CN101566988A - Method, system and device for searching fuzzy semantics - Google Patents

Abstract

The invention discloses a method for searching fuzzy semantics, which comprises the following steps: key words standing for fuzzy concepts and key words standing for mood arithmetic operators, which are input by a user, are reasoned and calculated by using a fuzzy body knowledge base, the key words generated by reasoning and calculating are combined as an expanded query condition, a query result which conforms to the condition is searched in a resource indexing base according to the expanded query condition, and meanwhile, the invention also discloses a method and a device for searching fuzzy semantics. The invention can expand the key words into equivalent concepts, sub-concepts, and the like through the key word analysis based on the body and can treat the fuzzy key word to a certain extent in order to increase the recall ratio.

Description

A kind of searching fuzzy semantics, system and equipment

Technical field

The present invention relates to communication technical field, relate in particular to a kind of searching fuzzy semantics, system and equipment.

Background technology

Web (network) search technique has obtained popularization and application, but recall ratio and degree of accuracy still can not satisfy user's demand.Present existing search engine major part is based on keyword or based on the retrieval of content of text, can not gives full expression to semantic information.The semantic search technology can be improved the search effect of current search engine, comprises recall ratio and precision ratio, and as one of main application of following Semantic Web, semantic search will produce great influence to people's life.

The fusion ontology can be realized semantic search to a certain extent, but has brought new problem based on the ontology knowledge storehouse of classical description logic, promptly can't be described and reasoning fuzzy message.Though the description logic descriptive power is very strong, body is used also very extensive, and description logic can't be handled the fuzzy concept in the fuzzy message, as " excellence ", " youth " etc.Therefore description logic is blured expansion, realize having very strong Practical significance based on the fuzzy semantics search of fuzzy domain body.

In addition, the number of times that occurs based on keyword in the document of traditional information retrieval (IR) technology great majority.Though the XML searching system considers that also it is simple that the relative semantic net of its data model structure is wanted with the combining of structure query and content retrieval, existing method can not satisfy semantic web retrieval fully.Current increasing Web information resources have been carried out semantic tagger, and to meet RDF (Resource DescriptionFramework, resource description framework) or the semantic net language of XML (EXtensible Markup Language, extensible markup language) grammer be described.In the face of a large amount of semantic net informations, how to provide and retrieve more effective visit and more rational result for retrieval than conventional information and become one of major issue that semantic search faces.

The overwhelming majority adopts the classical description logic based on the system of description logic at present, comprises its representation of knowledge and knowledge reasoning.Along with the popularization of OWL (ontology describing language), present most of main body system also all adopts the classical description logic to support as its basic logic.Yet as above saying, the classical description logic has the defective that is difficult to overcome in the face of fuzzy message, and therefore the fuzzy expansion to description logic also becomes the research focus gradually.Yet the vague description logic also only rests on conceptual phase, is also never implemented.In addition, the fuzzy research of expanding also only only limits to the 1-pattern paste expansion of description logic at description logic, promptly uses the degree of membership value of determining to describe ambiguity, can not well be applied to real world applications.

In realizing process of the present invention, the inventor finds:

Existing search engine still has bigger room for promotion looking on complete, the precision ratio, simultaneously, can't realize complicated constraint inquiry based on the tradition inquiry of keyword, and can't realize the complex relationship inquiry between resource, causes the recall ratio of user semantic search low.

Summary of the invention

The embodiment of the invention provides a kind of searching fuzzy semantics, system and equipment, to improve the recall ratio of user semantic search.

The embodiment of the invention provides a kind of searching fuzzy semantics, may further comprise the steps:

To the keyword of the representative fuzzy concept of user input and the keyword of represent mood operator, utilize described fuzzy ontology knowledge base to carry out reasoning and calculation, and the keyword sets cooperation that reasoning and calculation produces is the querying condition after expanding;

According to the querying condition after the described expansion, in described resource index storehouse, retrieve qualified Query Result.

The embodiment of the invention provides a kind of fuzzy semantics search system, comprises semantic search node, at least one resource website and internal network, and described semantic search node specifically comprises:

Ontology knowledge storehouse node is used to store fuzzy domain body, and described domain body is realized with the OWL file;

The crawl device node is used for being responsible for the centralized intranet resources of creeping, and obtains resource content and sets up index, safeguards URL information and the setting scope of creeping;

Search node is used for the user and selects inquiry mode voluntarily;

Inference node is used for reasoning and realizes semantic search reasoning service, and returns The reasoning results and submit to traditional search engines or directly return to the user.

