CN103258008A - Multilevel service resource discovery method and system based on user situations - Google Patents

Abstract

The invention relates to multilevel service resource discovery method and system based on user situations. The method includes: dividing the situation into user basic information, environmental information and platform information; providing multilevel service resource discovery based on user function demands, user situation reasoning and user situation clustering while fully considering user situations. The method has the advantages that early service discovery results are screened and ranked by predefined situation reasoning meta-rules on the basis of service function matching, services suitable for the user situations can be returned first, and the service query results can be refined and optimized further; by user situation clustering, resource search space can be reduced greatly and candidate services can be located quickly; and the method has great theoretical research significance and practical significance.

Description

Multi-level Service Source discover method and system based on user context

Technical field

The present invention relates to the service discovery in the Computer Processing, refer to a kind of multi-level Service Source discover method and system based on user context particularly.

Background technology

Service discovery mainly is to based on user's services request, feeds back corresponding information and service.Traditional Web service based on UDDI finds to be based on the service coupling of keyword search, simple to operate, but because the semantic ambiguity of term, system is difficult to correct understanding user's services request, and extendability is relatively poor, can not reflect user's demand information comprehensively.Along with the appearance based on the service discovery of semanteme, realize that based on service ontology the semantic matches of service function (FQoS) effectively solves this drawback.On this basis, part scholar expands semantic matches, has studied the service discovery based on QoS of customer (QoS) demand and user context demand, becomes present main flow research direction.

The QoS of service refers to that can reflect an index set of serving all NOT-function attributes, and it is not only relevant with service itself, and with the network environment at service place close contact is arranged also.Service discovery based on QoS can be done further screening to intimate service according to user's request, thereby finds more suitably service; Situation is " all information that can be used for describing any one entity situation ", the consumer's residing " situation " that studies show that of Customer Praxiology (Context) selects the decision of product or service that material impact is arranged to it, the dynamic variation characteristic of situation makes that this influence is particularly evident, such as for the mobile subscriber recommends tourist attractions, just must consider " situation " factors such as current time, user's current location, traffic conditions, weather.Along with the development of Mobile business, at present increasing service recommendation is all implemented by portable terminal, and the ability that therefore possesses the situation sensitivity recommends accuracy just to become of crucial importance to improving service recommendation.But summarize present stage both at home and abroad for the research of service discovery method, find mainly to have the following disadvantages:

(1) the user profile utilization factor is not high.In service discovery method, current user profile is being born the role that service is filtered, but service discovery method is based on selection and the coupling that service function and quality are served mostly at present.The general consideration that also just increases customer location of current service discovery method based on user context, when the service coupling, its derivation relationship and classic method do not have essential distinction, still only utilize some inherent in the domain body more weak derivation relationships, and the grade of fit of service discovery is not high.

(2) do not take into account time efficiency and user's grade of fit.Based on the service discovery method of Semantic Web Services cluster, though improved the time efficiency that Web service is found to a certain extent, can not guarantee that the service discovery result is to user's grade of fit.By the service discovery method of reasoning from logic, though improved the grade of fit of service discovery result to the user, can not guarantee the time efficiency of service discovery.

Summary of the invention

In view of above-mentioned prior art problems, the present invention proposes a kind of multi-level Service Source discover method based on user context, this method at first is divided into situation user basic information (User Profile), environmental information (Environment) and three aspects of platform information (Platform), on the basis of comprehensive consideration user context, the multi-level Service Source discover method based on user function demand, user context reasoning and user context cluster has been proposed.

Realize that the technical scheme that the object of the invention adopts is: a kind of multi-level Service Source discover method based on user context comprises:

A: based on the service discovery of user function demand, namely carry out concept matching, structure matching and text matches between the FQoS of the Service Source that provides by Service Source and supplier to user's request, realize finding based on the Service Source of FQoS grade of fit;

B: based on the service discovery of user context reasoning, namely according to the incidence relation between user context and the Service Source, between service ontology model and user context ontology model, set up reasoning meta-rule, comprise filterableness rule, user preference rule, optimize selective rule, the candidate service collection is further optimized;

C: based on the service discovery of user context cluster, namely have characteristics of big similarity according to the similar user-selected service of situation, namely the historical user context property value according to service carries out cluster to the user, the frequency of utilization of all services in the class of statistics active user place, inverted order is arranged, and the service priority that frequency of utilization is high returns to the user.

