evaluation index setting Surely set q；

S33, according to evaluation index, between agriculture field Ontological concept, object properties relation is estimated, and by assessment result Record is in judging set q；

Relation on attributes set e in s34, output agriculture field body, relation on attributes concept pairAnd result of determination set q.

Preferably, in step s13, the structure of domain lexicon comprises the following steps:

S131, determine subject fields scope, and assembling sphere vocabulary；

S132, determine the classification characteristicses of vocabulary, mutual relation between organized words and vocabulary, form vocabulary structure, and specify Attribute dictionary；

S133, the vocabulary to dictionary and relation carry out examination ＆ verification and revise.

Preferably, in step s131, from document, subject classical works, Baidupedia, existing dictionary and expertise way Footpath obtains Field Words.

Preferably, in step s23, concept c₁And c₂Between semantic similarity sim (c₁, c₂) computing formula is as follows:

sim(c₁,c₂)=[sim_dist (c₁,c₂)]^α×[sim_depth(c₁,c₂)]^β×[sim_width(c₁,c₂) ]^γ,

Wherein, sim_dist (c₁, c₂) for semantic similarity and semantic distance corresponding relation, sim_depth (c₁, c₂ For the corresponding relation of semantic similarity and semantic depth, sim_width (c₁, c₂) it is semantic similarity and semantic node density Corresponding relation；α represents the weight of semantic distance, the weight of β representational level depth, and γ represents the weight of node density, α, β, γ For regulation parameter, and α, β, γ sum is 1.

Preferably,Wherein, dist (c₁, c₂) represent semantic distance, a For regulation parameter；dist(c₁,c₂)=len (c₁,c₂), len (c₁,c₂) represent concept c₁And c₂Between path.

Preferably,Depth (c) represents concept c Depth, depth (c)=depth (parent (c))+1, parent (c) represents the father node of concept c it is preferable that root node Depth depth (root)=1.

Preferably, the density of concept c is expressed as width (c), and it is directly proportional to classification degree of refinement, as width (c₁)≤ width(c₂),

$sim\_width(c_1,c_2)=\frac{width\left(c_1\right)}{width\left(c_2\right)}.$

Preferably, in step s26, if ε is ＜ sim (v_i, v_j) ＜ 1, then concept v_iWith concept v_jMeet classification relation requirement； If 0≤sim is (v_i, v_j) ＜ ε, then judge concept v_iWith concept v_jThere is classification error and be stored in judged result set t.

Preferably, in judged result set t, t_nCorresponding with concept vn, if concept vn meets evaluation index, correspond to t_n It is labeled；If concept vn does not meet evaluation index, the type of error that corresponding tn mark is judged.

Preferably, between agriculture field Ontological concept, the evaluation index of object properties relation includes:

Reversibility: there is relation r between concept a and concept b, meanwhile, there is relation r between concept b and concept a^-1, R is then claimed to have reversibility；

Relation inheritance: there is relation r between concept a and father's concept b, there is also between concept a and sub- concept c meanwhile Relation r, then claim relation r to have relation inheritance；

Inverse relationship inheritance: there is relation r between father's concept and concept a, meanwhile, between its sub- concept and concept a There is also relation r, then claim relation r to have inverse relationship inheritance；

Transitivity: if there is relation r between concept a and concept b, there is relation r between concept b and concept c, meanwhile, generally Read and between a and concept c, there is relation r, then claim relation r to have transitivity；

Symmetry: if concept a and concept b have relation r, be denoted as a=r (b), then concept b and concept a have relation r, are denoted as b =r (a)；

Reflexivity: concept a has relation r, if r has reflexivity, a=r (a).