The embodiment of the invention provides a kind of semantic search node, comprising:

Ontology knowledge storehouse node is used to store fuzzy domain body, and described domain body is realized with ontology describing language OWL file;

The crawl device node is used for being responsible for the centralized intranet resources of creeping, and obtains resource content and sets up index, safeguards resource locator URL information and the setting scope of creeping;

Search node is used for the user and selects inquiry mode voluntarily;

Inference node is used for reasoning and realizes semantic search reasoning service, and returns The reasoning results and submit to traditional search engines or directly return to the user.

In the embodiment of the invention,, can be its equal notion, sub-notion etc. with keyword expansion, and can handle fuzzy keyword to a certain extent, thereby improve recall ratio by keyword resolution based on body.

Description of drawings

Fig. 1 is a semantic search system construction drawing in the embodiment of the invention;

Fig. 2 is the schematic flow sheet of creeping in the embodiment of the invention;

Fig. 3 is a searching fuzzy semantics process flow diagram in the embodiment of the invention;

Fig. 4 is an inference method process flow diagram in the embodiment of the invention;

Fig. 5 is a graphical customization semantic query process flow diagram flow chart in the embodiment of the invention;

Fig. 6 a is that the conceptual retrieval result generates page synoptic diagram in the embodiment of the invention;

Fig. 6 b is that the fuzzy concept result for retrieval generates page synoptic diagram in the embodiment of the invention;

Fig. 7 is a graphical customization conceptual schematic view in the embodiment of the invention;

Fig. 8 is a graphical customization data attribute synoptic diagram in the embodiment of the invention;

Fig. 9 is a graphical customized objects attribute synoptic diagram in the embodiment of the invention.

Embodiment

The embodiment of the invention provides a kind of searching fuzzy semantics based on the field, with Semantic Web Technology, fuzzy logic, combine with search engine technique, obtains the required information of user efficiently and accurately.The embodiment of the invention is a point of penetration to solve the description logic limitation, and the 2-pattern that has proposed the support description logic SHOIN (D) of OWL is stuck with paste extended method.Though be widely used based on the reasoning algorithm of description logic body owing to its powerful descriptive power and maturation, yet the classical description logic is confined to handle definite notion and relation, thereby causes the fuzzy knowledge in the large paper edition systems such as description logic intractable similar semantic net.Though 1-type fuzzy set can alleviate the influence that uncertainty is brought to a certain extent, it adopts method formula that definite degree of membership value decides blur level inadequately accurately.By comparison, the system based on 2-type fuzzy set can utilize the degree of membership interval more accurately to describe fuzzy message.

Based on this, embodiment of the invention utilization when the user does not know the detail of submitting inquiry to, can be provided with and submit to the fuzzy query request based on the expression and the reasoning of fuzzy logic.In conjunction with the fuzzy concept in the body, can submit the search service that has certain fog-level to the user by the triplet sets that has fuzzy message that fuzzy reasoning generates, thereby reduce the raising that realizes recall ratio and precision ratio under the semantic situation of losing.

Because the customization of fuzzy query statement is complicated, also needs subsidiary condition such as mood operator sometimes, therefore adopts the method for keyword to realize the fuzzy query poor effect.Therefore the embodiment of the invention is also utilized patterned way customization semantic query, and the user can be provided with complicated constraint condition inquiry, utilizes patterned way to realize the user customizable query statement, thereby accurately obtains semantic information, reaches the purpose that improves precision ratio.

It is Type-2Fuzzy SHOIN (D) (2-FSHOIN (D)) that the embodiment of the invention provides the 2-pattern of support description logic SHOIN (D) to stick with paste extended method.

2-FSHOIN (D) defines A, and C and R are the atom fuzzy concept, complicated fuzzy concept and the set of fuzzy role relation.Can draw

With

It is fuzzy concept.Mapping relations among the 2-FSHOIN (D) can be expressed as I=(Δ ^I, ^I), wherein ^IBe with fuzzy concept and the relationship map mapping function to the membership values interval: C ^I=Δ ^I→ [a, b] and R ^I=Δ ^I* Δ ^I→ [a, b], a wherein, b satisfies 0≤a≤b≤1.The mapping of 2-FSHOIN (D) ^IMust satisfy following equation:

For any example d ∈ Δ ^IHave:

⊥ ^I(d)＝[0，0]

C ^I(d)＝[μ ^L(C(d))，μ ^U(C(d))]

$\⫬C^I\left(d\right)=[1-{\mathrm{\μ}}^U\left(C\left(d\right)\right),1-{\mathrm{\μ}}^L\left(C\left(d\right)\right)]$

${(\∀R.C)}^I\left(d\right)={{\inf }_{d^{,}\∈\mathrm{\Δ}}}^I\left[S\right\{1-{\mathrm{\μ}}^U\left(R(d,d^{,})\right),{\mathrm{\μ}}^L\left(C\left(d^{,}\right)\right)\},S\{1-{\mathrm{\μ}}^L\left(R(d,d^{,})\right),$

