CN102819575A - Personalized search method for Web service recommendation - Google Patents

Abstract

The invention discloses a personalized search method for Web service recommendation. The personalized search method comprises the following steps of: 1, preprocessing a WSDL (Web Services Description Language) file, i.e., forming a bag of words through two preprocessing steps of removing stop words and extracting stems; 2, extracting user interest, i.e., calculating weight of each word in the bag of words by using an improved TF-IDF (Term Frequency-Inverse Document Frequency) formula, and multiplying by a time decay factor of the word to obtain a new weight; selecting previous k words according to the weight from large to small as interest words of a user and corresponding weight of each word to form a k-dimension user interest vector; 3, calculating interest similarity, i.e., setting a similarity threshold and selecting the users with interest similarity exceeding the threshold as neighbor users of a target user; and 4, ordering service search results, calculating a recommended predicted value of the service according to similarity of neighbor users and the frequency of selecting service of the users, and arranging the searched results in a descending order according to the recommended predicted value, thereby obtaining the personalized search result.

Description

A kind of individuation search method that is used for the Web service recommendation

Technical field

The present invention relates to web search, recommendation in a kind of computer software technical field, particularly a kind of individuation search method that is used for the Web service recommendation.

Background technology

For the demand of the dirigibility, expansibility, correctness and the robustness that constantly satisfy software systems, the practice of soft project is progressively developed and certain methods, make that the structure of software systems can be based on existing software resource, but not from the beginning all develops.These methods have successfully been accelerated the tempo of development of software systems, have improved production efficiency.At the technological layer of method, the Function Decomposition that software is realized is some simple relatively reusable functional modules, also for soft project a kind of better software administration technology is provided.

Current, widely accepted software reuse technology be based on assembly soft project (Components-Based Software Engineering, CBSE).(Service Oriented Computing SOC) is a kind of new software development normal form based on assembly in service-oriented calculating; The infrastructure of SOC be service-oriented architectural framework (Service Oriented Architecture, SOA); Web service and SOA are a kind of realization versions of SOC.

As a kind of emerging, towards the distributed computing model of Internet, SOC provides the technology that better enables for structure loose coupling, inter-organization integrated application.Service-Oriented Architecture Based provides basic guarantee through the pattern of " issuing-search-bind " for using Service Source.Yet service user and ISP are separated, and increased user's understanding, have obtained and use the difficulty of required service.Particularly when user's demand changed along with the evolution of application construction process, how letting the user obtain proper service was the problem that needs solve.To this problem, the demand for services that the traditional services discovery technique mainly initiatively provides the mode of query requests to obtain the user through the user, perhaps directly let user oneself in resource collection according to manual the searching of certain taxonomic hierarchies.When resource collection constantly expands, the operation of manually searching service will become loaded down with trivial details, time-consuming, fallibility.At present, Web service search technique comprises based on UDDI registration center, through Web service website (like XMethods, RemoteMethods etc.), uses universal search engine (like Google, Yahoo etc.) and use four kinds of modes of professional search engine (like seekda, Merobase etc.).These ways of search are mainly supported the key search mode, in the retrieving and no user participate in, thereby result for retrieval has nothing to do with user interest, more can not change with the variation of user interest.

Different with the thinking of traditional search technique; The personalized search technology can be analyzed and compare with user's interest the service page in the Search Results; The help user therefrom finds out more interested service and it preferentially is presented in the search result list, thereby improves search efficiency of users.As in the Google personalized search; Look & feel (the rank that comprises information filtering that system allows customization oneself to like; Speech selection and query suggestion customization etc.); The personalized Subscribed Links of Google allows the user in the Google search engine of oneself, to create self-defined result, for the client represents service chaining. The personalized search of releasing allows the user to search for interest information according to the factum mode, and supports the user to the management of result for retrieval with share.The user can add note, can classify to the Web webpage and ordering etc. according to individual need.