A kind of agriculture field body efficiency assessment method that the present invention provides, refers in conjunction with recall ratio r, precision ratio p and f1 Several agriculture field Ontological concept is estimated, in conjunction with concept between semantic similarity classify between agriculture field Ontological concept pass The concordance of system, accuracy, terseness are estimated, according to evaluation index between agriculture field Ontological concept object properties relation It is estimated.So, to solve three major issue bodies present in agriculture field body efficiency assessment general for emphasis The efficiency assessment of the efficiency assessment, the efficiency assessment of Ontology relation and body non-categorical relation read.

Agriculture field body efficiency assessment method system proposed by the present invention, can with agriculture field body as object of study Effectively assess the quality of agriculture field body, lifting Agricultural ontology builds efficiency.

Brief description

Fig. 1 is a kind of agriculture field body efficiency assessment method flow diagram proposed by the present invention；

Fig. 2 is assessment agriculture field Ontological concept method flow diagram；

Fig. 3 is classification relation method flow diagram between assessment agriculture field Ontological concept；

Fig. 4 is non-categorical relational approach flow chart between assessment agriculture field Ontological concept；

Fig. 5 assesses prototype system overall framework figure for tea-plant pests body.

Specific embodiment

With reference to Fig. 1, a kind of agriculture field body efficiency assessment method proposed by the present invention, comprise the following steps.

S1, assessment agriculture field Ontological concept；

Classification relation between s2, assessment agriculture field Ontological concept；

Non-categorical relation between s3, assessment agriculture field Ontological concept.

Concept, as the basic element of ontology model, is the core of expression domain knowledge.Therefore to domain body concept Assessment is the pith completing body assessment.

With reference to Fig. 2, in present embodiment, assessment agriculture field Ontological concept be step s1 specifically include following step by step.

S11, agriculture field body to be assessed is parsed according to concept, obtained concept set.This step is to defeated The agriculture field body to be assessed entering carries out body pretreatment, specifically can be led agricultural to be assessed by body analytical tool Domain body is parsed according to concept.

S12, setting Concept Evaluation index, Concept Evaluation index includes recall ratio r, precision ratio p and f1 index.

Recall ratio r refers to the correct concept number in body and the ratio of the sum of the concept in domain lexicon,

Precision ratio p refers to the correct concept number in body and the ratio of the sum of the concept in body,

It is defined on recall ratio and in the case that precision ratio occurs contradiction, this recall ratio and precision ratio are weighted be in harmonious proportion The average value obtaining is f index, and f1 index is the specific value of f index.F index is specifically calculated as follows with f1 index:

$f=\frac{({\mathrm{\&alpha;}}^2+1)p\&times;r}{{\mathrm{\&alpha;}}^2(p+r)};$

F1 index is f index during α=1,

$f1=\frac{2\&times;p\&times;r}{p+r}.$

The evaluation index of one group of scientific and reasonable concept should be able to evaluate the level of coverage of concept in constructed body, general Read the accuracy of term, i.e. concept set in body to be assessed should farthest comprise the important terms in this field. And the concept defined in body not only needs to have integrity in addition it is also necessary to ensure the correctness of Ontological concept.

Recall ratio, precision ratio and f index as reflection Concept Evaluation effect important indicator, f index can comprehensively this two Individual evaluation index, for the index that concentrated expression is overall, f1 index is clear and definite to f index.In present embodiment, complete to look into Rate, precision ratio and f1 index, as Concept Evaluation index, are conducive to the levels of precision of Concept Evaluation.

S13, the ambit scope according to body to be assessed, select or build corresponding domain lexicon.

Domain lexicon refers to the term of specific area or the set of vocabulary, can reflect information in certain specific field, The elements such as concept, element therein can extract from related corpus and refine.With conventional dictionaries simply simply by word Converge with organizing according to morphology or lexicographic order and compare, modern dictionary not only concept to be comprised, more to consider general in the middle of dictionary Read the relation and concept between.The accuracy of Concept Evaluation result depends on the selection of domain lexicon, therefore, the choosing of domain lexicon Fixed extremely important.

In step s13, the structure of domain lexicon comprises the following steps:

S131, determine subject fields scope, and assembling sphere vocabulary；Specifically, Field Words can be classical from document, subject Works, Baidupedia, existing dictionary and expertise approach obtain.