${\mathrm{\μ}}^U\left(C\left(d^{,}\right)\right)\left\}\right]$

${(\∃R.C)}^I\left(d\right)={{\sup }_{d^{,}\∈\mathrm{\Δ}}}^I\left[T\right\{{\mathrm{\μ}}^L\left(R(d,d^{,})\right),{\mathrm{\μ}}^L\left(C\left(d^{,}\right)\right)\},T\{{\mathrm{\μ}}^U\left(R(d,d^{,})\right),$

${\mathrm{\μ}}^U\left(C\left(d^{,}\right)\right)\left\}\right]$

(d wherein _i≠ d _j):

${(\≤\mathrm{nR})}^I\left(d\right)=\⫬{(\≥n+1R)}^I\left(d\right)$

${(\≤\mathrm{nR}.C)}^I\left(d\right)=\⫬{(\≥n+1R.C)}^I\left(d\right)$

T wherein, S is called the triangle modular arithmetic in the fuzzy set, and in fuzzy set theory, the computing of fuzzy set can only be determined with its subordinate function, and the computing meeting of different definition produces different results.The various nonidentity operations of therefore setting up fuzzy set can adapt to different bloomings, and modular arithmetic is the most general form of fuzzy set computing.

Mapping I is called three angle moulds, if satisfy condition:

(1)I(0，0)＝0，I(1，1)＝1

(2) $a\≤c,b\≥d\⇒I(a,b)\≤I(c,d)$

(3)I(a，b)＝I(b，a)

(4)I(I(a，b)，c)＝I(a，I(b，c))

When three angle moulds satisfy I (a, 1)=a (a ∈ [0,1]), be called the T mould; When three angle moulds satisfy I (0, a)=a (a ∈ [0,1]), be called the S mould.

Based on this, the embodiment of the invention provides a kind of searching fuzzy semantics based on fuzzy domain body.Semantic search is not that the resource website on all Internet is searched for, but the resource website in a certain field is searched for.The described searching fuzzy semantics based on fuzzy domain body of the embodiment of the invention comprises following steps:

Step 101 is set up renewable field fuzzy ontology knowledge base.

Step 102, crawl device is set up the resource index storehouse to inner resource website.

Step 103 receives the query requests that the user proposes.

Step 104, inference engine is analyzed the query requests that the user proposes, and finishes necessary ontology knowledge storehouse reasoning, and The reasoning results is returned to search utility as the querying condition after expanding.

Step 105, the expansion condition that returns of engine retrieves qualified Query Result in index database by inference.

Step 106 to the Search Results rearrangement, is combined into the complete results page and submits to the user in conjunction with The reasoning results.

The embodiment of the invention provides a kind of fuzzy semantics search system based on fuzzy domain body, as shown in Figure 1, comprises semantic search node, resource website (can be a plurality of arbitrarily) and internal network three parts.Wherein the semantic search node comprises: ontology knowledge storehouse node 101, crawl device node 102, search node 104 and inference node 103.

Wherein, ontology knowledge storehouse node 101 is used to store fuzzy domain body, and domain body realizes that with the OWL file ontology file is to determine body in this field.In order to operate the OWL body efficiently, adopt the persistent storage instrument of SQLServer as body.Owing to read and resolve in the OWL file and infer RDF graph structure information and extremely waste resource at every turn, (Jena API only provides MySQL to store the OWL graph structure interface of SQLServer database backstage into, the database interface of Oracle), directly reading ontology model from database like this when using just can save time and resource.

Set up the fuzzy ontology knowledge base following step is specifically arranged: at first, set up classical body K, adopt automanual body constructing method; Then, in the Tbox of body K, add fuzzy concept, and the degree of membership of atom fuzzy concept is set.Degree of membership is defined as: example is to the subjection degree of notion, and 2-FSHOIN (D) uses the interval between [0,1] to describe fuzzy membership; At last, increase the example of fuzzy concept among the K, and the calculated examples property value, draw or be provided with the be subordinate to interval of example to fuzzy concept, finish the structure of fuzzy ontology, make classical ontology knowledge storehouse K expand to fuzzy ontology knowledge base K '.

Crawl device node 102 is used for being responsible for the centralized intranet resources of creeping, and obtains resource content and sets up index, safeguards URL (resource locator) information and the setting scope of creeping.Workflow as shown in Figure 2, concrete steps are as follows:

1. obtain beginning resource node URL information.

2. obtain this website host and the scope of creeping is set, be set to creep in the in-house network scope herein.

