CN103198114B - A kind of WEB service matching process based on overlapping directed graph - Google Patents

Abstract

The invention provides a kind of WEB service matching process based on overlapping directed graph, relate to computerized information analysis and data processing field, by the description file (WSDL) of substantial amounts of WEB service node is loaded into server end internal memory, form magnanimity WEB and describe source data, in order to match the service specified, establish room seed Matching Model, and build overlapping directed graph on this basis, by connective and overlap several two key indexs, seed is evaluated, find optimum seed, it is finally completed the coupling of WEB service, it it is a kind of WEB service method that fast searching is suitable。The present invention can be widely used in WEB service node to be existed in a large number, WEB service is found and requires in higher environment, in the environment such as Enterprise SOA (SOA), cloud computing。

Description

A kind of WEB service matching process based on overlapping directed graph

Technical field

The invention belongs to computerized information analysis and data processing field, be specifically related to a kind of WEB service matching process based on overlapping directed graph。

Background technology

Web service (WebService) is based on a kind of service of XML and HTTP, each web services is a corresponding WSDL(WebServicesDescriptionLanguage) file is described。WSDL describe file be one for describing Web service and the XML language file how to communicate with Web service of client being described。

Each WSDL describe file encapsulate some be available for calling interface and illustrate document, but when WSDL describe file exist in a large number time, how to quickly find the service interface of needs, be need solve a problem。

1.WEB service describing

Service-oriented application (Service-OrientedArchitecture, SOA) brings a kind of new Integrated, can construct the framework of service-centric flexibly according to it。SOA is by the characteristic of its loose coupling, it is possible to existing for enterprise application is issued as service, and service has the interface independent of hardware platform, operating system and programming language。Therefore, can by the communication realized alternately between system of the good service of this definition between different system, it is possible not only to realize the integrated collaborative work with heterogeneous service systems of mass data, can also add according to modular mode and newly service or update existing service, to solve new business demand。

The core of SOA is exactly WEB service, and it outwardly exposes one can by the Web application programming interfaces (ApplicationProgramInterface, API) called, it is possible to call this application program by the method for programming by Web。Web service platform is a set of standard, it defines how application program realizes interoperability on Web。You with any language that you like, can write Web service on any platform that you like, as long as we can pass through Web service standard, these services are inquired about and are accessed。

Web service platform needs a set of agreement to realize the establishment of distributed application program。Any platform has its data presentation technique and type system。Realizing interoperability, web service platform must provide for standard set type system, for linking up the dissimilar system in different platform, programming language and component model。In traditional distributed system, provide certain methods based on the platform of interface (interface) and describe interface, method and parameter (the IDL language as in COM and COBAR)。Same, web service platform also must provide for a kind of standard to describe Web service, allows client can obtain enough information to call this Web service。

The WSDL how key words from client existed in a large number with service end in network describes file and carries out mating and find the WSDL that these key wordses mate most to describe file thus realizing suitable web services is the Internet information search techniques field problem demanding prompt solution。

2.WEB Service Matching

The quantity of the WEB service on network may be very big, how to search out the WEB service of coupling according to keyword or essential information, is a highly important problem。The WSDL of WEB service is described file as data source, it would be desirable to the keyword of coupling and essential information, as fundamental, find the process of suitable WEB service, i.e. the coupling of WEB service。

In order to quickly find suitable WEB service (namely WSDL describes file) when mass data source, present invention introduces the information local search approach based on room seed, complete the coupling of WEB service。This method is based primarily upon prior art " sequence alignment method " and matching sequence is estimated。

The definition of sequence alignment is to be mated between two by two sequences character。Upper at matching process, in fact it could happen that following three kinds of situations: (1) character instead of another character, namely there occurs that sudden change produces character mismatch；(2) one or more character is inserted；(3) one or more character is deleted。Therefore, in sequence alignment process, be not simple character one_to_one corresponding, except coupling and mismatch, also can introduce in the sequence room (representing by symbol "-") reflect the 2nd, 3 kind of change。

Such as two sequence A ATCTATA and AAGATA, shown below is wherein 3 kinds of sequence alignment situations:

AATCTATAAATCTATAAATCTATA

AAG-AT-AAA-G-ATAAA--GATA

In order to sequence alignment result is quantified, introduce mark mechanism and sequence alignment is given a mark, to obtain the sequence alignment of optimum。After introducing room, according to different comparison situations, it is possible to obtain sequence alignment score。Marking its definition of rule is as follows:

Score(S₁,S₂) represent two sequence S₁、S₂Matching score, i is natural number certain position for representing in sequence, n represent longer sequence max number of characters, described character can also be able to be English character for Chinese character。

Wherein k_i>=0 (1≤i≤3) represent the situation in coupling, mismatch and room respectively。From marking formula it can be seen that coupling will obtain a positive mark, and mismatch or room will obtain 0 point or negative point。Owing to possible sequence alignment situation is ever-changing, different comparisons is likely to be obtained identical mark。

WEB service technology is widely used in the modern information technologies such as Distributed Calculation, WebService, SOA, in the environment that WEB service magnanimity exists, the WEB service of coupling how is quickly found out according to keyword, it it is the important means of modern WEB framework development, but the method inefficiency of traditional keyword search, it is impossible to adapt to the situation of magnanimity WEB information inquiry。

Summary of the invention

In order to overcome the drawback of tradition WEB service matching process, the present invention devises a kind of WEB service matching process based on overlapping directed graph, adopt the matching technique based on room seed and overlapping Directed Graph Model, WEB service is mated, is greatly improved WEB service rate matched and treatment effeciency。

The technical scheme is that a kind of WEB service matching process based on overlapping directed graph, comprise the following steps:

Step 1: describe file and the wsdl document of WEB service are loaded into the internal memory of server end by network, form WEB and describe source data；

Step 2: set up room seed Matching Model, describes WEB service source data and is described with the form of room seed with target service data, obtain room seed the matching frame；

Step 3: build the overlapping directed graph towards different rooms seed, calculates its connective and overlapping number for each overlapping directed graph；

Step 4: find optimum room seed with maximum connectedness and minimum overlapping number for standard, and describe optimum room seed by overlapping directed graph weight ODW；

Step 5: with target WEB service, optimum room seed is described source data and mates, obtains the WEA Service Matching result of optimum, finds the WEB service node of needs。

As preferably: step 1 method particularly includes: adopt HTML (Hypertext Markup Language) to connect long-range wsdl document address in need, WEB service describes file resolved by document object tree-model, it resolves form is hypertext tag format, for the document object tree resolved, all nodes comprising<wsdl:documentation>are loaded into the internal memory of server end。

As preferably: the room seed described in step 2, it is defined as:

One room seed S is defined on character set Α={ 1, * }, and specifies to start with 1, the 1 fixed model string terminated；Wherein 1 represents coupling, and * is an asterisk wildcard, represents in this position it can is 1 coupling or 0 mismatch；S is expressed as:

$S={\left(1\right)}^{i_1}{(*)}^{k_1}{\left(1\right)}^{i_2}{(*)}^{k_2}...{\left(1\right)}^{i_{n-1}}{(*)}^{k_{n-1}}{\left(1\right)}^{i_n}$

Wherein (1)^j(*)^kRepresent continuous j 1 and continuous k *；

The length (length) of room seed is in S all 1 and the number of *, is expressed as | S |:

$\mathrm{length}\left(S\right)=\left|S\right|=\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{i}_j+\underset{j=1}{\overset{n-1}{\mathrm{\&Sigma;}}}{k}_j$

The weight w of room seed is defined as the number of all 1:

$\mathrm{weight}\left(S\right)=w=\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{i}_j$

The pattern of room seed is that * all in S replace with the expression string of 1 or 0 formation, always has 2^|s|-wIndividual different pattern。

As preferably: step 3 comprises the following steps:

A) set up structure directed graph: top-down node be according to seed corresponding element from left to right 1 or 0, wherein 1 representation node 1, * represents two nodes: 1 and 0, and every directed edge indicates similarity p or q=1-p；

B) markers step: in order from the sequence number 1 each node of beginning label from top node to bottom, be put into after nodal value and bracket with bracket, then deletes labelling p or q, is marked on every directed edge by head node ID meanwhile；

C) overlep steps: overlep steps is the process of a series of iteration, successively using every node layer value be 1 node as start node structural texture directed graph, next step is from second layer node intermediate value 1 node, by that analogy, eventually settle at L-1 layer, after overlep steps, obtain the overlapping Directed Graph Model of room seed。

As preferably: the connectedness of overlapping directed graph is defined as:

There are two limit e having common node₁And e₂, i.e. e₁End node be e₂Start node, if B₁It is e₁The set of labelling, B₂It is e₂The set of labelling, claims B₁And B₂And if only if to have connectedness or connection