S132, determine the classification characteristicses of vocabulary, mutual relation between organized words and vocabulary, form vocabulary structure, and specify Attribute dictionary；

S133, the vocabulary to dictionary and relation carry out examination ＆ verification and revise.

S14, it is that recall ratio, precision ratio and f1 exponent pair concept are estimated according to Concept Evaluation index, obtain correctly general Read number.

S15, the assessment result of output recall ratio, precision ratio and f1 index, obtain Concept Evaluation result.

Relation is that in body, the one kind between concept and concept contacts, and the classification relation between agriculture field Ontological concept is concrete Show as subclassof relation, describe the relation between father's concept and sub- concept.Classification relation system is general as tissue The basis read, is the backbone structure of body.Therefore, how research is effectively assessed to classification relation is to be highly desirable to 's.

With reference to Fig. 3, in present embodiment, between assessment agriculture field Ontological concept classification relation be step s2 specifically include with Under step by step.

S21, by body analytical tool by agriculture field body to be assessed according to concept between subclassof relation Parsed, and be saved as data structure be figure g=<v, e>data file, v be Ontological concept set, e be classification close Assembly is closed.

S22, three evaluation indexes of setting classification relation assessment, are concordance, accuracy and terseness respectively.

In present embodiment, three evaluation indexes have been directed to common three kinds of typical faults in classification relation and classify not Unanimously, classification error and redundancy error.Concordance is mainly used to occur in that conflicting classification relation in detection body；Accurately Property mainly in Ontology system occur concept classification Problem-Error；Terseness is primarily directed in Ontology system There is the redundancy error of same concept in middle discovery.

Semantic similarity between s23, calculating concept, and optimal threshold ε is set.In present embodiment, optimal threshold ε can tie Close Conventional wisdom and analysis expert to obtain through test.

Classification relation system in body is to arrange concept with tree according to subclassof, traditional based on language The Concept Semantic Similarity algorithm of adopted distance can reasonably utilize this type of organization.Traditional general based on semantic distance Read on the basis of Arithmetic of Semantic Similarity, the present invention has carried out certain optimization and adjustment, by level depth, node density Take into account, and be provided with regulation parameter, using a kind of improved Concept Semantic Similarity algorithm.

Semantic similarity between concept is calculated using Concept Semantic Similarity algorithm.Concept c₁And c₂Between semantic phase Like degree sim (c₁, c₂) computing formula is as follows:

sim(c₁,c₂)=[sim_dist (c₁,c₂)]^α×[sim_depth(c₁,c₂)]^β×[sim_width(c₁,c₂) ]^γ,

Wherein, sim_dist (c₁, c₂) for semantic similarity and semantic distance corresponding relation, sim_depth (c₁, c₂For Semantic similarity and the corresponding relation of semantic depth, sim_width (c₁, c₂) right for semantic similarity and semantic node density Should be related to；α represents the weight of semantic distance, the weight of β representational level depth, and γ represents the weight of node density, and α, β, γ are Regulation parameter, and α, β, γ sum is 1.

Concept Semantic Similarity value and semantic distance are inversely proportional to.

Wherein, dist (c₁, c₂) representing semantic distance, a is to adjust Parameter；If the weight of each edge is the same, weight is all set to 1, then

dist(c₁,c₂)=len (c₁,c₂),

len(c₁,c₂) represent concept c₁And c₂Between path.

Concept Semantic Similarity value is directly proportional to level depth difference, with level depth and being inversely proportional to.

Depth (c) represents the depth of concept c, Depth (c)=depth (parent (c))+1, parent (c) represents the father node of concept c.Depth depth of root node (root)=1.