3. creep webpage and set up index.

4. regularly carry out and 1.-3. upgrade search node server-side index file.

Inference node 103 is used for reasoning and realizes semantic search reasoning service, and returns The reasoning results and submit to traditional search engines or directly return to the user.Should be able to inquire about and Query By Example by real concept, more powerful inference function is provided.

Owing to adopt fuzzy ontology to carry out the reasoning service, traditional reasoning algorithm improved.The realization of reasoning supposes that based on the Tableau algorithm that has proposed conceptual description is C and D, and the Tableaux algorithm uses negates but not directly judge the two relation of inclusion: $C\⊆D$ And if only if Be unsatisfiable, can satisfy relation thereby the notion containment relationship is converted into.Before utilization Tableaux algorithm, use De Morgan's laws etc. make that to carrying out conversion in the conceptualization formula all negate only to appear at before the notion name.If E is Negative normal form, be satiable if the Tableaux algorithm attempts to prove E, must explain that I makes by one of structure so

Promptly at Δ ^IIn must to have individuality be E ^IAn element.

Description logic needs to carry out six rules in the tableau on basis algorithm implementation, and judges whether to generate bottom notion ⊥, and the bottom notion that the reasoning of fuzzy ontology generates also has the fuzzy concept of blur level interval less than threshold value except ⊥.Fuzzy ontology has defined the fuzzy concept class in order to handle fuzzy message, and reasoning algorithm by the interval computation rule that is subordinate to of each subclass definitions of fuzzy concept class, is derived the fuzzy value that example is subordinate to fuzzy concept in order to determine the fuzzy membership interval.The fuzzy membership interval of example is described in the interval of system's usable range in [0,1].Concrete Fuzzy Logic Reasoning Algorithm flow process is as follows:

(1) extracts fuzzy concept and belong to the example collection A of this fuzzy concept according to the fuzzy threshold value retrieval that is provided with;

(2) find the notion of the keyword correspondence of removing fuzzy concept and therewith conceptual dependency father's notion, equate the example collection B of notion and sub-notion correspondence;

(3) get set C=A ∩ B and preservation.Hypothesis instance d is [μ to the interval that is subordinate to of the fuzzy word D of fuzzy concept correspondence ^L _D(d), μ ^U _D(d)], then d to the degree of membership μ of the fuzzy word E that sews mood operator _E, (d)=[μ ^L _D(d) ^t, μ ^U _D(d) ^t].When t＞1 is called the centralization mood operator, otherwise be called the undisciplineization mood operator.According to the t value of mood operator correspondence, the example among the C is carried out the calculating of degree of membership;

(4) according to the degree of membership threshold value example after calculating is filtered, promptly be subordinate to the interval and from example collection C, reject less than the example of threshold value.The following rule of relatively employing wherein interval and numerical value: for interval [a, b] and numerical value t, if a＞t then claim the interval greater than threshold value; If b＜t then claims interval less than threshold value; Otherwise interval and threshold value can't compare;

(5) return example collection C in proper order from large to small according to the interval, the comparison rule between the interval is as follows: for interval C ₁=[a, b], C ₂=[c, d] has $h_1=\frac{(a+b)}{2},$ $h_2=\frac{(c+d)}{2},$ If h ₁＜h ₂Then claim interval C ₁＜C ₂If instead h ₁＞h ₂Claim C ₁＞C ₂If h ₁=h ₂If then (a+b)＜(c+d) then claim C ₁＜C ₂If instead (a+b)＞(c+d) then claim C ₁＞C ₂If (a+b)=(c+d) then claim C ₁=C ₂

Search node 104 is used for the user and selects three kinds of different inquiry modes voluntarily: inquiry, concept queries based on keyword reach by the self-defined fuzzy query of patterned way.

Submit queries condition (as keyword etc.), inquiry is sent to inference engine, inference engine is under knowledge base is supported, searching keyword is expanded, realization is based on the keyword resolution of body, keyword after keyword and the parsing is sent to search utility together, retrieve the link that meets querying condition in the search utility indexed file, and the result after will sorting returns to the user.

The user submits concept queries to, and inquiry is sent to inference engine, and inference engine obtains the conceptional tree of conceptual dependency therewith under knowledge base is supported, infer all examples that belong to this notion simultaneously.Example is sent to search utility as searching keyword, retrieves the link that meets querying condition in the search utility indexed file, and the result after will sorting is combined into the Query Result page with The reasoning results and returns to the user.

