CN101799822A - Method for modeling layered network knowledge model and method for establishing knowledge warehouse system - Google Patents
Method for modeling layered network knowledge model and method for establishing knowledge warehouse system Download PDFInfo
- Publication number
- CN101799822A CN101799822A CN201010113228A CN201010113228A CN101799822A CN 101799822 A CN101799822 A CN 101799822A CN 201010113228 A CN201010113228 A CN 201010113228A CN 201010113228 A CN201010113228 A CN 201010113228A CN 101799822 A CN101799822 A CN 101799822A
- Authority
- CN
- China
- Prior art keywords
- knowledge
- attribute
- notion
- knowledge model
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a method for modeling a layered network knowledge model, which comprises the following steps of: confirming a knowledge field to be modeled; establishing a knowledge model of the knowledge field; and detecting whether the knowledge model is logic or not. The invention also discloses a method for establishing a knowledge warehouse system by utilizing the knowledge model, comprising the following steps of: establishing a knowledge database group for storing knowledge concepts and examples thereof; establishing a non-knowledge database for recording the storage position of each example in the knowledge database group; and establishing a knowledge edition database for managing different knowledge editions in the knowledge warehouse. The invention uses the layered network knowledge modeling method to facilitate data modeling, provides maximum expandability for knowledge models and organizes all knowledge related in the system by adopting distributed storage and multi-computer automatic mutual backup strategies to form a data warehouse in a true meaning, thereby improving the robustness, the fault tolerance and the recoverability of the whole system.
Description
Technical field
The present invention relates to the e-commerce technology field, particularly a kind of layered network knowledge model modeling method and knowledge warehouse system method for building up.
Background technology
Different based on the internet, applications that document represents with great majority, e-commerce system need store and handle mass data.These data comprise product information, Business Information, Bidder Information and product stream information etc.Integration, the analysis how safely and efficiently these data to be carried out on the logic aspect are the matters of utmost importance that e-commerce system faces.In addition, how intelligent the and automaticity of the information processing of elevator system also is that the decision e-commerce system is runed the key of success factor to reduce system research and development and maintenance cost, to shorten the R﹠D cycle.
As everyone knows, present e-commerce system and other Internet application system all use relational database to carry out data storage, and for example: well-known relational database systems such as ORACLE, MYSQL, DB2, SQL Server are being played the part of data manager's role in e-commerce field.In these Database Systems, data are stored in respectively among many forms (Table).System Architect comes data are integrated by the major key in the specific data form (PrimaryKey), external key (Foreign Key) or data directory (Index), to improve data retrieval efficient.When research and development are used based on the Web of relational database, at first need veteran System Architect and domain expert to link up fully, determine by architect which data form native system needs, and needs which data field in each form then.This is very complicated, difficult and make mistakes an easily process normally.The e-commerce system of a maturation often needs up to a hundred data forms, if because and the domain expert links up comprehensively, the architect experience is abundant inadequately or form between data relationship do not set correct, the capital largely has influence on the exploitation of upper layer application, causes the fatal result of customer experience difference the most at last.Even avoided in issuable problem of database table designs stage by increase R﹠D costs, the method for prolongation R﹠D cycle, this pattern still faces test in application and development.Design Mode according to MVC (represent, logic and data separating), in the Web page, should there be initial data base link and the used Sql statement of Query Database, this just needs to use storing process (Procedure/Package) that data are conducted interviews, this has strengthened the system research and development workload to a great extent, is unfavorable for controlling R﹠D costs and cycle.To have under numerous senior domain experts, System Architect and programming personnel's the participation also be very difficult even building the intelligent and sufficiently high application system of automaticity of a cover on the basis of present relational database.