In addition, the invention still further relates to a kind of multi-level Service Source based on user context and find system, it is characterized in that, comprising:

Service discovery module based on the user function demand, carry out concept matching, structure matching and text matches between the FQoS of the Service Source that provides by Service Source and supplier to user's request, realization is found based on the Service Source of FQoS grade of fit, obtain initial service Candidate Set, represent with CS1;

Service discovery module based on the user context reasoning, according to the incidence relation between user context and the Service Source, between service ontology model and user context ontology model, set up reasoning meta-rule, comprise filterableness rule, user preference rule, optimize selective rule, CS1 is optimized to the candidate service collection, obtains serving Candidate Set CS2;

Service discovery module based on the user context cluster, similar user-selected service has the characteristics of big similarity based on situation, historical user context property value according to service Candidate Set CS2 carries out cluster to the user, the frequency of utilization of all services in the class of statistics active user place, inverted order is arranged, the services set CS3 that frequency of utilization is high preferentially returns to the user.

The inventive method is on the basis of service function coupling, by predefined situation reasoning meta-rule, service discovery result to early stage screens and sorts, the service priority of the user context that suits can be returned, realize further refining and optimizing of service-seeking result, and by user context cluster, the search volume that can greatly reduce resource, realize the quick location of candidate service, have very strong theoretical significance and realistic meaning.

Description of drawings

Fig. 1 is the multi-level Service Source discover method synoptic diagram that the present invention is based on user context;

Fig. 2 is the service discovery synoptic diagram based on the user context cluster.

Embodiment

The present invention is described in further detail below in conjunction with the drawings and specific embodiments.

As shown in Figure 1, the multi-level Service Source discovery system that the present invention is based on user context comprises 3 functional modules, be respectively: A, based on the service discovery module of user function demand, carry out concept matching, structure matching and text matches between the FQoS of the Service Source that provides by Service Source and supplier to user's request, realized finding based on the Service Source of FQoS grade of fit, obtained candidate service collection CS1; B, based on the service discovery module of user context reasoning, according to the incidence relation between user context and the Service Source, between service ontology model and user context ontology model, set up reasoning meta-rule, comprise filterableness rule, user preference rule, optimize selective rule, CS1 further optimizes to the candidate service collection, obtains candidate service collection CS2; C, service discovery module based on the user context cluster, the characteristics that have big similarity according to the similar user-selected service of situation, utilization is based on the clustering method of fuzzy price relation, historical user context property value according to service Candidate Set CS2 carries out cluster to the user, the frequency of utilization of all services in the class of statistics active user place, set up the service inverted index table of active user place class, the service priority that frequency of usage is high is returned, can obtain serving Candidate Set CS3, this step has effectively been dwindled the scope of service search, has improved the efficient of service search.

Present embodiment needs the hotel reservation service to describe the present invention in detail with a user.The user context information that this reservation service can be drawn into comprises: (1) user's characteristic information.Be divided into user basic information and preference information, essential information mainly comprises user's name, sex, age, occupation, place, purpose and companion's information, as shown in table 1, because need ID (identity number) card No. during the online ordering ticket, therefore can infer user companion's name, sex and the age; Preference information mainly is the information that clearly provides of user and by excavating the information that the historical consume record obtains, and with the user preference information in hotel is example, and is as shown in table 2.(2) the residing environmental information of user.As shown in table 3.(3) user platform feature.Comprise user equipment information and networking mode.Subscriber equipment is notebook, and networking mode is that switch inserts.