The user also can pass through patterned way customized graphics semantic query, and the semantic query that customization is good is submitted to inference engine, determines that by inference engine the user need inquire about and satisfy the example of constraint condition.The user adopts patterned way customization fuzzy query, search system at first is provided with fuzzy threshold value, its implication is: when example to the degree of membership of fuzzy concept during greater than this fuzzy threshold value, think that this example can return as Search Results, otherwise think that the degree of membership of this example is not enough so that it becomes Search Results.At first choose the fuzzy concept in the fuzzy ontology when user uses, select to be fit to the mood operator of query requests degree then, as " very ", " very ", " summary " etc.After the submit queries, mood operator is at first judged by system, and the fuzzy threshold value that changes system settings according to mood operator (centralized mood operator increases fuzzy threshold value, careless and sloppy formula mood operator reduces fuzzy threshold value), then the fuzzy threshold value after example degree of membership and the change is compared, return the example list of degree of membership greater than fuzzy threshold value.

Embodiment of the invention specific embodiments: consult Fig. 3, handle according to following steps when the user uses this searching method: (the ontology knowledge storehouse should change the fuzzy ontology knowledge base into)

301, the user selects the inquiry mode based on keyword;

302, the user selects the concept queries mode;

303, the user selects patterned way customization constraint fuzzy query mode;

304, user's inputted search keyword;

305, notion that the user selects or input is retrieved;

306, the user is by the complete constraint fuzzy query of patterned way customization;

307, the query requests after handling is submitted to inference engine;

308, the etendue critical word set of submitting to The reasoning results to produce arrives keyword retrieval;

309, The reasoning results is submitted to the results page generation module;

310, the record that meets keyword retrieval is submitted to the sort result module;

311, the result after the ordering submits to the results page generation module;

312, the Search Results that the user generates by the access interface visit.

The reasoning embodiment, the reasoning server node is accepted the keyword query that the user submits to from search node, the reasoning server node is at first regarded it as notion, carry out the notion reasoning, if there is this notion, then return its equal notion, sub-notion and father's notion, carry out case-based reasoning then, retrieve all examples that belong to these notions; If this notion does not exist, the reasoning server node is considered as example with it, carries out case-based reasoning, retrieves the example that equates with it.The reasoning server returns to the user with the sample result of reasoning gained, also it is submitted to traditional search engines simultaneously.If this notion does not exist yet, then directly this keyword is committed to text retrieval in knowledge base.Reasoning flow process of the present invention as shown in Figure 4.

The reasoning server is realized the reasoning in ontology knowledge storehouse by calling pellet API.

Import the pellet.jar kit of increasing income: the example ontology of definition org.semanticweb.OWL.model.OWLOntology, the example reasoner of org.mindswap.pellet.OWLapi.Reasoner.Use ontology to read in the URI of OWL:

Ontology=OntologyHelper.getOntology (URI.create (uri)) (wherein uri is the address of ontology file).Then body is loaded on inference machine: reasoner.setOntology (ontology).Then body is carried out consistency detection: reasoner.isConsistent () (returning Boolean).This process realizes by the Tableau algorithm, has generated relevant triplet sets in internal memory, calls API then tlv triple is read, for example.

Enumerate all classes: Set classSet=reasoner.getClasses ();

Enumerate all examples: Set individuaSet=reasoner.getIndividuals ();

Type determined property node according to node belongs to notion or example, specifically, if the type attribute has only OWL:Class to be class, if the type attribute comprises other classes, then is example.As for obtaining of the associated class of class, at first with parent, subclass equates that class is defined as associated class.Obtaining associated class promptly can realize by following API:

OWLCLASS?class；

Set sup=reasoner.getSuperClass (class); (obtaining parent)

Set sub=reasoner.getSubClass (class); (obtaining subclass)

Set equipment=reasoner.getEquipmentClass (class); (obtaining equal class)

Obtain the example of class:

Set inds=reasoner.getIndividuals (class); (obtaining example)

Remove outside these API, all right compatible RDQL statement of reasoner, pass through method:

ResultSet set=reasoner.excuteQuery (RDQL statement);

Can realize the OWL tlv triple is inquired about, and realize above-mentioned each method.In addition, carry out " Select? a,? b,? c where (? a,? b,? c) " can obtain all triplet sets of body, this set is write specified database, can realize the persistence of body based on this.

If the user is by the self-defined constraint of patterned way inquiry, method for customizing is as follows: at first the notion in the body is selected, all therewith the attribute of conceptual dependency be presented, for expansion.Click the attribute expanding query figure that chooses then and limit this attribute, select the codomain notion of this attribute, the rest may be inferred finishes up to the inquiry customization, sets the notion that needs inquiry at last.Graphical customization procedure as shown in Figure 5.

Below by specific embodiment plain language justice search procedure.Set up laboratory ontology file idc_onto.OWL.