Specifically, there is following defective in present internet electronic business system aspect data modeling and the access:
1, data modeling difficulty.The data of a perfect e-commerce system need nearly hundred database tables to store usually, after needing architect and domain expert to link up efficiently, association between the structure of form and the table could determine, but often the domain expert is not IT expert, architect is known little about it for commercial field again, resulting between the two communication disorders will influence the accurate type and the validity of data model to a great extent, and this problem tends to directly have influence on building and customer experience of whole application system;
2, data integrity is poor, and logicality is poor, and relevance is low between the data.Design Mode based on database table, every database table is relatively independent, but actual application system often requires these data that are distributed among the different pieces of information form can produce various associations, uses in order to search engine, purchase guiding system or the commending system on upper strata.Present method is that the mode by outer key assignments and given query condition will produce hard affinity between the data form, and considers that from the modeling initial stage data integrity, relevance and the logicality of total system are very difficulty or even the target that can not reach.This intellectuality service, robotization service that has just caused application system to provide is greatly limited.
3, system research and development and maintenance cost are restive, and the R﹠D cycle is long.The design of present electronic commerce data system comprises following several cost: domain expert's design data cost, domain expert and IT expert's communication cost, IT expert's database design cost, cost that the data base programmer builds database table, set up cost related between the database table and the data base programmer writes the cost that storing process is operated database.These complicated processes not only can consume a large amount of funds, R﹠D cycle of system is become be difficult to control.
Summary of the invention
(1) technical matters that will solve
The purpose of this invention is to provide a kind of layered network knowledge model modeling method and utilize this knowledge model to set up the method for knowledge warehouse system, to solve the data modeling difficulty, data integrity is poor, logicality is poor, relevance is low between the data, technical matterss such as system research and development and maintenance cost are restive, and the R﹠D cycle is long.
(2) technical scheme
For achieving the above object, the invention provides a kind of layered network knowledge model modeling method, may further comprise the steps:
S1: the ken of determining to want modeling;
S2: described ken is set up knowledge model by following formula, and described knowledge model KM comprises top layer knowledge model u (k), middle level knowledge model m (k) and bottom knowledge model l (k), and D represents ken,
KM＝f(D，u(k)∪m(k)∪l(k))
Described u (k) comprises the notion among the D, the sub-notion among m (k) the definition u (k), and l (k) comprises the example of described notion and sub-notion;
S3: utilize RACER reasoning from logic instrument that described knowledge model is verified, for the knowledge model that does not satisfy the logical condition requirement, need make amendment makes it can be by the checking of RACER.
Wherein, top layer knowledge model u (k) modeling comprises step among the described step S2:
Determine notion in the top layer knowledge model according to ken;
Judge whether to be notion specified limit and the note in the top layer knowledge model then direct if desired specified limit and note;
The attribute of the notion in the definition top layer knowledge model carries out the Attribute domain of attribute and codomain relatedly simultaneously, and described attribute comprises concept attribute and data attribute;
Judge whether to be described attribute specified limit, if desired then directly defined attribute restriction;
Utilize RACER reasoning from logic instrument that described top layer knowledge model is verified, for the top layer knowledge model that does not satisfy the logical condition requirement, need make amendment makes it can be by the checking of RACER.
Wherein, middle level knowledge model m (k) modeling comprises step among the described step S2:
The sub-notion of the notion in the definition top layer knowledge model;
Judge whether to be described sub-notion specified limit and note then direct if desired specified limit and note;
Define the attribute of described sub-notion, simultaneously the Attribute domain of attribute and codomain are carried out relatedly, described attribute comprises sub-concept attribute and data attribute;
Judge whether to be described attribute specified limit, if desired then directly defined attribute restriction;
Utilize RACER reasoning from logic instrument that described middle level knowledge model is verified, for the middle level knowledge model that does not satisfy the logical condition requirement, need make amendment makes it can be by the checking of RACER.