Name

Age

Sex

Occupation

The place

Purpose

The companion

Zhang San

40

The man

The project manager

Wuhan

Spend a holiday

Father and mother, wife, child

Table 1

Star

Price

Traffic

The position

Check out

Breakfast

The broadband

Three-star

Medium

Convenient

The urban district

Afternoon

Provide

Provide

Table 2

Time	Weather	Temperature	Season	The network bandwidth	Internet security
						12.8	Sunny	8℃	Winter	2M	High

Table 3

The specific implementation process of present embodiment is:

A. based on the service discovery of user function demand, concrete operations and employing algorithmic notation are as follows:

The concept similarity of A1:FQoS calculates.Calculating is such as the similarity of the FQoS of atomic tag types such as " classification of service ".Adopt a kind ofly based on string matching algorithm, based on the comprehensive semantic similarity algorithm that WordNet and information theory combine, be specially:

A11: the concept semantic similarity based on string matching calculates.The basic concept of this method is to realize similarity coupling between the element by the editing distance of calculating character string.The character string that editing distance refers to a concept converts the minimal action number of times that needs in the process of another concept place character string to, and these operations mainly contain replacement, insertion and deletion etc.The Maedche ﹠amp that quotes; The similarity algorithm that Staab proposes is used sim ₁(C ₁, C ₂) represent the semantic similarity based on string matching between concept C1 and the C2, this algorithm can be expressed as so:

${\mathrm{sim}}_1(c_1,c_2)=\max (0,\frac{\min \left(\right|C_1|,|C_2\left|\right)-D(C_1,C_2)}{\min \left(\right|C_1|,|C_2\left|\right)})$

Wherein, | C| represents the shared character length of concept C, D (C ₁, C ₂) editing distance between two concepts of expression.In computation process, basic edit operation (comprising insertion, deletion and the replacement of character) the performed cost weight between the character string all is 1.

A12: the concept semantic similarity based on WordNet calculates.WordNet be one one based on the English semantic dictionary of cognitive linguistics.It in an organized way gathers noun, adjective, verb and adverbial word and forms a plurality of synonym set, and interconnects by dissimilar semantic relation mutual restriction between each synonym set, has constituted the network of a vocabulary.Semantic relation between each synonym set can be that part, hyponymy, synonymy or antonymy etc. are quoted.The semantic similarity algorithm of Wu-Palmer is used sim ₂(C ₁, C ₂) represent the semantic similarity based on WordNet between concept C1 and the C2, this algorithm can be expressed as so:

${\mathrm{sim}}_2(C_1,C_2)=\frac{2\×\mathrm{depth}\left(\mathrm{lso}(C_1,C_2)\right)}{\mathrm{depth}\left(C_1\right)+\mathrm{depth}\left(C_2\right)}$

Wherein, depth (C ₁) and depth (C ₂) refer to concept C1 and the degree of depth of concept C2 in the WordNet semantic tree, lso (C ₁, C ₂) what describe is nearest common ancestor's concept of concept C1 and concept C2.

A13: the concept semantic similarity based on information theory calculates.During semantic similarity calculates the similarity of two concepts with they between to share the degree of information relevant.Use sim ₃(C ₁, C ₂) represent the semantic similarity based on information theory between concept C1 and the C2, this algorithm can be expressed as so:

${\mathrm{sim}}_3(C_1,C_2)=\frac{2\×\mathrm{share}(C_1,C_2)}{\mathrm{IC}\left(C_1\right)+\mathrm{IC}\left(C_2\right)}$

Wherein, IC (C) refers to the information content of concept C, may be defined as: IC (C)=-log (Prob (C)), Prob (C)=Freq (C)/N wherein, N represents the summation of all concepts in the body corpus, Freq(C) refers to the frequency of concept C in the body, the number of times that occurs in body for concept C and its all child nodes, according to the information theory of Ross, a concept C in the body may be defined as: Freq (C)=Σ { occur (C _i) | C ∈ Ancestors (C _i).Share (C ₁, C ₂) referring to that concept C1 and C2 share information, the semantic similarity between two concepts is relevant with their degree of shared information, so their shared information can represent with the information content of the ancestor node that the information similarity is the highest between two concepts, i.e. share (C ₁, C ₂)=max{IC (A) | a ∈ sub (C ₁, C ₂), sub (C wherein ₁, C ₂) refer to the concept that comprises C1 and C2 simultaneously.