It is as follows to set up segment:

<？xml?version＝″1.0″？>

<rdf:RDF

xmlns:rdf＝″http://www.w3.org/1999/02/22-rdf-syntax-ns#″

xmlns:rdfs＝″http://www.w3.org/2000/01/rdf-schema#″

xmlns:OWL＝″http://www.w3.org/2002/07/OWL#″

xmlns＝″http://www.OWL-ontologies.com/unnamed.OWL#″

xml:base＝″http://www.OWL-ontologies.com/unnamed.OWL″>

<OWL:Ontology?rdf:about＝″″/>

＜OWL:Class rdf:ID=" secondary research field " 〉

<rdfs:subClassOf>

＜OWL:Class rdf:ID=" research field "/〉

</rdfs:subClassOf>

</OWL:Class>

＜OWL:Class rdf:ID=" master " 〉

<rdfs:subClassOf>

＜OWL:Class rdf:ID=" student "/〉

</rdfs:subClassOf>

</OWL:Class>

…

<OWL:Class?rdf:ID＝″FuzzyConcept″>

<rdfs:subClassof>

<OWL:Class?rdf:about＝″#Resource″/>

</rdfs:subClassof>

<OWL:Restriction>

<OWL:onProperty>

<OWL:DatatypeProperty?rdf:about＝fuzzy:lower_degree/>

</OWL:onProperty>

<OWL:maxCardinality

rdf:datatype＝″http://www.w3.org/2001/XMLSchema#int″

>1</OWL:maxCardinality>

<OWL:onProperty>

<OWL:DatatypeProperty?rdf:about＝fuzzy:upper_degree/>

</OWL:onProperty>

<OWL:maxCardinality

rdf:datatype＝″http://www.w3.org/2001/XMLSchema#int″

>1</OWL:maxCardinality>

</OWL:Restriction>

</OWL:Class>

＜OWL:Class rdf:ID=" youth " 〉

<rdfs:subClassof>

<OWL:Class?rdf:about＝″#FuzzyConcept″/>

</rdfs:subClassof>

…

＜doctor rdf:ID=" Xiao Sun " 〉

<hasAge>27</hasAge>

The results page that the user finally obtains is to be combined by The reasoning results and result for retrieval, supposes to have set up above-mentioned laboratory body in sphere of learning, and has the corresponding instance set.

If input is based on the inquiry of keyword, as import computing machine, the reasoning server thinks that at first this speech is a notion, knows that by the reasoning of ontology knowledge storehouse computer is the equal notion of this speech, then computer and computing is returned to searcher simultaneously, the result that retrieval satisfies condition in index.

If the doctor is as conceptual retrieval in input, then reasoning server retrieves goes out father's notion, equal notion and the sub-notion of conceptual dependency therewith.This The reasoning results shows in the upper right side of the page.The server of reasoning simultaneously also need further retrieve pairing all examples of this notion of doctor, and all examples are returned to searcher as search key, and searcher is retrieved the result who satisfies condition in index.The structure of the net result page is shown in Fig. 6 a.In following example, doctor's example has comprised Xiao Tang, Xiao Sun and Xiao Yu etc., and searcher is inquired about as keyword Xiao Tang, Xiao Sun and Xiao Yu etc. in index simultaneously, and the link that satisfies condition is presented under the corresponding example.Simultaneously, provide the brief description of all examples in the lower right of the page, these essential informations derive from the ontology knowledge storehouse.

If input is the fuzzy concept for example " very young doctor " that has a mood operator, then the reasoning server is isolated fuzzy concept " youth " and mood operator " very ", still carries out reasoning as keyword according to above-mentioned step with " doctor ".Find then and return fuzzy concept " youth " and therewith conceptual dependency father's notion, equate the example collection of notion and sub-notion correspondence.Supposing to be provided with the interval threshold value of fuzzy membership is 0.2, and promptly system thinks that the degree of membership interval (is interval upper limit μ less than 0.2 example ^U＜0.2) no longer belongs to this fuzzy class.According to the corresponding t=2 of mood operator " very ", the example among the C is carried out the calculating of degree of membership, the degree of membership interval that hypothesis instance Xiao Sun is under the jurisdiction of " youth " is [0.35,0.43], then has:

μ ^U _{" very young "}, (Sun Xiaolin)=[μ ^U _{" youth "}, (Sun Xiaolin)] ²=(0.43) ²≈ 0.185

μ ^L _{" very young "}, (Sun Xiaolin)=[μ ^L _{" youth "}, (Sun Xiaolin)] ²=(0.35) ²≈ 0.123

Then example Xiao Sun be under the jurisdiction of the degree of membership of " very young " interval for [0.123,0.185] since 0.185＜0.2 can obtain example " Xiao Sun " and should not belong to " very young doctor ", then from set C, pick out this example.All examples filter among the traversal C, return the result set after the filtration and have different degree of membership intervals according to remaining example among the set C, will sort from large to small according to the comparative approach in degree of membership interval in ordering.Final page is looked like shown in the 6b.