Personalized recommendation technology tap/dip deep into user's personalization preferences; The information of the formula of taking the initiative " propelling movement " mode; Provide to robotization the information that meets individual requirements to the user; Rather than need user oneself from the Web information of magnanimity, to seek own interested content, thereby improve the efficient that user's effective information obtains.1992, first commending system Tapestry was born, and it is used for the collaborative filtering of Email and has obtained good effect.After this, commending system has obtained increasing concern with its wide application value.1996, Yahoo introduced portal website with commending system, added personalized user inlet MyYahoo, had proposed personalized service to different user; 1997, the AT&T laboratory proposed personalized recommendation system Referral Web and the PHOAKS based on collaborative filtering; Calendar year 2001, IBM Corporation has increased personalized recommendation system in its e-commerce platform Websphere, so that businessman's exploitation individual electronic business web site; Similar products like also has GroupLens, Amazon, Netflix etc., and application relates to electronic mail filtering, ecommerce class website, theme of news class website, search engine, online DVD lease website and some web2.0 socialization websites etc.

What personalized search used in a large number is the ultimate principle in the personalized recommendation; And personalized recommendation also needs a large amount of basic fundamentals of using for reference in the personalized search; Both are as tight association in the personalized service and two technology of core the most; Can satisfy the differentiation information requirement of different user largely, be with a wide range of applications.

Search engine can help the content that the user gets access to oneself efficiently, quickly from magnanimity Web resource need, thereby greatly improve the efficient that the user obtains information as the instrument of effective information retrieval.Along with enriching constantly of Web service resource and further developing of search engine technique, under the driving of user's actual need, individuation search method becomes the focus of search field research gradually.Individuation search method to Web service; Its core is according to the interest of user's personalization, preference; The service result for retrieval is carried out the screening and the ordering of " varying with each individual ", thereby the result for retrieval output of the differentiation that satisfies its individual demand is provided for different user.

Yet, how in the Web Internet resources, to find a kind of comparatively objective and accurate searching method, accurately service implementation pushes, and satisfies the needs of different the main consuming bodies, is a difficult point.

Summary of the invention

Goal of the invention: technical matters to be solved by this invention is to long defective of search out of true time in the prior art, and a kind of individuation search method that Web service is recommended that is used for is provided.

In order to solve the problems of the technologies described above, the invention discloses a kind of individuation search method that Web service is recommended that is used for, may further comprise the steps:

Step 1; Pre-service WSDL WSDL (Web Service Description Language; WSDL) document; From user's service recorder, obtain the WSDL document that it was selected,, form speech bag (bag ofwords) through removing stop words and extracting two pre-treatment step of stem;

Step 2 extracts user interest, uses improved TF-IDF formula to calculate the weight of each speech in the speech bag, and multiply by the time decay factor, obtains new weight δ _IjSelect new weight δ _IjBy the interest speech of big extremely little preceding k speech as the user, and the respective weights δ of each speech _Ij, the user interest of forming the k dimension is vectorial; The excellent weights of k before selecting, and corresponding speech constitutes the user interest vector together.This measure helps reducing the dimension of user interest vector space and makes its dimension consistent, helps calculating efficiently the interest similarity between per two users.

Step 3 is calculated similarity, uses the vector angle cosine formula to calculate cosine between per two users apart from as its similarity; Set similarity threshold, the user who surpasses threshold value goes into to elect as targeted customer's neighbours user; The setting range of similarity threshold is 0 ~ 1.

Step 4, ordering service result for retrieval: the targeted customer submits services request to, goes out all by the Web service search engine retrieving and meets requested service; According to neighbours user select these services number of times and with targeted customer's similarity, adopt the weighted mean predictor formula to calculate the recommendation predicted value of each result for retrieval; With result for retrieval according to recommending the predicted value descending sort, thereby obtain the personalized search result.