Comprehensive semantic similarity between A14: concept C1 and the C2 calculates.With sim (C ₁, C ₂) representing the comprehensive semantic similarity between concept C1 and the C2, can be expressed as:

sim(C ₁,C ₂)=w ₁sim ₁(C ₁,C ₂)+w ₂sim ₂(C ₁,C ₂)+w ₃sim ₃(C ₁,C ₂)

Wherein, sim ₁(C ₁, C ₂), sim ₂(C ₁, C ₂) and sim ₃(C ₁, C ₂) represent the similarity result based on character string, WordNet and information theory coupling respectively.w ₁, w ₂And w ₃Being divided into is the weight coefficient of three kinds of different concept matching algorithms, represents every conception of species similarity algorithm shared influence degree in aggregate concept semantic similarity algorithm, and w ₁+ w ₂+ w ₃=1.

The structural similarity of A2:FQoS calculates.Structure matching need be analyzed more complicated tree structure (for example XML dom tree shape structure) on the basis of atomic tag.Quote the method for Sun Xia and Cheng Hongbin proposition the XML tree structure is carried out similarity calculating.This method at first utilizes the hierarchical structure relation of XML document tree to extract corresponding file characteristics information, according to the hierarchical model after the modeling node element of different levels is carried out different weight allocation then, make high-level node element weight greater than the node element weight of low level, utilize calculating formula of similarity that two XML document are mated at last.With sim (X ₁, X ₂) structural similarity of expression XML document X1 and X2, this algorithm can be expressed as so:

L wherein ₁And L ₂The number of plies that is worth document X1 and document X2 respectively, w _1iAnd w _2jThe weight of representing j layer node among the weight of i layer node among the document X1 and the document X2 respectively, s _1iIdentical nodal point number among i layer node and the document X2 among the expression document X1, s _2jThen represent among the document X2 identical nodal point number in the j layer node and document X1, s _1mAnd s _2nThe node sum of representing n layer among the node sum of m layer among the document X1 and the document X2 respectively.

The Text similarity computing of A3:FQoS.Text matches is used for handling the FQoS such as the text data type of " note " etc., can be achieved by the static text digging technology, mainly is by calculating the metric that similarity between two or more texts obtains its matching degree.Can carry out the similarity coupling to the FQoS of text by the vector space model in the statistical method (VSM).In VSM, text D can be expressed as D={t with the characteristic item set ₁, t ₂T _k... t _n, t wherein _kRefer to k characteristic item of text, n refers to the number of characteristic item.w _kFinger is according to t _kSignificance level in text and the weight of giving, and D={t ₁, w ₁, t ₂, w ₂T _k, w _k... t _n, w _nJust can constitute the text vector space of n dimension.

With sim (D ₁, D ₂) structural similarity of expression text D1 and D2, this algorithm can be expressed as: $\mathrm{sim}(D_1,D_2)=\cos \mathrm{\θ}=\frac{\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{w}_{1k}\×w_{2k}}{\sqrt{\left(\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{w_{1k}}^2\right)\×\left(\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{w_{2k}}^2\right)}}$

W wherein _1kRefer to the weighted value of characteristic item k in text D1; w _2kRefer to the weighted value of characteristic item k in text D2.They all can calculate by the TF-IDF formula:

$w_{\mathrm{ik}}=\frac{{\mathrm{tf}}_{\mathrm{ik}}\log (N/n_k+0.5)}{\sqrt{\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{\left({\mathrm{tf}}_{\mathrm{ik}}\right)}^2\×{\left[\log (N/n_k+0.5)\right]}^2}}$

W wherein _IkRefer to the weighted value of characteristic item k in text Di; Tf _IkRefer to the number of times that characteristic item k occurs in text Di, just word frequency; Log (N/n _k+ 0.5) describe be feature word k arrange the text frequency, N is the number (being 2 here) of matched text, n _kRefer to the amount of text that contains characteristic item k.

A4: comprehensive FQoS grade of fit is calculated the (FQoS with sim ₁, FQoS ₂) expression, namely have

$\mathrm{sim}({\mathrm{FQoS}}_1,{\mathrm{FQoS}}_2)=\frac{\mathrm{sim}(C_1,C_2)+\mathrm{sim}(X_1,X_2)+\mathrm{sim}(D_1,D_2)}{3}$

A5: will calculate the FQoS grade of fit and the user preset threshold value compares, and be designated as service Candidate Set CS1 greater than the Service Source of predetermined threshold value.