Wherein, end top layer knowledge model l (k) modeling comprises step among the described step S2:
Determine notion or sub-notion that example is affiliated;
Specify unique ID for described each example and represent this example;
Judge whether and to specify note for described example, then directly specify note if desired;
The attribute of described example institute's categorical conception or sub-notion is carried out instantiation, comprise Attribute domain instantiation and codomain instantiation;
Utilize RACER reasoning from logic instrument that described bottom knowledge model is verified, for the bottom knowledge model that does not satisfy the logical condition requirement, need make amendment makes it can be by the checking of RACER.
In order to achieve the above object, the present invention also provides a kind of method of the ecommerce knowledge warehouse system of setting up based on above-mentioned knowledge model, may further comprise the steps:
Set up the knowledge data base group, described knowledge data base group comprises n platform knowledge data base server, is used for storing the notion of top layer knowledge model and middle level knowledge model and described notion corresponding example in the bottom knowledge model, and wherein n is a positive integer;
Set up non-knowledge data base, creation of knowledge example routing table in described non-knowledge data base is used for writing down the IP address of each example at knowledge data base group's knowledge data base server;
Set up knowledge edition data storehouse, described knowledge edition data storehouse be used for according to top layer knowledge model and middle level knowledge model variation control described data warehouse version.
Wherein, described foundation comprises the step that every knowledge data base server stores example amount threshold is set in the knowledge data base group step, when the storage of first station server reaches threshold value, then stores newly-increased example into second station server.
Wherein, described method also comprises sets up backup database, is used for described knowledge data base group's data are backed up.
Wherein, described method also comprises the administration module of creation of knowledge database cluster, non-knowledge data base, knowledge edition data storehouse and backup database respectively.
Wherein, described method also comprises creates the interface that is used for middleware visit knowledge data base group and non-knowledge data base, and described middleware comprises: JENA semantic net middleware and data management middleware.
(3) beneficial effect
The present invention has following beneficial effect:
1, creates extendible logic knowledge model.The present invention uses the layered network knowledge modeling method of original creation, has created the knowledge model of e-commerce field.This knowledge model have can expand, advantages such as integrality is strong, logicality is strong, unambiguity.
2, provide the smart electronics business service.The present invention makes the ecommerce upper layer application to carry out knowledge acquisition and management by the JENA middleware by complete knowledge warehouse framework is provided.This can significantly advance the intellectuality and the automated process of e-commerce system search engine, purchase guiding system and commending system etc.
3, reduce the data system design cost, shorten the research and development time.The domain expert can directly carry out the visual modeling of domain knowledge according to the layered network knowledge modeling method.This will significantly reduce the cost with IT expert's communication cost, the design cost that reduces data system, the cost that reduces the modification design data and assurance data consistency.The present invention meets international state-of-the-art semantic net knowledge description standard OWL2.0, this design can be supported JENA middleware standard fully, carry out the knowledge data visit by JENA and can reduce the size of code of data access to a great extent, thereby significantly shorten the system research and development time.
4, the knowledge warehouse system fault-tolerance is strong, robustness is strong, restorability is strong.Knowledge model multimachine quilt is mutually adopted in the design of knowledge warehouse among the present invention, the knowledge instance load balancing, and knowledge data is stored the theory of separating with visit with non-knowledge data.Such design has reduced negative the influencing each other between data access, makes whole knowledge warehouse system have extremely strong fault-tolerance, robustness and restorability.
Description of drawings
Fig. 1 is according to layered network knowledge model modeling method process flow diagram of the present invention;
Fig. 2 is according to layered network knowledge model modeling method top layer knowledge model modeling process flow diagram of the present invention;
Fig. 3 is according to layered network knowledge model modeling method of the present invention middle level knowledge model modeling process flow diagram;
Fig. 4 is according to layered network knowledge model modeling method bottom knowledge model modeling process flow diagram of the present invention;
Fig. 5 is an embodiment according to layered network knowledge model modeling method bottom knowledge model of the present invention;
Fig. 6 sets up ecommerce knowledge warehouse system process flow diagram according to of the present invention based on above-mentioned knowledge model;
Fig. 7 is the ecommerce knowledge warehouse system embodiment frame diagram of setting up according to the present invention;
Fig. 8 is the experimental result picture of ecommerce knowledge warehouse system among Fig. 7.