Among the present invention, improved TF-IDF (Term Frequency-Inverse Document Frequency, document-anti-document frequency) formula is following:

$\mathrm{tf}\left(t_{\mathrm{ij}}\right)=\frac{\mathrm{freq}(t_{\mathrm{ij}},D_i)}{\left|D_i\right|},$

$\mathrm{idf}\left(t_{\mathrm{ij}}\right)=\log \frac{\left|D\right|}{\left|\right\{D_i:t_{\mathrm{ij}}\∈D_i\left\}\right|},$

ω _ij＝tf(t _ij)*idf ²(t _ij)，

Wherein, t _IjBe i j speech in user's speech bag, tf (t _Ij) be speech t _IjDocument frequency, D _iBe i user's speech bag, freq (t _Ij, D _i) be speech t _IjAt speech bag D _iThe middle number of times that occurs, | D _i| be D _iThe number of middle speech, idf (t _Ij) be speech t _IjAnti-document frequency, | D| is the number of WSDL document in the corpus, | { D _i: t _Ij∈ D _i| expression speech t _Ijω appearred in what users' speech bag _IjBe speech t _IjWeight;

The computing method of time decay factor are following:

Decay＝2-e ^α*t，

Wherein, Decay express time decay factor, e is the end of natural logarithm, generally uses numerical value 2.718.α is an attenuation rate, and span is [0,0.1], for example can be set at 0.1.When the α value is 0, Decay=1, the expression weights are decay in time not, and the α value is big more, and it is fast more to decay, and t is current time and the distance users difference between the last time of selecting to serve.Corresponding to the decay characteristics in time that user interest had, designed the time decay factor.New weight is the product of the value and the time decay factor of former weight, and for non-selected speech of a specified duration, its weight decays to 0 gradually.

Speech t in each user's speech bag _IjNew weight δ _IjComputing formula is:

δ _ij＝ω _ij*Decay。

Among the present invention, it is following to calculate the similarity formula:

$\mathrm{sim}(u_a,u_b)=\frac{{\mathrm{\Σ}}_{j=1}^k{\mathrm{\δ}}_{\mathrm{aj}}*{\mathrm{\δ}}_{\mathrm{bj}}}{\sqrt{{\mathrm{\Σ}}_{j=1}^k{\mathrm{\δ}}_{\mathrm{aj}}^2}*\sqrt{{\mathrm{\Σ}}_{j=1}^k{\mathrm{\δ}}_{\mathrm{bj}}^2}},$

Wherein, u _aWith u _bBe two different users, sim (u _a, u _b) refer to the similarity between these two users, δ _AjAnd δ _BjBe respectively user u _aWith user u _bThe speech bag in the weight of j speech, k is the number of user interest speech.

Among the present invention, the formula of the recommendation predicted value of each result for retrieval of employing weighted mean predictor formula calculating is following:

$P_{u_t,s_t}={\stackrel{\‾}{c}}_{u_t}+\frac{{\mathrm{\Σ}}_{u_i\∈N}(c_{u_i,s_t}-{\stackrel{\‾}{c}}_{u_i})*\mathrm{sim}(u_t,u_i)}{\sqrt{{\mathrm{\Σ}}_{u_i\∈N}\mathrm{sim}{(u_t,u_i)}^2}},$

Wherein, u _tBe the targeted customer, s _tBe destination service, the service of recommendation predicted value promptly to be calculated,

Be targeted customer u _tTo destination service s _tThe recommendation predicted value,

With

Be respectively targeted customer u _tWith neighbours user u _iThe number of times of average selection service,

Be neighbours user u _iSelect target service s _tNumber of times, sim (u _t, u _i) be targeted customer u _tWith neighbours user u _iThe interest similarity, N is targeted customer u _tNeighbours set.