The overlapping number of overlapping directed graph is defined as:

If B is the tag set of limit e, the overlapping number of definition:

oc(e)=|B|。

As preferably: calculate overlapping directed graph weight ODW and comprise the following steps:

A) overlapping penalty is built:

Wherein c=oc (e), it is seen that when p < when 1,It it is the subtraction function of c；

B) " reward function " of rewarding limit connectedness is built:

WhereinIt it is overlapping penaltyValue, for node N, definition w (N)=w (e) * w (e '), for the weights of calculating crunode；

C) overlapping directed graph weight is built:

$\mathrm{ODW}\left(G\right)=\underset{j=1}{\overset{\mathrm{\&epsiv;}}{\mathrm{\&Sigma;}}}w\left(N_j\right)$

According to this model, for the room seed of equal length and weight, more the seed of high-quality has higher overlapping directed graph weight, just can find high quality seed afterwards at the ODW (G) calculating candidate seed。

As preferably: step 5 comprises the following steps:

A) first determine that a stop value d, continuous seed length w and threshold value t, a d value are normally based on statistical principle and indicate an intended termination E value, then on the basis of consideration search background character, calculate suitable d value；

B) according to room spermotype, the WEB service in internal memory is described source data and carry out local matching, when the score value having a coupling string is higher than t, then have found a word string chosen, namely strengthen a little；

C) after a score value is chosen higher than the word string of t, carry out extending optimizing based on the local of dynamic programming algorithm, the minimum score value of regulation comparison is d, some negative score values can be run into when comparison extends, the score value making comparison declines, and when the score value declined is less than d, the extension of hit will terminate, the fragment at this moment in this section of comparison with highest score just becomes the coupling of one section of hit, thus finding satisfactory WEB service node。

The invention have the benefit that the present invention devises a kind of description at substantial amounts of WSDL and mates order calibration method in file, it is proposed that room Seed model, and find optimum room seed by overlapping directed graph, complete the coupling to target WEB service。When needs describe and find target at magnanimity WSDL in file, overcome traditional slow-footed shortcoming that matching way exists one by one successively, the present invention passes through the room seed that design performance is excellent, Seed model is utilized to go coupling WSDL to describe file, thus being greatly improved matching efficiency, and then it being effectively improved the performance of Web service, growing Internet service extension is had great significance by this。

Accompanying drawing explanation

Fig. 1 is the structure directed graph of room seed 1*1*1；

Fig. 2 is the structure directed graph of room seed 11**1；

Fig. 3 is the amendment first of room seed 1*1*1 structure directed graph；

Fig. 4 is the amendment first of room seed 11**1 structure directed graph；

Fig. 5 is the overlapping directed graph of room seed 1*1*1；

Fig. 6 is the overlapping directed graph of room seed 11**1。

Detailed description of the invention

For the ease of the understanding of those skilled in the art, below in conjunction with accompanying drawing and specific embodiment, the present invention is described further。

The present invention by according to WSDL describe file loading, the foundation of room seed, overlapping oriented map generalization, connective and overlapping number calculating, be described in detail based on five steps of coupling of optimum room seed:

Step 1: the server end WEB server substantial amounts of WSDL of loading from network describes file and describes file source data at server end formation magnanimity WSDL, and this step is prior art。

Step 1 method particularly includes: adopt HTML (Hypertext Markup Language) to connect long-range wsdl document address in need, WEB service describes file resolved by document object tree-model, it resolves form is hypertext tag format, for the document object tree resolved, all nodes comprising<wsdl:documentation>are loaded into the internal memory of server end。

Step 2: the foundation of room seed

1) room seed is defined according to below equation:

One room seed S is defined on character set Α={ 1, * }, and specifies to start with 1, the 1 fixed model string terminated。Wherein 1 represents coupling, and * is an asterisk wildcard, represents in this position it can is 1 (coupling) or 0 (mismatch)。S can be expressed as:

$S={\left(1\right)}^{i_1}{(*)}^{k_1}{\left(1\right)}^{i_2}{(*)}^{k_2}...{\left(1\right)}^{i_{n-1}}{(*)}^{k_{n-1}}{\left(1\right)}^{i_n}$