B. based on the service discovery of user context reasoning.According to above-mentioned user context information, available part inference rule is as follows.

B1: definition filterableness rule, according to case assumed condition, available part inference rule is expressed as follows:

Rule1：R(maximum)→filterS。Wherein, R=Ins_Load ().Certain monitor of server end is responsible for serving the perception of real-time load, and the service of load saturation is filtered out.

B2: definition user preference rule comprises:

Rule2： $isPreferred:Q^R(3\text{\&\#160;}star)\→isChosen(S^n)\mathrm{\Λ}\underset{i=1}{\overset{n}{\mathrm{\Λ}}}(Q(x\text{\&\#160;}star)\mathrm{\Λ}(isPreferred:$ $Q^R=Q)\mathrm{\Λ}(x\text{\&\#160;}equals\text{\&\#160;}3)$ 。Q wherein ^R, Q represents the star in hotel.

Rule3 $isPreferred:Q^R(\text{convenient})\→isChosen(S^n)\mathrm{\Λ}\underset{i=1}{\overset{n}{\mathrm{\Λ}}}(Q(x\text{\&\#160;})\mathrm{\Λ}(isPreferred:$ $Q^R=Q)\mathrm{\Λ}(x\text{\&\#160;}is\text{\&\#160;}convenient)$ 。Q wherein ^R, Q represents the traffic around the hotel.

Rule4： $isPreferred:Q^R(downtown)\→isChosen(S^n)\mathrm{\Λ}\underset{i=1}{\overset{n}{\mathrm{\Λ}}}(Q(x\text{\&\#160;})\mathrm{\Λ}(isPreferred:$ $Q^R=Q)\mathrm{\Λ}(x\text{\&\#160;}is\text{\&\#160;}downtown)$ 。Wherein Expression residing position, hotel.

Rule5： $isPreferred:Q^R(afternoon)\→isChosen(S^n)\mathrm{\Λ}\underset{i=1}{\overset{n}{\mathrm{\Λ}}}(Q(x\text{\&\#160;})\mathrm{\Λ}(isPreferred:$ $Q^R=Q)\mathrm{\Λ}(x\text{\&\#160;}is\text{\&\#160;}afternoon)$ 。Wherein $Q^R=Q)\mathrm{\Λ}(x\text{\&\#160;}is\text{\&\#160;}afternoon)$ The check-out time of expression hotel regulation.

B3: selective rule is optimized in definition, comprising:

Rule6： $R(familiy)\→isChosen(S^n)\mathrm{\Λ}\underset{i=1}{\overset{n}{\mathrm{\Λ}}}(Q(\text{apartotel}))$ ， $R=user\_companion$ ， $Q=type()$ 。If when expression user companion is the household, preferentially select hotel apartment.

B4: the grade precedence between the inference rule is set.According to the priority reasoning successively of inference rule, in above-mentioned 6 rules, user-defined priority orders is during deduce machine work:

, at first deduce machine can filter out the flight service that the departure time is evening, then, according to the user preference of hotel's grade, position, check-out time and traffic is selected hotel service, then, and according to companion's type selecting hotel type.

B5: whether check between the inference rule contradiction.Noncontradictory on inspection forwards step B6 to.

B6: through above-mentioned service discovery based on the user context reasoning, obtain the candidate service collection CS2 of hotel reservation task, be S _Hotel={ S1, S2, S10}.

C. based on the service discovery of user context cluster, realization flow as shown in Figure 2, concrete operations are expressed as follows:

C1: create customer data base, the contextual information of the historical user of all services among active user's contextual information and the candidate service collection CS2 is joined in the customer data base simultaneously.In order to reduce the complexity of situation cluster, user context is mainly considered these 5 factors of user's annual income, companion's number, companion's type, purpose and time span, wherein, Ui (i=1,2 ..., 9) and represent historical user, U10 represents the active user, and is as shown in table 4.