Among the present invention, remove stop words and refer to: in information retrieval, stop words is meant the speech that the frequency of occurrences is too high, do not have too overall search meaning.It is a step of vectorial participle in the knowledge extraction process that stop words is handled, speed and quality that its individual processing meeting speed up document is handled.At present, some English of publishing vocabulary of stopping using has been arranged, wherein more famous is inactive vocabulary that Van Rijsbergen delivers and the Brown Corpus vocabulary of stopping using.Chinese stop using vocabulary more famous Harbin Institute of Technology stop using vocabulary, Sichuan University machine intelligence laboratory stop using dictionary, the inactive vocabulary of Baidu etc. are arranged.The general vocabulary of stopping using not only comprises some general stop words, like a, and by, is; At etc., and be included in some vocabulary that the Web service field often occurs, service for example, soap; Response, request, set; Get etc., these speech are discrimination and little for Web service, and introduces easily and disturb.The speech that is contained in this table is removed from the WSDL document.WSDL document 7 important parameters: types, import, message, portType, operation, binding and service.These parameters are nested in the definitions root element.Adopt WSDL4J (Web Services Description Language for Java Toolkit; The JAVA kit of WSDL) the WSDL document of the user being selected is resolved; The content that parses is removed stop words, extract stem, form this user's speech bag.

Among the present invention, stem is meant that all inflectional affixes are removed the remaining part in back, and extracting stem is to remove the process that affixe obtains root.The present invention in the baud stem algorithm in univ cambridge uk's computer laboratory invention in 1979, carries out the extraction of stem for the speech in the WSDL document according to Martin doctor Poter, so that more accurately nothing repeatedly extracts the interest speech.

Compare with existing individuation search method; This method has three characteristics: the one, and not only implicit expression extracts user's interest itself; And obtained the relation between the different user interest through calculating the interest similarity; And adopt the collaborative filtering technology, and the service searching result is carried out the personalized ordering based on interest, improved the accuracy and the correlativity of Search Results to a certain extent; The 2nd, in the process that interest forms, added the time decay factor, represented the characteristics that user interest develops in time more exactly; The 3rd, but all off-line completion of first, second and third step of method are very little to the influence of recall precision.

The present invention just is being to use the ultimate principle in the personalized recommendation, and the collaborative filtering technology is applied to the personalized search of Web service, has improved user satisfaction and retrieval precision.Particularly, the present invention collects user's search records, describes from its Web service of selecting and extracts user interest the document, and form interest vector; According to the similarity of the cosine distance metric user interest of interest vector, the user who selects similarity with the targeted customer to surpass certain threshold value forms this user's neighbours; When the targeted customer submits the service search request to; Service recommendation system adopts one of above certain search technique to retrieve the service of a plurality of keyword matching for it; But directly result for retrieval is not returned to the user; But, by descending sort, return to the user then according to the recommendation predicted value that neighbours' selection experience and interest similarity thereof are calculated these result for retrieval.Like this, participate in service search result's customization user transparent, adopted the service recommendation method to accomplish personalized service search.

Beneficial effect: effect of the present invention is embodied in: the extraction of user interest does not need frequently to inquire the user or obtains the explicit feedback of user, thereby can obtain the approval and the use of more users user transparent.User interest and time correlation; The weight of the long-time interest that does not repeat to select decays gradually; Finally withdraw from the user interest vector, and the service interest of up-to-date frequent selection the user interest vector can be in time added to, thereby the variation of user interest can be expressed and follow the tracks of more accurately.Adopt the method for collaborative filtering that Search Results is recommended prediction and ordering,, also can from the experience of other similar users, obtain personalized recommendation even the targeted customer does not have the correlation experience of current required service.Can be widely used in the personalization of Web service search, support service recommendation, belong to computer software technical field.

Description of drawings

Below in conjunction with accompanying drawing and embodiment the present invention is done specifying further, above-mentioned and/or otherwise advantage of the present invention will become apparent.

Fig. 1 is a kind of process flow diagram that is used for the individuation search method of Web service recommendation of the present invention.

Embodiment

As shown in Figure 1, the invention discloses a kind of individuation search method that Web service is recommended that is used for, may further comprise the steps:

Step 1, pre-service WSDL document obtains the WSDL document that it was selected from user's service recorder, through removing stop words and extracting two pre-treatment step of stem, form the speech bag.