Embodiment
A kind of layered network knowledge model modeling method that the present invention proposes is described as follows in conjunction with the accompanying drawings and embodiments.
Layered modeling method and netted modeling are the cores of this method in the layered network knowledge model modeling method of the present invention.Wherein layered modeling is meant that the establishment with the domain knowledge model is divided into " top layer knowledge model " in logic, " middle level knowledge model " and " bottom knowledge model ".The process of knowledge Modeling will be carried out one by one according to the order " top-down " of top layer modeling, middle level modeling and bottom modeling.Netted modeling represents that the domain expert is that notion is set concept attribute and data attribute flexibly, with extendible knowledge network of the final formation of field concept mesh topology.Below be the interrelated logic definition of knowledge model:
Definition 1: field
The field is the part of objective world, is the finite union of all notions in the problem space.
D≡c
1∪c
2∪...∪c
n，
(1) in the formula, D represents the field, equals c on the ≡ presentation logic
iThe expression notion, Ω problem of representation space.
Definition 2: knowledge
Knowledge is promptly in the union of all notions of a certain field and all axioms.
K≡D∪A，A＝f(D，p，r，a) (2)
(2) in the formula, K represents knowledge, and A represents axiom, and f represents abstract function (g hereinafter represents abstract function equally), and p represents the attribute of notion in the field, and r represents in the field that for the restriction of knowledge, a represents the supplemental text explanation for knowledge.So-called axiom for the asserting of knowledge in the field, is that the prejudgementing character of notion is described promptly, will introduce every content of axiom below in detail.
Definition 3: notion and example
A class has the set of the example of common property in the representation of concept objective world in the knowledge.For example, in e-commerce field, digital product, household electrical appliance, cosmetics etc. all are notions.
Example is certain individuality that exists in the objective world, and each example is under the jurisdiction of a notion at least.For example, jason is an example of client's notion.
Definition 4: attribute
Attribute be connect two notion examples or connect a notion example and data between the predicate relation.Attribute is divided into concept attribute and data attribute two classes, and concept attribute is represented two relations between the notion, and data attribute is represented the relation between a notion and the data.Can represent by formula (3):
For example, a concept attribute " hasBrand " (" having brand ") can be expressed as hasBrand (certain product, certain brand) in e-commerce field.A data attribute " hasSize " (" size for ") can be expressed as hasSize (certain part clothes, L).Attribute be based upon between two notions often or notion and data type between related, first notion c wherein
1The territory, source (Domain) that is called attribute, second notion c
2Or data type is called the codomain (Range) of attribute.
In fact, in real world, attribute is to specify in related between two notion examples or notion example and the data instance, but when modeling, for the abstractness and the general modeling person that guarantee model need be at notion or data type level definition attributes.For example: hasBrand (commodity, brand).When creating example, again attribute is specialized, for example: hasBrand (certain mobile phone, NOKIA).Wherein certain mobile phone is an example of commodity and " NOKIA " is an example of brand.
Definition 5: restriction
Limit (r) and describe for assigning the mathematical logic on notion or attribute, its purpose is the example establishment of qualification notion or some characteristic of attribute.Formula (4) has provided the definition of restriction:
Define the restriction of 5.1 notions
Employed notion restriction is mainly and equates restriction (equal) and separation limit (disjoint) among the present invention.
(1) if two notions satisfy following condition: then c
i≡ c
j
(2) if n notion satisfies following condition:
c
1∩c
2∩...∩c
n≡φ (6)
Then this n notion is separated, and its implication is that the example of any two notions in n the notion is all different.