Table 4

C2: set up the inverted index table of user's services selection according to table 2, as shown in table 5.

The user	The service frequency of utilization
		UI	S1[1]
U2	S2[1]
		U3	S1[1]，S2[1]
U4	S3[1]
		U5	S3[1]，S4[1]，S5[1]
U6	S5[2]，S6[2]，S3[1]，S4[1]，?S7[1]
		U7	S5[2]，S6[2]，S8[2]，S3[1]，?S7[1]，?S9[1]
U8	S7[1]，S8[1]，S9[1]，S10[1]
		U9	S8[1]，S10[1]

Table 5

C3: the user carries out cluster according to situation in the his-and-hers watches 1.Cluster result as shown in Figure 3.For property value is quantized.In the purpose attribute, spend a holiday=0, go on business=1.In companion's attribute, be designated as 0 alone, friend=1, colleague=2, household=3.Obtain active user place class, for (U6, U7, U10).

C4: the class by the active user place is screened, thereby obtains the inverted index table of active user place class, and is as shown in table 6.The frequency of utilization of all services has S3[2 in the class of statistics active user place], S4[2], S5[4] and, S6[4], S7[2] and, S8[2], S9[1].

The user	The service frequency of utilization
		U6	S5[2]，S6[2]，S3[1]，S4[1]，?S7[1]
U7	S5[2]，S6[2]，S8[2]，S3[1]，?S7[1]，?S9[1]
		U10

Table 6

C5: the frequency of utilization according to service is carried out descending sort, selects the service of rank preceding 5, obtains the candidate service collection CS3hotel=(S5 of hotel service, S6, and S3, S7 S8), returns the candidate service collection CS3hotel of hotel service.

Claims (8)

1. multi-level Service Source discover method based on user context is characterized in that comprising:

A: based on the service discovery of user function demand, namely carry out concept matching, structure matching and text matches between the FQoS of the Service Source that at first provides by Service Source and supplier to user's request, realization is found based on the Service Source of FQoS grade of fit, obtain preliminary candidate service collection, be designated as CS1;

B: based on the service discovery of user context reasoning, then according to the incidence relation between user context and the Service Source, between service ontology model and user context ontology model, set up reasoning meta-rule, comprise filterableness rule, user preference rule, optimize selective rule, realization obtains serving Candidate Set CS2 to the optimization of candidate service collection CS1;

C: based on the service discovery of user context cluster, selected service has the characteristics of big similarity based on the situation similar users, historical user context property value according to service Candidate Set CS2 carries out cluster to the user, the frequency of utilization of all services in the class of statistics active user place, inverted order is arranged, the services set CS3 that frequency of utilization is high preferentially returns to the user.

2. the multi-level Service Source discover method based on user context according to claim 1 is characterized in that described steps A specifically comprises based on the service discovery of user function demand:

A1: the concept similarity that calculates FQoS;

A2: the structural similarity that calculates FQoS;

A3: the text similarity that calculates FQoS;

A4: calculate comprehensive FQoS grade of fit;

A5: will calculate the FQoS grade of fit and the user preset threshold value compares, and be designated as service Candidate Set CS1 greater than the Service Source of predetermined threshold value.

3. the multi-level Service Source discover method based on user context according to claim 2 is characterized in that, the concept similarity of described calculating FQoS comprises:

The comprehensive semantic similarity algorithm that employing combines based on string matching algorithm, based on WordNet and information theory, utilize the semantic similarity algorithm to solve the ambiguous problem of semantic differentiation that exists in the string matching algorithm, utilize string matching algorithm to reduce the error that the compound word coupling is brought simultaneously, with sim (C ₁, C ₂) represent the comprehensive semantic similarity between concept C1 and the C2, be expressed as:

sim(C ₁,C ₂)=w ₁sim ₁(C ₁,C ₂)+w ₂sim ₂(C ₁,C ₂)+w ₃sim ₃(C ₁,C ₂)

Wherein, sim ₁(C ₁, C ₂), sim ₂(C ₁, C ₂) and sim ₃(C ₁, C ₂) represent the similarity result based on character string, WordNet and information theory coupling respectively; w ₁, w ₂And w ₃Being divided into is the weight coefficient of three kinds of different concept matching algorithms, represents every conception of species similarity algorithm shared influence degree in aggregate concept semantic similarity algorithm, and w ₁+ w ₂+ w ₃=1.