Concept Semantic Similarity value and node density are inversely proportional to.In classification relation system, density higher finger classification is thinner Cause, degree of refinement is higher, and between concept and concept, similarity degree is lower.The density of concept c is expressed as width (c), its with classification Degree of refinement is directly proportional, as width (c₁)≤width(c₂),

$sim\_width(c_1,c_2)=\frac{width\left(c_1\right)}{width\left(c_2\right)}.$

So:

$\begin{array}{c}sim(c_1,c_2)={\&lsqb;sim\_dist(c_1,c_2)\&rsqb;}^{\mathrm{\&alpha;}}\&times;{\&lsqb;sim\_depth(c_1,c_2)\&rsqb;}^{\mathrm{\&beta;}}\&times;{\&lsqb;sim\_width(c_1,c_2)\&rsqb;}^{\mathrm{\&gamma;}}\\ ={\&lsqb;\frac{a}{a+dist(c_1,c_2)}\&rsqb;}^{\mathrm{\&alpha;}}\&times;{\&lsqb;1-\frac{|depth\left(c_1\right)-depth\left(c_2\right)|}{depth\left(c_1\right)+depth\left(c_2\right)}\&rsqb;}^{\mathrm{\&beta;}}\&times;{\&lsqb;\frac{width\left(c_1\right)}{width\left(c_2\right)}\&rsqb;}^{\mathrm{\&gamma;}}\end{array}.$

S24, set up Concept Semantic Similarity set s (s₁, s₂..., s_n) and concordance, accuracy and terseness judgement knot Fruit set t (t₁, t₂..., t_n), s_i=sim (v_i, v_j), i, j are natural numbers, and concept vi is concept vj hypernym, and v_i≠ v_j.

In judged result set t, t_nCorresponding with concept vn, if concept vn meets evaluation index, correspond to t_nEnter rower Note, specifically can label symbol " √ "；If concept vn does not meet evaluation index, the type of error that corresponding tn mark is judged, Type of error is classification error or redundancy error.

If s25 concept vi hypernym number is more than 1, judge the classification error of concept vi and in judged result set t (t₁, t₂..., t_n) in corresponding t_iRecord sort mistake.

S26, by sim (v_i, v_j) compare with optimal threshold ε, and judge whether concept vi and concept vj accord with according to comparative result Close classification relation, meet, then by s_iIt is stored in Concept Semantic Similarity set s (s₁, s₂..., s_n)；Do not meet and then judge concept vi Classification error simultaneously will determine that result is stored in judged result set t (t₁, t₂..., t_n).

Specifically, if ε is ＜ sim (v_i, v_j) ＜ 1, then concept v_iWith concept v_jMeet classification relation requirement, by s_i=sim (v_i, v_j) it is stored in Concept Semantic Similarity set s；If 0≤sim is (v_i, v_j) ＜ ε, then judge concept v_iWith concept v_jThere is classification Mistake, is unsatisfactory for accuracy, corresponding t in judged result set t_i、t_jRecord sort mistake.

S27, judge whether sim (v_i, v_j)=sim (v_i, v_k), if it is present judging concept v_jWith concept v_kIt is There is redundancy error and be stored in judged result set t (t in synonym₁, t₂..., t_n), specifically, in judged result set t Corresponding t_j, t record redundancy error.

S28, repeat step s25 to s27, until all calculating and having judged Ontological concept set v and classification relation set e Finish.

S29, output Similarity Measure results set s and concordance, accuracy and terseness judged result set t.

According to relation not same-action in the body, classification relation and two kinds of non-categorical relation can be divided into.Non-categorical closes System illustrates all other all relations being not belonging to classification relation such as cause effect relation, relation on attributes, example relationship.Non-categorical closes System enhances completeness and the complexity of ontology knowledge expression, makes body really become network structure.Therefore, to non-categorical relation The efficiency assessment that carries out have important theory significance and researching value.

Non-categorical relation has many kinds, in agriculture field body, wherein function the strongest, using most be object properties Relation.The present invention is mainly estimated to the object properties relation in non-categorical relation system, proposes a kind of theoretical based on figure Non-categorical relationship assessment method between agriculture field Ontological concept.