For an embodiment: if 1**1*1 is exactly a room seed。

Wherein (1)ⁱ(*)^kRepresent continuous i 1 and continuous k *, i_jRepresenting the different values of i, j ∈ [1, n], wherein n is natural number；K_jRepresenting the different values of k, j ∈ [1, n-1] wherein n is natural number。It can be seen that room seed S multiple 1 " 1 piece (1blocks) " formed and continuous multiple * forms continuously by several " * block (* blocks) " is constituted。Owing to regulation S must end up with 1 beginning and 1, so the number of " 1 piece " adds 1 equal to the number of " * block "。

The length (length) of room seed S is in S all 1 and the number of *, is expressed as | S |:

$\mathrm{length}\left(S\right)=\left|S\right|=\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{i}_j+\underset{j=1}{\overset{n-1}{\mathrm{\&Sigma;}}}{k}_j$

The weight w of room seed is defined as the number of all 1:

$\mathrm{weight}\left(S\right)=w=\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{i}_j$

The pattern of room seed is that * all in S replace with the expression string of 1 or 0 formation。Obviously, 2 are always had^|s|-wIndividual different pattern。

Step 3: overlapping oriented map generalization:

In order to be quickly found out optimum or the room seed close to optimum, we introduce the concept that seed is overlapping。The hit of seed can be overlapping, but the hit of overlap will detect the coupling of same place。Therefore, seed sensitivity is inversely proportional to overlapping hits, and namely good seed should have low overlapping hits。Even if different room seeds has identical length and weight, it still has different sensitivity and plyability。

1) structure room kernel texture directed graph:

The top-down directed graph produced by kernel texture。Top-down node be according to seed corresponding element from left to right 1 or 0, wherein 1 representation node 1, * represents two nodes: 1 and 0。Every directed edge indicates similarity p or q=1-p, pronounces " occurring with Probability p (or q) "。

In order to explain that different seeds has the different structure of identical length and weight, we do not adopt simplest room seed 1*1 as our first example because its only one of which candidate seed, i.e. 1*1, its length be 3 and weight be 2。We start with length for the room seed that 5 weights are 3。Here there are two candidate seed (removing symmetry): 1*1*1 and 11**1。

Fig. 1 and Fig. 2 shows the different kernel texture directed graph of corresponding for seed 1*1*1 and 11**1 two。

Definition according to room seed, each structure directed graph is by node 1 and terminates。For the room seed of a length L and weight w, L layer, each layer is had all to have at least a node from top-level node to bottom, such as node 1。Lower level node is connected with limit p (lower level node is for 1) or limit q (lower level node is for 0) with upper layer node。Obviously, a basic structure directed graph has w+2 (L-w) individual node。

The structure directed graph of room seed is a basic model for detecting the architectural difference between equal length and weight seed。Value in node is 1 or 0, is called nodal value。The overlapping directed graph of room seed, is the extension to structure directed graph。In order to make directed graph preserve more kernel texture information, it would be desirable to it is modified, including markers step and overlep steps。

2) markers step

First, we in order from the sequence number 1 each node of beginning label, are put into after nodal value and bracket with bracket from top node to bottom, then delete labelling p or q, because we can (1 corresponding p, 0 corresponding q) analysis obtains from lower level node；Meanwhile, head node ID is marked on every directed edge。Owing to this is the first step setting up overlapping directed graph, therefore all of limit all carries out labelling with start node (1), and from second step, start node will be changed to the node that lower floor's intermediate value is 1。

Fig. 3 and Fig. 4 shows the amendment result of room seed 1*1*1 and 11**1。The start node of structure directed graph is top-level node。This directed graph can as the primary version of overlapping directed graph, i.e. non-overlapping directed graph。

3) overlep steps

Overlep steps is the process of a series of iteration。Iterative process based on step, successively using every node layer value be 1 node as start node structural texture directed graph, but the point that start node adopts lower floor's intermediate value to be 1。The amendment first of structure directed graph is to proceed by from ground floor, therefore next step must from second layer node intermediate value 1 node, by that analogy。

When to last layer, we can see that its ground floor is identical, therefore we are parked in L-1 layer。

It has been created that overlapping directed graph after overlep steps operates。

Fig. 5 and Fig. 6 is the directed graph full release of room seed 1*1*1 and 11**1。The number of plies of overlapping directed graph: N (layers)=2L-2。

Step 4: calculate connective and overlapping number

1) connective calculating: have two limit e having common node₁And e₂, i.e. e₁End node be e₂Start node, if B₁It is e₁The set of labelling, B₂It is e₂The set of labelling。Claim B₁And B₂And if only if to have connectedness or connection