If the user needs inquiry " research direction is that database and native place are the teacher in Wuhan, Hubei ", then can realize by graphical custom-built query mode.At first, the user selectes notion " teacher ", and utilizes pattern technology to be presented on the page this notion and further this figure is operated for the user, and as shown in Figure 7, the user can limit query concept by the chooser classification.

Show teacher's all properties, as shown in Figure 8, the user can select any attribute is retrained, and in this embodiment, can retrain " research direction " and " native place " these two data attributes.

In addition also can retrain, as shown in Figure 9 " delivering works " such object properties.After constraint is finished, need to confirm the notion of inquiry at last.The reasoning server infers the example that satisfies condition according to the constraint inquiry of submitting in knowledge base, and is shown to the user, its result's display structure identical with concept queries (also identical with the fuzzy query flow process).

The raising of recall ratio.By keyword resolution, can be its equal notion, sub-notion etc. with keyword expansion, and can handle fuzzy keyword to a certain extent, thereby improve recall ratio based on body.Comprise the fuzzy concept inquiry by concept queries, can accurately retrieve all examples that belong to the related notion collection or the example that satisfies the degree of membership threshold value by reasoning.By patterned way customization semantic search, can fully discern and keep the semantic information in the user inquiring, thereby reach the purpose that improves precision ratio.Though classical description logic can increase the semantic feature of search engine return results, but can't accurately understand and find the result for fuzzy searching request.Though 1-type fuzzy set can utilize the size of degree of membership value to handle fuzzy problem to a certain extent, make the fuzzy expression ability of 1-type fuzzy set be short of to some extent owing to use determined value to describe the limitation that ambiguity brings.Therefore will apply to search engine in conjunction with the 2-FSHOIN (D) that 2-type fuzzy set proposes with the classical description logic and can handle fuzzy problem more exactly, and owing to the powerful descriptive power of description logic can give the semantic search engine powerful semanteme support.

Through the above description of the embodiments, those skilled in the art can be well understood to the present invention and can realize by hardware, also can realize based on such understanding by the mode that software adds necessary general hardware platform, technical scheme of the present invention can embody with the form of software product, it (can be CD-ROM that this software product can be stored in a non-volatile memory medium, USB flash disk, portable hard drive etc.) in, comprise that some instructions are with so that a computer equipment (can be a personal computer, server, the perhaps network equipment etc.) carry out the described method of each embodiment of the present invention.

More than disclosed only be several specific embodiment of the present invention, still, the present invention is not limited thereto, any those skilled in the art can think variation all should fall into protection scope of the present invention.

Claims (13)

1, a kind of searching fuzzy semantics is characterized in that, may further comprise the steps:

To the keyword of the representative fuzzy concept of user input and the keyword of represent mood operator, utilize described fuzzy ontology knowledge base to carry out reasoning and calculation, and the keyword sets cooperation that reasoning and calculation produces is the querying condition after expanding;

According to the querying condition after the described expansion, in described resource index storehouse, retrieve qualified Query Result.

2, the described searching fuzzy semantics of claim 1 is characterized in that, the described fuzzy ontology knowledge base of setting up specifically comprises:

Notion in classical body is added fuzzy word, makes described notion convert fuzzy concept to, and the interval that is subordinate to of atom fuzzy word is set;

Increase the example of described fuzzy concept, and, draw or be provided with the be subordinate to interval of example described fuzzy concept according to being subordinate to of described fuzzy concept of interval calculated examples property value.

3, the described searching fuzzy semantics of claim 1, it is characterized in that, the keyword of described representative fuzzy concept to user input and the keyword of representing mood operator, utilize described fuzzy ontology knowledge base to carry out reasoning and calculation, and for the querying condition concrete steps after expanding be the keyword sets cooperation that reasoning and calculation produces:

From the fuzzy ontology knowledge base, extract fuzzy concept, and belong to the example collection A of described fuzzy concept according to the fuzzy threshold value retrieval that is provided with;

Find the notion of the keyword correspondence of removing described fuzzy concept and with father's notion of described conceptual dependency, equate notion and the corresponding example collection B of sub-notion;

Get set C=A ∩ B;

According to default degree of membership threshold value the example in the ontology knowledge storehouse is compared, will be subordinate to the interval and from example collection C, reject less than the example of threshold value;

Return example collection C from large to small in proper order according to the interval.