Step 2 extracts user interest, uses improved TF-IDF formula to calculate the weight of each speech in the speech bag, and multiply by the time decay factor of this speech, obtains new weight; Select weight by the interest speech of big extremely little preceding k speech as the user, and the respective weights of each speech, the user interest vector of composition k dimension.

Step 3 is calculated similarity, uses the vector angle cosine formula to calculate cosine between per two users apart from as its similarity; Set similarity threshold, the user who surpasses threshold value goes into to elect as targeted customer's neighbours user.

Step 4, ordering service result for retrieval: the targeted customer submits services request to, goes out all by the Web service search engine retrieving and meets requested service; According to neighbours user select these services number of times and with targeted customer's similarity, adopt the weighted mean predictor formula to calculate the recommendation predicted value of each result for retrieval; With result for retrieval according to recommending the predicted value descending sort, thereby obtain the personalized search result.

Improved TF-IDF formula is following:

$\mathrm{tf}\left(t_{\mathrm{ij}}\right)=\frac{\mathrm{freq}(t_{\mathrm{ij}},D_i)}{\left|D_i\right|},$

$\mathrm{idf}\left(t_{\mathrm{ij}}\right)=\log \frac{\left|D\right|}{\left|\right\{D_i:t_{\mathrm{ij}}\∈D_i\left\}\right|},$

ω _ij＝tf(t _ij)*idf ²(t _ij)，

Wherein, t _IjBe i j speech in user's speech bag, tf (t _Ij) be speech t _IjDocument frequency, D _iBe i user's speech bag, freq (t _Ij, D _i) be speech t _IjAt speech bag D _iThe middle number of times that occurs, | D _i| be D _iThe number of middle speech, idf (t _Ij) be speech t _IjAnti-document frequency, | D| is the number of WSDL document in the corpus, | { D _i: t _Ij∈ D _i| mean t _Ijω appearred in what users' speech bag _IjBe speech t _IjWeight.

The computing method of time decay factor are following:

Decay＝2-e ^α*t，

Wherein, Decay express time decay factor, e is the end of natural logarithm; α is an attenuation rate, and span is [0,0.1]; When the α value is 0, Decay=1, the expression weights are decay in time not; The α value is big more, and it is fast more to decay, and t is current time and the distance users difference between the last time of selecting to serve;

Speech t in each user's speech bag _IjNew weight calculation formula be:

δ _ij＝ω _ij*Decay。

Among the present invention, it is following to calculate the similarity formula:

$\mathrm{sim}(u_a,u_b)=\frac{{\mathrm{\Σ}}_{j=1}^k{\mathrm{\δ}}_{\mathrm{aj}}*{\mathrm{\δ}}_{\mathrm{bj}}}{\sqrt{{\mathrm{\Σ}}_{j=1}^k{\mathrm{\δ}}_{\mathrm{aj}}^2}*\sqrt{{\mathrm{\Σ}}_{j=1}^k{\mathrm{\δ}}_{\mathrm{bj}}^2}},$

Wherein, u _aWith u _bBe two different users, sim (u _a, u _b) refer to the similarity between these two users, δ _AjAnd δ _BjBe respectively user u _aWith user u _bThe speech bag in the weight of j speech, k is the number of user interest speech.

Among the present invention, the formula of the recommendation predicted value of each result for retrieval of employing weighted mean predictor formula calculating is following:

$P_{u_t,s_t}={\stackrel{\‾}{c}}_{u_t}+\frac{{\mathrm{\Σ}}_{u_i\∈N}(c_{u_i,s_t}-{\stackrel{\‾}{c}}_{u_i})*\mathrm{sim}(u_t,u_i)}{\sqrt{{\mathrm{\Σ}}_{u_i\∈N}\mathrm{sim}{(u_t,u_i)}^2}},$

Wherein, u _tBe the targeted customer, s _tBe destination service, the service of recommendation predicted value promptly to be calculated, Be targeted customer u _tTo destination service s _tThe recommendation predicted value,

With

Be respectively targeted customer u _tWith neighbours user u _iThe number of times of average selection service,

Be neighbours user u _iSelect target service s _tNumber of times, sim (u _t, u _i) be targeted customer u _tWith neighbours user u _iThe interest similarity, N is targeted customer u _tNeighbours set.