Define 5.2 attribute limits
Employed attribute limits is divided into concept attribute restriction and data attribute restriction among the present invention.The concept attribute restriction is applied to concept attribute, and what mainly use is concept attribute value restricted number.Concept attribute value restricted number is mainly used to limit the number that an attribute can have property value, and such restriction can be divided into three kinds:
(1) minimum concept attribute value restricted number, the value of arranging certain attribute comprises the quantity of example at least.The logical expression of this restriction is:
≥ObjectProperty(number?of?instance) (7)
Wherein ObjectProperty represents certain concept attribute, and the independent variable in this function is the example quantity that logic allows.For example: can assign restriction 〉=hasSeller (1) in hasSeller (commodity, the dealer) attribute, the logic implication of this restriction is " every commodity have a dealer at least ".
(2) maximum concept attribute value restricted number, the value of arranging certain attribute comprises the quantity of example at most.The logical expression of this restriction is:
≤ObjectProperty(number?of?instance) (8)
For example: can assign restriction≤hasBrand (1) in hasBrand (commodity, the brand) attribute, the logic implication of this restriction is " every commodity have a brand at the most ".
(3) accurate concept attribute value restricted number is arranged the exact magnitude of the example that value allowed of certain attribute.The logical expression of this restriction is:
＝ObjectProperty(number?of?instance) (9)
For example: hasSex (client, sex) attribute can be assigned restriction=hasSex (1), and the logic implication of this restriction is " client has and have only a sex ".
Seemingly restricted with concept attribute, the data attribute restriction that the present invention mainly uses is a data attribute value restricted number, and difference only is that data attribute value restricted number is mainly used in the desirable data number of restricting data attribute.But data attribute value restricted number simple table is shown:
≥(≤，＝)DatatypeProperty(number?of?data) (10)
Wherein DatatypeProperty represents certain data attribute, and the independent variable in this function is the data bulk that logic allows.For example: hasProductID (commodity, String) attribute can be assigned restriction=hasProductID (1), and its logic implication is " every commodity have and have only an ID who describes with character string ".
Definition 6: note
Note is the complementarity textual description that notion, attribute or example is carried out in the character string mode.
Layered network knowledge model modeling method of the present invention is followed logical theory described above, process flow diagram as shown in Figure 1:
Step S101 determines the ken of modeling, and this step need be determined the object of knowledge model, i.e. certain subclass of objective world according to the demand of application system.What the present invention will determine is the ecommerce ken, for example ken can be decided to be digital product field, ceramic product field etc.
Step S102 sets up knowledge model according to determined ken, by the thought of layered modeling knowledge model is divided into the modeling of top layer knowledge model, middle level knowledge model modeling and the modeling of bottom knowledge model.
As shown in Figure 2, be top layer knowledge model modeling process flow diagram.Determine notion in the top layer knowledge model, i.e. top layer notion according to selected ken.This step needs by the domain expert deep research to be carried out in the field, take out the top layer notion in this area, top layer knowledge comprises in the ecommerce: key concepts such as product, client, dealer, brand, geography information, these notions are absolutely necessary for e-commerce system, be portrayal middle level ontologies, ensure the basic of intelligent business system operation; The domain expert judges whether to be top layer notion specified limit and note, then direct if desired specified limit and note, good knowledge restriction is set guarantees whole architectonic integrality and logicality, particularly when increasing new knowledge content, ambiguity problem can not be produced, the consistance of knowledge system and objective world can be kept; The attribute of definition top layer notion, these attribute descriptions between the notion or getting in touch between notion and the data, as the top layer concept definition following attribute: purchaseBy (product, the client), semanteme is " purchased ", represented between product and the client it is purchased relation, simultaneously the Attribute domain of attribute and codomain carried out relatedly that described attribute comprises concept attribute and data attribute; The domain expert judges whether to be described attribute specified limit, then directly defined attribute restriction if desired; To verify the top layer knowledge model at last, utilize RACER reasoning from logic instrument that described top layer knowledge model is verified, for the top layer knowledge model that does not satisfy the logical condition requirement, need make amendment makes it can be by the checking of RACER, this step help the very first time find in the knowledge mistake, ambiguity is arranged or violates the content of knowledge restriction.