With reference to Fig. 4, in present embodiment, between assessment agriculture field Ontological concept, non-categorical relation is that step s3 specifically includes Below step by step.

S31, parsed according to object properties relation pair agriculture field body using analytical tool, will the genus that obtains of parsing Sexual intercourse figure saves as data structureData file, and represented with adjacency matrix a；Wherein, v is this Body concept set, e be relation on attributes set,For relation on attributes concept pair.

The evaluation index of object properties relation between s32, setting agriculture field Ontological concept, and sentenced according to evaluation index setting Surely set q.

S33, according to evaluation index, between agriculture field Ontological concept, object properties relation is estimated, and by assessment result Record is in judging set q.

In present embodiment, between agriculture field Ontological concept, the evaluation index of object properties relation includes following six.

1. reversibility: there is relation r between concept a and concept b, meanwhile, there is relation r between concept b and concept a^-1, R is then claimed to have reversibility.

The judgement of " reversibility " evaluation index can be carried out using the theoretical relevant knowledge of figure.If in adjacency matrix a=(a_ij) in Find a_ij=1, a_ji=1, and fromFind a_ijCorresponding relation on attributes r1 and a_jiCorresponding relation on attributes r2 is inverse each other closing System, then meet " reversibility ".Will determine that result record in set of matrices q.

2. relation inheritance: there is relation r between concept a and father's concept b, also deposit between concept a and sub- concept c meanwhile In relation r, then relation r is claimed to have relation inheritance.

The judgement of relation inheritance feature needs combining classification relation system.If a_ij=1, a_ik=1, and fromFind a_ijWith a_ikCorresponding relation on attributes name r is also identical, and finds v_jAnd v_kBelong to father's subclass relation in classification relation system, then meet and " close It is inheritance "；If a_ij=1, a_ik≠ 1, then it is unsatisfactory for " relation inheritance ".And will determine that result record in set of matrices q.

3. inverse relationship inheritance: there is relation r between father's concept and concept a, meanwhile, its sub- concept and concept a it Between there is also relation r, then claim relation r there is inverse relationship inheritance.

The judgement of inverse relationship inheritance feature needs combining classification relation system.If a_ik=1, a_jk=1, and fromFind a_ikAnd a_jkCorresponding non-categorical relation name r is also identical, and finds v_iAnd v_jBelong to father's subclass relation in classification relation system, then full Foot " the reverse inheritance of relation "；If a_ik=1, a_jk≠ 1, then it is unsatisfactory for " inverse relationship inheritance ".And will determine that result record exists In set of matrices q.

4. transitivity: if there is relation r between concept a and concept b, between concept b and concept c, there is relation r, meanwhile, There is relation r between concept a and concept c, then claim relation r to have transitivity.

If in adjacency matrix a=(a_ij) in find a_ij=1, a_jk=1, and a_ik=1, and fromFind relation on attributes name r Identical, i.e. concept v_i, v_kBetween relation on attributes contradict with " there is no transitivity " this property；?In find out a_ij, a_jk, a_ikCorresponding relation on attributes name, and will determine that result record in set of matrices q.

5. symmetry: if concept a and concept b have relation r, be denoted as a=r (b), then concept b and concept a have relation r, are denoted as B=r (a).

If in adjacency matrix a=(a_ij) in find a_ij=1, a_ji=1, and fromFind that relation on attributes name is also identical, that is, generally Read v_iAnd v_jContradict with relation on attributes " there is no symmetry " this property, find concept v_iAnd v_jAnd institute is right between them The relation on attributes answered, and will determine that result record in set of matrices q.

6. reflexivity: concept a has relation r, if r has reflexivity, a=r (a).

If in adjacency matrix a=(a_ij) in found a_iiWhen=1 (1≤i≤n), then this result and relation on attributes " do not have Having reflexivity " this property contradicts, and finds a_iiCorresponding concept and relation on attributes, and will determine that result record in matrix stack Close in q.