Embodiment

The substance of present embodiment is from Web service supermarket (http: // 125.221.225.2:8080/WSSM/) background data base.

Present embodiment comprises following four steps:

(1) pre-service WSDL document

From the background data base in Web service supermarket, extract 200 users' service recorder, obtain raw data, the service recorder of certain customers is following:

Table 1 user service recorder (part)

Enumerated four users in the table 1, user name is respectively: " tailaoliu ", and " fangfang ", " zww ", " skh " selected some Web services respectively.Document is described in the Web service that download and each user of pre-service selected, and removes stop words according to the inactive vocabulary that Van Rijsbergen delivers, and adopts Martin doctor's Poter poter stem algorithm to extract stem, formation speech bag.Selected service " BookingService " by name, " JasonsBooking ", three Web services of " HotelBookingEngine " like " taolaoliu "; From the WSDL document of three Web services of corresponding service website, download, be " render (84), hotel (99); reservation (40), invoice (33), room (269); city (81), client (13), book (194); ticket (13), basket (42), rate (25) " with extracting the speech bag that forms behind the stem through removing stop words.This speech bag comprises 11 speech altogether, wherein in the bracket behind each speech mark be the number of times that this speech occurs in document.

(2) extract user interest

All users' speech bag is formed corpus, uses improved TF-IDF formula to calculate the weight of each speech in the speech bag; The weight of the speech in each user's speech bag multiply by the time decay factor, obtain new weight.Excellent speech and the corresponding weights thereof of k formed the user interest vector before the weight.Like " render " speech in the speech bag of " taolaoliu "; The number of times that in the speech bag of " taolaoliu ", occurs is 84 times; In 200 users' speech bag, one has in 68 users' the speech bag and this speech occurred, therefore; The weights that calculate " render " according to improved TF-IDF formula in the claim step 2 are following

$\mathrm{tf}(\′\′\mathrm{render}\′\′)=\frac{84}{11}=7.64,$

$\mathrm{idf}(\′\′\mathrm{render}\′\′)=\log \frac{200}{68}=0.47,$

ω _“render”＝7.64*0.47 ²＝1.68，

Continue decay factor computing time, the α value is 0.05, and the t value is an initial value 1, and Decay calculates as follows:

Decay＝2-e ^0.05*1＝0.95，

New weights calculate as follows:

δ _“render”＝1.68*0.95＝1.59

Likewise, calculate the weights of remaining speech in the speech bag, and preceding 6 maximum (being k=6) speech of weighting value, the interest vector that obtains user " taolaoliu " is: < (basket, 6.42), (hotel, 4.03), (room; 3.15), (book, 2.82), (render, 1.59), (information, 1.24)＞, user's " fangfang " interest vector is: < (book; 3.31), (price, 3.26), (title, 3.23), (author, 3.17); (ISBN, 2.15), (infomation, 1.13)＞, user's " zww " interest vector is: < (weather, 4.42), (city; 3.33), (forecast, 3.29), (replication, 2.12), (add, 1.12); (id, 1.11)＞, user's " skh " interest vector is: < (weather, 3.39), (comment, 3.31), (forecast; 2.27), (city, 2.22), (replication, 1.20), (add, 1.10)＞.Wherein, the new weights of this speech of numeral behind the interest speech.

(3) calculate the interest similarity

Use the vector angle cosine formula to calculate cosine between per two users apart from as its similarity; Set similarity threshold, the user who surpasses threshold value goes into to elect as targeted customer's neighbours.For example, adopt calculating formula of similarity calculating user " taolaoliu " and user's " fangfang " similarity following:

$\mathrm{sim}(\&prime;\&prime;\mathrm{taolaoliu}\&prime;\&prime;,\&prime;\&prime;\mathrm{fangfang}\&prime;\&prime;)=$

$\frac{2.82*3.31+1.24*1.13}{\sqrt{{6.42}^2+{4.03}^2+{3.15}^2+{2.82}^2+{1.59}^2+{1.24}^2}*\sqrt{{3.31}^2+{3.26}^3+{3.23}^2+{3.17}^2+{2.15}^2+{1.13}^2}}=0.17,$

Setting similarity threshold is 0.15, then " taolaoliu " and user " fangfang " neighbours user each other.