Such as, in the overlapping directed graph of room seed 1*1*1, there is connectedness on two limits of common node node (3), because B₁={ 1} and B₂={ 1} is soBut two limits (Far Left) for having common node (2) do not connect, because B₁={ 1} and B₂={ 2} is so

On path, all of limit is all the path that the path connected is called connection。L bar limit in the path of a connection that and if only if all comprises certain identical labelling, just says in path the pattern comprising room seed。

2) overlapping number calculates: set the tag set that B is limit e, the overlapping number of definition:

oc(e)=|B|

Connective and overlapping number is two principal elements judging room seed quality。If certain room seed has high connectedness, claiming it is high-quality, because more communication path can be detected such that it is able to find more pattern。On the other hand, high-quality room seed to have low overlapping number, because the hit of overlap will detect the coupling of same place, then low overlapping number will have high sensitivity。

Step 5: based on the coupling of optimum room seed

1) calculating of limit weight: the weight on limit is the function of connective and overlapping number。Define as follows:

Wherein oc (e) is the overlapping number of e,It is " overlapping penalty "；

2) definition overlap punishment and connection reward function:

Wherein c=oc (e)。Visible when p < when 1,It it is the subtraction function of c。

It is reward limit connective " reward function ", say, that if the limit that e has common node with it connects, w (e) should increase。If e ' is certain of e, there is the limit of common node, in the present invention, definitionAs follows:

WhereinIt it is overlapping penaltyValue。

For node N, definition w (N)=w (e) * w (e '), for the weights of calculating crunode。

3) overlapping directed graph weight is calculated

Overlapping directed graph weight definition is the weights sum of each edge in figure。A given overlapping directed graph G, if the weight ODW (G) that ε is the so overlapping directed graph of limit number of G defines as follows:

$\mathrm{ODW}\left(G\right)=\underset{j=1}{\overset{\mathrm{\&epsiv;}}{\mathrm{\&Sigma;}}}w\left(N_j\right)$

ODW (G) value of certain overlapping directed graph is based primarily upon the structure of room seed, overlapping number, connectedness and similarity etc.。Owing to relating to more parameter and allowing User-Defined Functions to calculate, overlapping Directed Graph Model will be more complete, accurately and flexibly。

According to this model, for the room seed of equal length and weight, more the seed of high-quality has higher overlapping directed graph weight, just can find high quality seed afterwards at the ODW (G) calculating candidate seed。

4) based on the coupling of optimum room seed

It is calculated for each candidate seed according to overlapping directed graph weight maximum, finds optimum room seed, and to carry out WEB service profile matching for standard a bit, it is characterised in that comprise the following steps:

A) a stop value d, continuous seed length w and a threshold value t are first determined。D value is normally based on statistical principle and indicates an intended termination E value, then calculates suitable d value on the basis of consideration search background character。

B) according to room spermotype, the WEB service in internal memory is described source data and carry out local matching, when the score value having a coupling string is higher than t, then have found a word string chosen, namely strengthen a little。

C) after a score value is chosen higher than the word string of t, carry out local and extend optimizing, the minimum score value of regulation comparison is d, some negative score values can be run into when comparison extends, the score value making comparison declines, and when the score value declined is less than d, the extension of hit will terminate, the fragment at this moment in this section of comparison with highest score just becomes the coupling of one section of hit, thus finding satisfactory WEB service node。

Those of ordinary skill in the art is it will be appreciated that embodiment described here is to aid in reader understanding's principles of the invention, it should be understood that protection scope of the present invention is not limited to such special statement and embodiment。Those of ordinary skill in the art can make various other various concrete deformation and combination without departing from essence of the present invention according to these technology disclosed by the invention enlightenment, and these deformation and combination remain in protection scope of the present invention。

Claims (2)

1. the WEB service matching process based on overlapping directed graph, it is characterised in that: comprise the following steps:

Step 1: describe file and the wsdl document of WEB service are loaded into the internal memory of server end by network, form WEB service and describe source data；

Step 2: set up room seed Matching Model, describes WEB service source data and is described with the form of room seed with target service data, obtain room seed the matching frame；

Described room seed, it is defined as:

One room seed S is defined on character set Α={ 1, * }, and specifies to start with 1, the 1 fixed model string terminated；Wherein 1 represents coupling, and * is an asterisk wildcard, represents in this position it can is 1 coupling or 0 mismatch；S is expressed as:

$S={\left(1\right)}^{i_1}{(*)}^{k_1}{\left(1\right)}^{i_2}{(*)}^{k_2}...{\left(1\right)}^{i_{n-1}}{(*)}^{k_{n-1}}{\left(1\right)}^{i_n}$