Relation on attributes refers to the attribute that certain concept is another concept, describes " property of things and its mutual pass System ".Wherein, object properties relation is used for associating two concepts, the non-categorical relation between expression concept, and relation on attributes has inverse The constraint such as reflexive, symmetry, transitivity, its domain of definition and codomain determine two classes of association respectively, and attribute-name can be according to reality Situation is named.Six properties being had according to relation on attributes: have reversibility, relation inheritance, inverse relationship inheritance and There is no transitivity, symmetry and reflexivity.In present embodiment, six evaluation index reflexivitys of setting, symmetry, can Inverse property, transitivity, relation inheritance, inverse relationship inheritance are true to the assessment of the relation on attributes in non-categorical relation system The real extent of reaction is very high.

S34, according to above step by all concepts in Ontological concept set v and relation on attributes set e and relation on attributes All after the completion of decision analysis, the relation on attributes set e in output agriculture field body, relation on attributes concept pairAnd judge Results set q.

Specifically commented using the tea-plant pests body of above agriculture field body efficiency assessment method below in conjunction with a kind of Estimate system above method is further explained.

By verifying feasibility and the effectiveness of proposed agriculture field body appraisal procedure, java language specifically can be used Speech, calls ontology development bag jena, completes the prototype system of tea-plant pests body assessment in eclipse development platform.Should The development environment that system is used is window7, and development platform is eclipse3.3, and jdk version for 1.8, jena version is jena2.6.4.This prototype system can be parsed to tea-plant pests body and be assessed.

Tea-plant pests body assessment system includes tea-plant pests body pretreatment module, the Concept Evaluation of tea-plant pests body Non-categorical relationship assessment between the concept of classification relation evaluation module and tea-plant pests body between module, the concept of tea-plant pests body Module.With reference to Fig. 5, the operating procedure of the system includes.

A) the tea-plant pests body of structure is imported body assessment system.

B) the tea-plant pests body that imports enters body pretreatment module, by call jena bag to the concept in body, Classification relation, non-categorical relation etc. are parsed.

C) in the Concept Evaluation module of tea-plant pests body, using step 1) assessment agriculture field Ontological concept side Method, calculates recall ratio, precision ratio and f index respectively, and exports the assessment result of recall ratio, precision ratio and f1 index.

D) in classification relation evaluation module between the concept of tea-plant pests body, using step 2) assessment agriculture field this The method of classification relation between body concept, obtain classification relation between agriculture field Ontological concept Similarity Measure results set s and Concordance, accuracy and terseness judged result, and export.

E) in non-categorical relationship assessment module between the concept of tea-plant pests body, using step 3) assessment agriculture field Between Ontological concept, the method for non-categorical relation is estimated respectively to six evaluation indexes of relation on attributes, and exports.This step Specifically, the relation on attributes in tea-plant pests body, after parsing, is shown in the form of relation on attributes name, domain of definition, codomain In systems, and using the non-categorical relationship assessment method theoretical based on figure six evaluation indexes of relation on attributes are carried out respectively Assessment, and show non-categorical relationship assessment result in systems.

By the test analysis to system, the present invention achieves good effect, demonstrates the feasibility of the method and has Effect property.Tea-plant pests body assess prototype system using based on domain lexicon Concept Evaluation method, based on Concept Semantic similar The classification relation appraisal procedure of degree and the non-categorical relationship assessment method based on figure theory can be with effectively solving agriculture field bodies Evaluation problem, the effectively quality of the assessment agriculture field body of structure, assess existing automatically or semi-automatically ontological construction The feasibility of method, instructs Ontology engineering teacher to construct higher-quality Agricultural ontology, provides more for agriculture field associated user Suitable Agricultural ontology, for the rationale carrying out Agricultural ontology assessment in a deep going way and key technology research, has certain showing Sincere justice and practical study are worth.