As shown in Figure 3, be middle level knowledge model modeling process flow diagram.To the top layer concept classification, this step need be carried out refinement to the top layer notion according to the demand of application system, this is mainly reflected in by the domain expert and designs the semantic accurate categorizing system of a cover, in e-commerce system, relate generally to classification, according to the sub-notion of the notion in the class definition top layer knowledge model for product type; The domain expert judges whether to be described sub-notion specified limit and note, then direct if desired specified limit and note; Define the attribute of described sub-notion, simultaneously the Attribute domain of attribute and codomain are carried out relatedly, described attribute comprises sub-concept attribute and data attribute; The domain expert judges whether to be described attribute specified limit, then directly defined attribute restriction if desired; The last mode the same with the top layer knowledge model will be verified the middle level knowledge model.
As shown in Figure 4, be bottom knowledge model modeling process flow diagram.Determine notion or sub-notion that example is affiliated, all must have example (example of sub-notion logically also is the example of upper strata notion) for each notion in top layer knowledge and the middle level knowledge; Specify unique ID for described each example and represent this example, the example modeling, promptly setting up with above-mentioned ID is the example of sign; Judge whether and to specify note for described example, then directly specify note if desired; The attribute of described example institute's categorical conception or sub-notion is carried out instantiation, comprise Attribute domain instantiation and codomain instantiation, promptly according to attribute type, the example and the property value example of concept attribute is associated, the example of data attribute and concrete data are associated; At last the bottom knowledge model is verified that this time checking is that knowledge base is done as a whole checking, utilizes RACER reasoning from logic instrument that the logic of knowledge is tested, verified once more after then solving corresponding problem as authentication failed, up to by checking.
The top layer notion that below is the realistic model Koeb (Knowledge ofElectronic Business) of an ecommerce knowledge Modeling is: commodity, client, dealer, concept attribute, brand and geography information.Wherein, concept attribute is the codomain of all concept attributes in the knowledge model.The mesh topology of knowledge is as follows in the top layer knowledge model:
(1) purchaseBy (product, client), semanteme is " purchased ".
(2) hasSeller (product, dealer), semantic is " having dealer ".
(3) hasBrand (product, brand), semantic is " having brand ".
(4) hasAddress (client/dealer, geography information), semantic is " having the address ".
(5) hasProducePlace (product, geography information), semanteme are " place of production is ".
(6) hasOriginPlace (brand, geography information), semanteme are " the brand cradle is ".
Also comprise following restriction in the top layer knowledge model:
(1) disjoint (product, client, dealer, brand, geography information), semantic is " separating between the top layer notion ".
(2)≤hasBrand(1)
(3)≥hasSeller(1)
The notion that comprises in the knowledge model of middle level of the present invention is the sub-notion of notion in the top layer knowledge model, comprises the classification of commodity, client's classification, the notions such as classification of dealer specifically.Wherein, the sub-notion of attributive concept is: client properties, dealer's attribute, universal product attribute, multiclass product attribute and segmentation product attribute.
(1) client properties is used to describe the character that the client has.
(2) dealer's attribute is used to describe the character that dealer has.
(3) universal product attribute is used to describe the character that all products all have.
(4) the multiclass product attribute is used to describe the character that several series products all have.
(5) the segmentation product attribute be used to describe certain series products exclusive character.
The netted knowledge topological structure of Koeb middle level knowledge model is described for the example of a perfume notion below.Perfume is the sub-notion of cosmetics notion in the commodity, is that the concept attribute of perfume appointment is as follows among the Koeb:
(1) purchaseBy (perfume, client), semanteme is " purchased ".This attribute only just comes into force when order generates.