Wherein (1)^j(*)^kRepresent continuous j 1 and continuous k *；

The length (length) of room seed is in S all 1 and the number of *, is expressed as | S |:

$length\left(S\right)=\left|S\right|=\underset{j=1}{\overset{n}{\&Sigma;}}{i}_j+\underset{j=1}{\overset{n-1}{\&Sigma;}}{k}_j$

The weight w of room seed is defined as the number of all 1:

$weight\left(S\right)=w=\underset{j=1}{\overset{n}{\&Sigma;}}{i}_j$

The pattern of room seed is that * all in S replace with the expression string of 1 or 0 formation, always has 2^|s|-wIndividual different pattern；

Step 3: build the overlapping directed graph towards different rooms seed, calculates its connective and overlapping number for each overlapping directed graph；

Described step 3 comprises the following steps:

A) set up structure directed graph: top-down node be according to seed corresponding element from left to right 1 or 0, wherein 1 representation node 1, * represents two nodes: 1 and 0, and every directed edge indicates similarity p or q=1-p；

B) markers step: in order from the sequence number 1 each node of beginning label from top node to bottom, be put into after nodal value and bracket with bracket, then deletes labelling p or q, is marked on every directed edge by head node ID meanwhile；

C) overlep steps: overlep steps is the process of a series of iteration, successively using every node layer value be 1 node as start node structural texture directed graph, next step is from second layer node intermediate value 1 node, by that analogy, eventually settle at L-1 layer, after overlep steps, obtain the overlapping Directed Graph Model of room seed；

The connectedness of described overlapping directed graph is defined as:

There are two limit e having common node₁And e₂, i.e. e₁End node be e₂Start node, if B₁It is e₁The set of labelling, B₂It is e₂The set of labelling, claims B₁And B₂And if only if to have connectedness or connection

The overlapping number of overlapping directed graph is defined as:

If B is the tag set of limit e, the overlapping number of definition:

Oc (e)=| B |；

Step 4: find optimum room seed with maximum connectedness and minimum overlapping number for standard, and pass through

Overlapping directed graph weight ODW describes optimum room seed；

Calculate overlapping directed graph weight ODW to comprise the following steps:

A) overlapping penalty is built:

Wherein c=oc (e), it is seen that as p < 1,It it is the subtraction function of c；

B) " reward function " of rewarding limit connectedness is built:

WhereinIt it is overlapping penaltyValue, for node N, definition w (N)=w (e) * w (e '), for the weights of calculating crunode；

The weight on limit is the function of connective and overlapping number, defines as follows:

Wherein oc (e) is the overlapping number of e,It is " overlapping penalty "；

C) overlapping directed graph weight is built:

$ODW\left(G\right)=\underset{j=1}{\overset{\mathrm{\&epsiv;}}{\&Sigma;}}w\left(N_j\right)$

According to this model, for the room seed of equal length and weight, more the seed of high-quality has higher overlapping directed graph weight, just can find high quality seed afterwards at the ODW (G) calculating candidate seed；

Step 5: with target WEB service, optimum room seed is described source data and mates, obtains the WEB service matching result of optimum, finds the WEB service node of needs；

Step 5 comprises the following steps:

A) first determine that a stop value d, continuous seed length w and threshold value t, a d value are normally based on statistical principle and indicate an intended termination E value, then on the basis of consideration search background character, calculate suitable d value；

B) according to room spermotype, the WEB service in internal memory is described source data and carry out local matching, when the score value having a coupling string is higher than t, then have found a word string chosen, namely strengthen a little；

C) after a score value is chosen higher than the word string of t, carry out extending optimizing based on the local of dynamic programming algorithm, the minimum score value of regulation comparison is d, some negative score values can be run into when comparison extends, the score value making comparison declines, and when the score value declined is less than d, the extension of hit will terminate, the fragment at this moment in this section of comparison with highest score just becomes the coupling of one section of hit, thus finding satisfactory WEB service node。

2. a kind of WEB service matching process based on overlapping directed graph according to claim 1, it is characterized in that: step 1 method particularly includes: adopt HTML (Hypertext Markup Language) to connect long-range wsdl document address in need, WEB service describes file resolved by document object tree-model, it resolves form is hypertext tag format, for the document object tree resolved, all nodes comprising<wsdl:documentation>are loaded into the internal memory of server end。