The above, the only present invention preferably specific embodiment, but protection scope of the present invention is not limited thereto, Any those familiar with the art the invention discloses technical scope in, technology according to the present invention scheme and its Inventive concept equivalent or change in addition, all should be included within the scope of the present invention.

Claims (10)

1. a kind of agriculture field body efficiency assessment method is it is characterised in that comprise the following steps:

S1, assessment agriculture field Ontological concept；

Classification relation between s2, assessment agriculture field Ontological concept；

Non-categorical relation between s3, assessment agriculture field Ontological concept；

Step s1 specifically include following step by step:

S11, agriculture field body to be assessed is parsed according to concept, obtained concept set；

S12, setting Concept Evaluation index, Concept Evaluation index includes recall ratio r, precision ratio p and f1 index, wherein,

It is defined on recall ratio and in the case that precision ratio occurs contradiction, harmonic average is weighted to this recall ratio and precision ratio The value obtaining is f index,

$f=\frac{({\mathrm{\&alpha;}}^2+1)p\&times;r}{{\mathrm{\&alpha;}}^2(p+r)}$

F1 index is f index during α=1,

$f1=\frac{2\&times;p\&times;r}{p+r}$

S13, the ambit scope according to body to be assessed, select or build corresponding domain lexicon；

S14, it is that recall ratio, precision ratio and f1 exponent pair concept are estimated according to Concept Evaluation index, obtain correct concept number；

S15, the assessment result of output recall ratio, precision ratio and f1 index；

Step s2 specifically include following step by step:

S21, by agriculture field body to be assessed according to concept between subclassof relation parsed, and be saved as Data structure be figure g=<v, e>data file, v be Ontological concept set, e be classification relation set；

S22, three evaluation indexes of setting classification relation assessment, are concordance, accuracy and terseness respectively；

Semantic similarity between s23, calculating concept, and optimal threshold ε is set；

S24, set up Concept Semantic Similarity set s (s₁, s₂..., s_n) and concordance, accuracy and terseness judged result collection Close t (t₁, t₂..., t_n), s_i=sim (v_i, v_j), i, j are natural numbers, and concept vi is concept vj hypernym, and v_i≠v_j；

S25, judge that whether concept vi hypernym number is more than 1, be then to judge the classification error of concept vi and be stored in judged result collection Close t (t₁, t₂..., t_n)；

S26, by sim (v_i, v_j) compare with optimal threshold ε, and concept vi is judged according to comparative result and whether concept vj meets point Class relation, meets, then by s_iIt is stored in Concept Semantic Similarity set s (s₁, s₂..., s_n)；Do not meet and then judge that concept vi is classified Mistake simultaneously will determine that result is stored in judged result set t (t₁, t₂..., t_n)；

S27, judge whether sim (v_i, v_j)=sim (v_i, v_k), if it is present judging concept v_jWith concept v_kOccur superfluous Remaining mistake is simultaneously stored in judged result set t (t₁, t₂..., t_n)；

S28, repeat step s25 to s27, finish until Ontological concept set v and classification relation set e is all calculated judgement；

S29, output Similarity Measure results set s and concordance, accuracy and terseness judged result set t；

Step s3 specifically include following step by step:

S31, parsed according to object properties relation pair agriculture field body, the attributed relational graph that parsing is obtained saves as number According to structure it isData file, and represented with adjacency matrix a；V is Ontological concept set, e is relation on attributes Set,For relation on attributes concept pair；

The evaluation index of object properties relation between s32, setting agriculture field Ontological concept, and collection is judged according to evaluation index setting Close q；

S33, according to evaluation index, between agriculture field Ontological concept, object properties relation is estimated, and by assessment result record In judging set q；

Relation on attributes set e in s34, output agriculture field body, relation on attributes concept pairAnd result of determination set q.