(2) hasSeller (perfume, dealer), semantic is " having dealer ".
(3) hasBrand (perfume, brand), semantic is " having brand ".
(4) hasProducePlace (perfume, geography information), semanteme are " place of production is ".
(5) hasFitPurpose (perfume is bought purpose), semanteme are " buying purpose is ".Buying purpose is the sub-notion of universal product attribute.
(6) hasFitPeopleType (perfume, customer type), semanteme are " customer type that is fit to is ".Customer type is the sub-notion of client properties.
(7) hasFitHobby (perfume, hobby), semanteme is " hobby that is fit to ".Hobby is the sub-notion of client properties.
(8) hasFitSex (perfume, sex), semanteme are " sex that is fit to ".Sex is the sub-notion of client properties.
(9) hasFitCharacter (perfume, personality), semanteme are " personality that is fit to ".Personality is the sub-notion of client properties.
(10) hasFitJob (perfume, occupation), semanteme are " occupation that is fit to ".Occupation is the sub-notion of client properties.
(11) hasFitTime (perfume, red-letter day), semanteme are " red-letter day that is fit to ".Be the sub-notion of multiclass product attribute red-letter day.
(12) hasPreTaste (perfume, odor type), semanteme are " preceding perfume (or spice) is ".Odor type is the sub-notion of multiclass product attribute.
(13) hasMidTaste (perfume, odor type), semanteme are " middle perfume (or spice) is ".Odor type is the sub-notion of multiclass product attribute.
(14) hasLastTaste (perfume, odor type), semanteme are " back is fragrant to be ".Odor type is the sub-notion of multiclass product attribute.
(15) hasRanking (perfume, commodity rank), semanteme are " the commodity rank is ".The commodity rank is the sub-notion of universal product attribute.
Be that perfume data designated attribute is as follows among the Koeb:
(1) (perfume, String), semanteme is " name of product is " to hasProductName.Its value is a character string data.
(2) (perfume, String), semanteme is " age bracket that is fit to is " to hasFitAge.Its value is a character string data.
(3) (perfume, Date), semanteme is " date of manufacture is " to hasProduceDate.Its value is date type data.
(4) (perfume, Date), semanteme is " guarantee date is " to hasExpDate.Its value is date type data.
(5) (perfume, String), semanteme is " capacity is " to hasVolume.Its value is a character string data.
According to the requirement of layered network knowledge modeling, in fact be exactly with upper strata notion and attribute instanceization for the modeling of bottom knowledge.Provide the knowledge model of a perfume example herein, its product IDs is " FPL010301000000000001 ".Fig. 5 is the description (with OWL2.0 standard provide) of knowledge model for this example.This example has all been carried out detailed instantiation description with the concept attribute and the data attribute of above-described perfume notion.Because the OWL document has high legibility, herein not in the description details that this example is described in detail in detail.
The invention allows for a kind of method of setting up knowledge warehouse system based on above-mentioned knowledge model, as shown in Figure 6, step S601 sets up the knowledge data base group, comprise a large amount of products and customer information in the e-commerce system, these information all will be stored in the knowledge model.The knowledge data base group is the distributed memory system of knowledge, finishes knowledge backup and load balancing task, to improve the knowledge access efficiency of upper layer application.Comprise n platform database server among the knowledge data base group, when example quantity acquired a certain degree, the n value was big more, and the knowledge access efficiency of distributed system is high more.The present invention all disposes top layer knowledge model and the middle level knowledge model in the knowledge model in every database server, the abstractdesription partial redundance that is about in the knowledge model is stored on each database server of knowledge data base group.In e-commerce system, the increase of example is generally from client or dealer.Newly-increased example is recorded in earlier in first database server, each newly-increased example can cause in the database data to increase m capable (example can relate to m bar axiom, every axiom corresponding data line in database.The numerical value of m is not fixed, and the attribute of this example and the many more m of restriction are big more).When wherein example quantity reached u, when promptly the number of data lines in the database was u * m (supposing corresponding m the axiom of each example here), new example information will be recorded in second database server, by that analogy.U is a threshold value, and determining and will deciding on concrete system situation (determinative has server performance, load, network service quality etc.) of its numerical value can increase space expense if numerical value is obtained too small, excessively then can increase time overhead.In addition, also need n platform database server to make dual-host backup respectively among the knowledge data base group at least, recover conveniently to carry out emergency operation and data.