4. according to the described multi-level Service Source discover method based on user context of claim 2, it is characterized in that the comprehensive FQoS grade of fit of described calculating comprises:

Concept similarity, structural similarity and the text similarity of the Service Source of user's request and the FQoS between service provider's service supplied resource averaged to be obtained, with sim (FQoS ₁, FQoS ₂) expression FQoS grade of fit, namely have

$sim\left(FQ\text{o}{S}_{\text{1}}{\text{,FQoS}}_{\text{2}}\right)\text{=}\frac{sim\left(C_1,C_2\right)+sim\left(X_1,X_2\right)+sim\left(D_1,D_2\right)}{3}$

Sim (FQoS wherein ₁, FQoS ₂) ∈ [0,1], C1, X1, D1 represent the corresponding concept type of the Service Source of user's request, structure type, text FQoS respectively; C2, X2, D2 then are respectively concept type, structure type, the text FQoS of service provider's service supplied resource correspondence; Sim (C ₁, C ₂) similarity between expression concept C1 and the C2; Sim (X ₁, X ₂) similarity between expression XML document X1 and the X2; Sim (D ₁, D ₂) similarity between expression destructuring document D 1 and the D2.

5. according to the described multi-level Service Source discover method based on user context of claim 1, it is characterized in that described step B specifically comprises based on the service discovery of user context reasoning:

B1: the filterableness rule is set, when certain attribute of user context or service equals certain value, directly should serves filtration, the meta-rule of described filterableness rule is expressed as: R (v ₀) → filerS,

V wherein ₀Refer to certain property value of user context or service, R is the predicate relational symbol, represents that certain situation or attributive concept have certain relation, and filter WS represents that service directly is filtered, and will not appear in the result set that returns to the user;

B2: the user preference rule is set, and the services set that property value is met user preference returns to the user, and user preference meta-rule is expressed as:

$isPreferred:Q^R(preferred\text{\&\#160;}vs)\→isChosen(S^n)\mathrm{\Λ}\underset{i=1}{\overset{n}{\mathrm{\Λ}}}(Q(vsi)\mathrm{\Λ}(isPreferred:Q^R=Q)\mathrm{\Λ}$

$(preferred\text{\&\#160;}vs\text{\&\#160;}metBy\text{\&\#160;}vsi))$

V wherein _sRefer to that the user is to the preference value of Service Properties i, v _SiThe value that refers to Service Properties i, isChosen (S ⁿ) expression n service selected (n ∈ N); IsPreferred:Q ^RBe the predicate relational symbol, certain situation property value of expression user preference service has certain condition; Q is the predicate relational symbol, and certain situation property value of the service that expression is returned has certain condition; MetBy represents that certain service situation condition of user preference is satisfied by the situation condition of service that chosen correspondence;

B3: the optimization selective rule is set, when user's sight attribute was certain certain value, certain property value of the service that meets the demands also will be certain certain value, perhaps when some attribute of service satisfies a certain condition, this service of preferential selection, the meta-rule that described optimization is selected is expressed as:

$R(vu)\→isChosen(S^n)\mathrm{\Λ}\underset{i=1}{\overset{n}{\mathrm{\Λ}}}(Q1(vsi)\mathrm{\Λ}Q2(vsi)\mathrm{\Λ}\mathrm{...}\mathrm{\Λ}{Q}_j(vsi)),\text{\&\#160;}n,j\∈N$

Namely when certain situation attribute of user is a certain value, will return one group of relevant services set, wherein, some situation attribute of each service should have particular value;

Perhaps, $Q(v_{si})\→isChosen(S^n)\mathrm{\Λ}\underset{i=1}{\overset{n}{\mathrm{\Λ}}}(Q1(vsi)\mathrm{\Λ}Q2(vsi)\mathrm{\Λ}\mathrm{...}\mathrm{\Λ}{Q}_j(vsi)),\text{\&\#160;}n,j\∈N$