2. agriculture field body efficiency assessment method as claimed in claim 1 is it is characterised in that domain term in step s13 The structure of allusion quotation comprises the following steps:

S131, determine subject fields scope, and assembling sphere vocabulary；

S132, determine the classification characteristicses of vocabulary, mutual relation between organized words and vocabulary, form vocabulary structure, and specify attribute Dictionary；

S133, the vocabulary to dictionary and relation carry out examination ＆ verification and revise.

3. agriculture field body efficiency assessment method as claimed in claim 2 is it is characterised in that in step s131, from literary composition Offer, subject classical works, Baidupedia, existing dictionary and expertise approach obtain Field Words.

4. agriculture field body efficiency assessment method as claimed in claim 1 is it is characterised in that in step s23, concept c₁ And c₂Between semantic similarity sim (c₁, c₂) computing formula is as follows:

sim(c₁,c₂)=[sim_dist (c₁,c₂)]^α×[sim_depth(c₁,c₂)]^β×[sim_width(c₁,c₂)]^γ,

Wherein, sim_dist (c₁, c₂) for semantic similarity and semantic distance corresponding relation, sim_depth (c₁, c₂For semanteme Similarity and the corresponding relation of semantic depth, sim_width (c₁, c₂) close for semantic similarity is corresponding with semantic node density System；α represents the weight of semantic distance, the weight of β representational level depth, and γ represents the weight of node density, and α, β, γ are to adjust Parameter, and α, β, γ sum is 1.

5. agriculture field body efficiency assessment method as claimed in claim 4 it is characterised in thatWherein, dist (c₁, c₂) representing semantic distance, a is regulation parameter；dist (c₁,c₂)=len (c₁,c₂), len (c₁,c₂) represent concept c₁And c₂Between path.

6. agriculture field body efficiency assessment method as claimed in claim 4 it is characterised in thatThe depth of depth (c) expression concept c, depth (c)= Depth (parent (c))+1, parent (c) represents the father node of concept c it is preferable that depth depth (root) of root node =1.

7. agriculture field body efficiency assessment method as claimed in claim 4 is it is characterised in that the density of concept c represents For width (c), it is directly proportional to classification degree of refinement, as width (c₁)≤width(c₂),

$sim\_width(c_1,c_2)=\frac{width\left(c_1\right)}{width\left(c_2\right)}.$

8. agriculture field body efficiency assessment method as claimed in claim 1 is it is characterised in that in step s26, if ε is ＜ sim(v_i, v_j) ＜ 1, then concept v_iWith concept v_jMeet classification relation requirement；If 0≤sim is (v_i, v_j) ＜ ε, then judge concept v_i With concept v_jThere is classification error and be stored in judged result set t.

9. agriculture field body efficiency assessment method as claimed in claim 1 is it is characterised in that in judged result set t, t_nCorresponding with concept vn, if concept vn meets evaluation index, correspond to t_nIt is labeled；If concept vn does not meet assessment referred to Mark, then correspond to the type of error that tn mark is judged.

10. agriculture field body efficiency assessment method as claimed in claim 1 is it is characterised in that agriculture field body is general Between thought, the evaluation index of object properties relation includes:

Reversibility: there is relation r between concept a and concept b, meanwhile, there is relation r between concept b and concept a^-1, then claim r There is reversibility；

Relation inheritance: there is relation r between concept a and father's concept b, meanwhile, there is also relation between concept a and sub- concept c R, then claim relation r to have relation inheritance；

Inverse relationship inheritance: there is relation r between father's concept and concept a, also deposit between its sub- concept and concept a meanwhile In relation r, then relation r is claimed to have inverse relationship inheritance；

Transitivity: if there is relation r between concept a and concept b, between concept b and concept c, there is relation r, meanwhile, concept a with There is relation r between concept c, then claim relation r to have transitivity；

Symmetry: if concept a and concept b have relation r, be denoted as a=r (b), then concept b and concept a have relation r, are denoted as b=r (a)；

Reflexivity: concept a has relation r, if r has reflexivity, a=r (a).