(4) ordering service result for retrieval

The targeted customer submits services request to, and the Web service supermarket retrieves all for it and meets requested service; According to neighbours' services selection experience and with targeted customer's similarity, adopt the weighted mean predictor formula to calculate the recommendation predicted value of each result for retrieval.For example; Services request for targeted customer " taolaoliu " submission; Comprise the Web service of service " BookStoreService " by name in the result for retrieval, if this service was only selected 3 times by " fangfang ", the number of times of user's " taolaoliu " average selection service is 2; The number of times of user's " fangfang " average selection service is 1.5, and the recommendation predictor calculation that then should serve is following:

$P_{\&prime;\&prime;\mathrm{BookStoreService}\&prime;\&prime;}=2+\frac{(3-1.5)*0.17}{\sqrt{{0.17}^2}}=3.5$

With result for retrieval according to recommending the predicted value descending sort, thereby the user can obtain meeting the personalized search result of its interest from first page of result for retrieval rapidly.

Implementation result:

User " zww " hopes to obtain the service of buying books on the net as the current goal user.Respectively to the seekda search system ( Http:// webservices.seekda.com/, belong to prior art) and after the Web service supermarket submitted services request key word " book " to, the Search Results of preceding 10 ranks that obtain was respectively shown in table 2 and table 3.

The service of table 2.seekda Search Results top 10

In the table 2, having only sequence number is that 2,4,5 service provides the function of buying books on the net, and user " zww " also will manually seek the service that meets own demand after obtaining the return results of system, and this process is consuming time often, uninteresting, fallibility.

The service of table 3.Web service supermarket Search Results top 10

In the table 3, except sequence number is 9 service, all the other services all with buy books on the net relevant.This shows that personalized search uses collaborative filtering mode calculation services to recommend predicted value, can improve service searching accuracy and user search efficient, improves the satisfaction of user to the Web service search engine.

The invention provides a kind of individuation search method that Web service is recommended that is used for; The method and the approach of concrete this technical scheme of realization are a lot, and the above only is a preferred implementation of the present invention, should be understood that; For those skilled in the art; Under the prerequisite that does not break away from the principle of the invention, can also make some improvement and retouching, these improvement and retouching also should be regarded as protection scope of the present invention.The all available prior art of each ingredient not clear and definite in the present embodiment realizes.

Claims (4)

1. one kind is used for the individuation search method that Web service is recommended, and it is characterized in that, may further comprise the steps:

Step 1, pre-service WSDL WSDL document: from user's service recorder, obtain the WSDL document that it was selected,, form the speech bag through removing stop words and extracting two pre-treatment step of stem;

Step 2 extracts user interest: calculate the weight of each speech in the speech bag, and multiply by the time decay factor, obtain new weight δ _IjSelect new weight δ _IjBy the interest speech of big extremely little preceding k speech as the user, and the respective weights δ of each speech _Ij, the user interest of forming the k dimension is vectorial;

Step 3 is calculated the interest similarity: calculate cosine between per two user interest vectors apart from as its interest similarity; Set similarity threshold, the user who surpasses threshold value goes into to elect as targeted customer's neighbours user;

Step 4, ordering service result for retrieval: the targeted customer submits services request to, goes out all by the Web service search engine retrieving and meets requested service; According to neighbours user select these services number of times and with targeted customer's similarity, adopt the weighted mean predictor formula to calculate the recommendation predicted value of each result for retrieval; With result for retrieval according to recommending the predicted value descending sort, thereby obtain the personalized search result.