Namely when some property value of service satisfies certain condition, preferentially select this service;

V wherein _uCertain the situation property value that refers to the user; v _SiThe value that refers to Service Properties i; IsChosen (S ⁿ) expression n service selected (n ∈ N); R, Q _jBe the predicate relational symbol; R represents that the user context property value has certain condition; Q _jJ property value of expression service has certain condition (j ∈ N);

B4: set priority according to filterableness rule, user preference rule, optimization selective rule precedence;

B5: whether check between the inference rule contradiction, if contradiction occurs in user preference rule and the optimization selective rule, then according to regular grade, deletion and the afoul optimization selective rule of user preference rule, rule is resequenced, new rule set after being optimized if do not have, forwards step B6 to;

B6: if rule-based reasoning is to carry out according to grade precedence, and noncontradictory between the rule, then finish based on the service discovery of user context reasoning, obtain serving Candidate Set CS2.

6. according to the described multi-level Service Source discover method based on user context of claim 5, it is characterized in that, comprise based on the service discovery of user context cluster:

C1: join in the customer data base during with the historical user profile of all services among the candidate services set CS2 and active user's information of same;

C2: create user's inverted index table according to customer data base, comprise user and service frequency of utilization;

C3: according to the record in the customer data base, the user is carried out cluster;

C4: according to the frequency of utilization of all services in the class of user's inverted index table statistics active user place;

C5: by the descending sort of service frequency of utilization, obtain candidate services set CS3, and the service result collection is returned to the user.

7. according to the described multi-level Service Source discover method based on user context of claim 1, it is characterized in that described step C3 specifically comprises based on the cluster operation of user context:

C31: user context information is expressed as d with vector form _i={ w _I1w _I2..., w _In, n the situation property value of expression user i adopts the cosine similarity algorithm to come compute vector d _iWith d _jBetween similarity, this algorithmic notation is:

$sim(d_i,d_j)=\frac{{\displaystyle \underset{k=1}{\overset{n}{\∑}}{w}_k(d_i)\×w_k(d_j)}}{\sqrt{({\displaystyle \underset{k=1}{\overset{n}{\∑}}{w}_k^2(d_i)})({\displaystyle \underset{k=1}{\overset{n}{\∑}}{w}_k^2(d_j)})}}$ ；

C32: use the fuzzy equivalence relation of constructing context information from multiplication, make the sight similarity relation have transitivity, for situation similarity relation S, with its involution

, if S ²≠ S continues involution, until satisfying condition

Till; At this moment, Be exactly a fuzzy equivalence relation, establish s _IjBe the element of the capable j row of i among the S, r _IjBe S ²In the element of the capable j of i row, involution is calculated by following formula: , finally obtain fuzzy equivalence relation matrix R, wherein R (d _i, d _j)=r _Ij

C33: with r _IjCompare with predetermined threshold value λ (0≤λ≤1), if r _IjDuring＞λ, then represent d _iWith d _jBe of equal value, to be classified as a class; By the value of control threshold value λ, the final many granularity division that realize contextual information in the opinion vector space.

8. the multi-level Service Source based on user context is found system, it is characterized in that, comprising:

Service discovery module based on the user function demand, carry out concept matching, structure matching and text matches between the FQoS for the Service Source that provides by Service Source and supplier to user's request, realization is found based on the Service Source of FQoS grade of fit, is obtained initial service Candidate Set CS1;

Service discovery module based on the user context reasoning, be used for according to the incidence relation between user context and the Service Source, between service ontology model and user context ontology model, set up reasoning meta-rule, comprise filterableness rule, user preference rule, optimize selective rule, CS1 is optimized to the candidate service collection, obtains serving Candidate Set CS2;

Service discovery module based on the user context cluster, for similar user-selected service has the characteristics of similarity greatly based on situation, historical user context property value according to service Candidate Set CS2 carries out cluster to the user, the frequency of utilization of all services in the class of statistics active user place, inverted order is arranged, the services set CS3 that frequency of utilization is high preferentially returns to the user.

Images