4, the described searching fuzzy semantics of claim 3 is characterized in that, the default degree of membership threshold value of described basis compares specifically the example in the ontology knowledge storehouse and comprises:

For being subordinate to interval [a, b] and threshold value is the example of t, if a＞t then claim interval greater than threshold value; If b＜t then claims interval less than threshold value.

5, the described searching fuzzy semantics of claim 3 is characterized in that, described according to the interval from large to small order return example collection C and specifically comprise:

For interval C ₁=[a, b], C ₂=[c, d] has $h_1=\frac{(a+b)}{2},$ $h_2=\frac{(c+d)}{2},$ If h ₁＜h ₂Then claim interval C ₁＜C ₂If instead h ₁＞h ₂Claim C ₁＞C ₂If h ₁=h ₂If then (a+b)＜(c+d) then claim C ₁＜C ₂If instead (a+b)＞(c+d) then claim C ₁＞C ₂If (a+b)=(c+d) then claim C ₁=C ₂

6, the described searching fuzzy semantics of claim 1 is characterized in that, described keyword is substituted by patterned way, utilizes described patterned way customization semantic query, and the user is provided with complicated constraint condition inquiry, realizes the user customizable query statement.

7, a kind of fuzzy semantics search system comprises semantic search node, at least one resource website and internal network, it is characterized in that, described semantic search node specifically comprises:

Ontology knowledge storehouse node is used to store fuzzy domain body, and described domain body is realized with ontology describing language OWL file;

The crawl device node is used for being responsible for the centralized intranet resources of creeping, and obtains resource content and sets up index, safeguards resource locator URL information and the setting scope of creeping;

Search node is used for the user and selects inquiry mode voluntarily;

Inference node is used for reasoning and realizes semantic search reasoning service, and returns The reasoning results and submit to traditional search engines or directly return to the user.

8, as fuzzy semantics search system as described in the claim 7, it is characterized in that described ontology knowledge storehouse node specifically comprises:

Classical body is set up the unit, is used to adopt automanual body constructing method to set up classical body;

The fuzzy concept adding device is used for adding fuzzy concept to body, and the degree of membership of atom fuzzy concept is set;

Fuzzy ontology knowledge base generation unit, be used for adding the example of fuzzy concept at described classical body, and the calculated examples property value, draw or be provided with the be subordinate to interval of example to fuzzy concept, finish the structure of fuzzy ontology, make classical ontology knowledge storehouse expand to the fuzzy ontology knowledge base.

9, as fuzzy semantics search system as described in the claim 7, it is characterized in that described crawl device node specifically comprises:

The address information acquiring unit is used for obtaining beginning resource node resource locator URL information;

The scope acquiring unit is used to obtain this website in-house network and the scope of creeping is set;

The unit set up in index, and webpage and set up index is used to creep;

Updating block is used to upgrade search node server-side index file.

10, as fuzzy semantics search system as described in the claim 7, it is characterized in that described search node specifically comprises:

The querying condition receiving element is used to receive the querying condition that the user submits to, and described querying condition is sent to inference engine;

Inference engine is used under knowledge base is supported querying condition being expanded, and realizes the query parse based on body, and the querying condition after querying condition and the parsing is sent to search utility together;

Search utility is used for indexed file and retrieves the link that meets querying condition, and the result after will sorting returns to the user.

As fuzzy semantics search system as described in the claim 10, it is characterized in that 11, described querying condition comprises keyword or graphical example.

12, as fuzzy semantics search system as described in the claim 7, it is characterized in that described inference node specifically comprises:

The set acquiring unit is used for extracting fuzzy concept from the fuzzy ontology knowledge base, and belongs to the example collection A of described fuzzy concept according to the fuzzy threshold value retrieval that is provided with; Find the notion of the keyword correspondence of removing described fuzzy concept and with father's notion of described conceptual dependency, equate notion and the corresponding example collection B of sub-notion; The common factor of getting example collection A and example collection B obtains example collection C C=A ∩ B;

Comparing unit is used for comparing according to the example of default degree of membership threshold value to the ontology knowledge storehouse, will be subordinate to the interval and reject from example collection C less than the example of threshold value;

Return the unit, be used for returning example collection C from large to small in proper order according to the interval.

13, a kind of semantic search node is characterized in that, comprising:

Ontology knowledge storehouse node is used to store fuzzy domain body, and described domain body is realized with ontology describing language OWL file;

The crawl device node is used for being responsible for the centralized intranet resources of creeping, and obtains resource content and sets up index, safeguards resource locator URL information and the setting scope of creeping;

Search node is used for the user and selects inquiry mode voluntarily;

Inference node is used for reasoning and realizes semantic search reasoning service, and returns The reasoning results and submit to traditional search engines or directly return to the user.

Images