2. a kind of individuation search method that Web service is recommended that is used for according to claim 1 is characterized in that, in the step 2, calculates the weight of each speech in the speech bag, and multiply by the time decay factor of this speech, obtains new weight δ _IjComprise the steps:

Use improved TF-IDF formula to calculate weights omega _Ij:

$\mathrm{tf}\left(t_{\mathrm{ij}}\right)=\frac{\mathrm{freq}(t_{\mathrm{ij}},D_i)}{\left|D_i\right|},$

$\mathrm{idf}\left(t_{\mathrm{ij}}\right)=\log \frac{\left|D\right|}{\left|\right\{D_i:t_{\mathrm{ij}}\&Element;D_i\left\}\right|},$

ω _ij＝tf(t _ij)*idf ²(t _ij)，

Wherein, t _IjBe i j speech in user's speech bag, tf (t _Ij) be speech t _IjDocument frequency, D _iBe i user's speech bag, freq (t _Ij, D _i) be speech t _IjAt speech bag D _iThe middle number of times that occurs, | D _i| be D _iThe number of middle speech, idf (t _Ij) be speech t _IjAnti-document frequency, | D| is the number of WSDL document in the corpus, | { D _i: t _Ij∈ D _i| expression speech t _Ijω appearred in what users' speech bag _IjBe speech t _IjWeight;

The computing method of time decay factor are following:

Decay＝2-e ^α*t，

Wherein, Decay express time decay factor, e is the end of natural logarithm; α is an attenuation rate, and span is [0,0.1]; When the α value is 0, Decay=1, the expression weights are decay in time not; The α value is big more, and it is fast more to decay, and t is current time and the distance users difference between the last time of selecting to serve;

Speech t in each user's speech bag _IjNew weight δ _IjComputing formula is:

δ _ij＝ω _ij*Decay。

3. a kind of individuation search method that Web service is recommended that is used for according to claim 1 is characterized in that, adopts following method to calculate the user interest similarity in the step 3:

$\mathrm{sim}(u_a,u_b)=\frac{{\mathrm{\&Sigma;}}_{j=1}^k{\mathrm{\&delta;}}_{\mathrm{aj}}*{\mathrm{\&delta;}}_{\mathrm{bj}}}{\sqrt{{\mathrm{\&Sigma;}}_{j=1}^k{\mathrm{\&delta;}}_{\mathrm{aj}}^2}*\sqrt{{\mathrm{\&Sigma;}}_{j=1}^k{\mathrm{\&delta;}}_{\mathrm{bj}}^2}},$

Wherein, u _aWith u _bBe two different users, sim (u _a, u _b) refer to the similarity between these two users, δ _AjAnd δ _BjBe respectively user u _aWith user u _bThe speech bag in the weight of j speech, k is the number of user interest speech.

4. a kind of individuation search method that Web service is recommended that is used for according to claim 1 is characterized in that, in the step 4, the formula of the recommendation predicted value of each result for retrieval of employing weighted mean predictor formula calculating is following:

$P_{u_t,s_t}={\stackrel{\&OverBar;}{c}}_{u_t}+\frac{{\mathrm{\&Sigma;}}_{u_i\&Element;N}(c_{u_i,s_t}-{\stackrel{\&OverBar;}{c}}_{u_i})*\mathrm{sim}(u_t,u_i)}{\sqrt{{\mathrm{\&Sigma;}}_{u_i\&Element;N}\mathrm{sim}{(u_t,u_i)}^2}},$

Wherein, u _tBe the targeted customer, s _tBe destination service, the service of recommendation predicted value promptly to be calculated,

Be targeted customer u _tTo destination service s _tThe recommendation predicted value,

With

Be respectively targeted customer u _tWith neighbours user u _iThe number of times of average selection service,

Be neighbours user u _iSelect target service s _tNumber of times, sim (u _t, u _i) be targeted customer u _tWith neighbours user u _iThe interest similarity, N is targeted customer u _tNeighbours set.

Images