Step S602 sets up non-knowledge data base, not only needs the knowledge data base group in the e-commerce system, also needs the storage of some non-knowledge datas simultaneously.These data will play index and auxiliary effect in the knowledge acquisition process.These non-knowledge datas are stored in the non-knowledge data base, and the most important non-knowledge data of using among the present invention is the knowledge instance routing table.The memory location of this each example of table essential record.The structure of this table is as follows:
| Ins_ID | Ins_Name | Concept | Database_1 | Database_n |
The ID of this table essential record example, the notion under the Chinese of example, example and the database ip address of example knowledge store (and all have the standby host address of identical knowledge content, support of the present invention is transferred to standby host with client activities and responds).When knowledge system increased example newly, system can be the storage specified database server of this example automatically according to knowledge data base group's storage condition.Simultaneously, the relevant information of this example can be recorded in the knowledge instance routing table of non-knowledge data base.Upper layer application is when carrying out knowledge acquisition, and system is retrieval knowledge example routing table at first, finds the memory address of object instance correspondence, carries out knowledge retrieval subsequently on the knowledge data base server of this memory address appointment.In addition, also preserve some other information in the non-knowledge data base, for example, transaction journal, system journal etc.
Step S603 sets up knowledge edition data storehouse, and domain knowledge is constantly to change, and the variation of example is more frequent, but this does not influence the version of knowledge model.If changing of top layer knowledge or middle level knowledge will cause the structural variation of knowledge model.The present invention uses knowledge edition data storehouse independently to describe document for the OWL of top layer knowledge in the model and middle level knowledge, knowledge edition data storehouse uses the SVN version control system to carry out the knowledge Version Control, with knowledge version before the convenient inquiry and rollback version when being necessary.
Particularly, described method also comprises the administration module of creation of knowledge database cluster, non-knowledge data base, knowledge edition data storehouse and backup database respectively.Described method also comprises creates the interface that is used for middleware visit knowledge data base group and non-knowledge data base, and described middleware comprises: JENA semantic net middleware and data management middleware.
As shown in Figure 7, the ecommerce knowledge warehouse system frame diagram of setting up for the method for setting up knowledge warehouse system of the present invention.
The native system framework can be divided into three levels (being) in logic from bottom to top: knowledge data accumulation layer, logic control layer and intelligent use middleware.
1, mainly comprises knowledge data base group, knowledge data base group standby host, data security database and standby host, knowledge version management database and standby host and non-knowledge data base and standby host in the knowledge store layer.Wherein knowledge model is distributed among the knowledge data base group, backs up in knowledge data base group standby host.Knowledge version management database and standby host use the SVN edition management system to carry out version management for the OWL document of abstract knowledge.Main stored knowledge example routing table and other non-more educated system datas in non-knowledge data base and the standby host,
2, the logic control layer mainly is responsible for knowledge acquisition and non-knowledge data and is extracted operation, comprises following system module:
(1) load balancing control module.This module mainly is responsible for judging load transfer condition, i.e. example threshold value.When the example quantity of storing in the knowledge data base server surpasses this threshold value, newly-increased example will be routed to the lower station server of load and store.In addition, this module also is responsible for client activities are routed to the pressure that the database cluster standby host brings to server with the balance client activities from the source database group.
(2) knowledge logical routing module.The main task of this module is when the knowledge acquisition demand of respective client, reads the knowledge instance routing table in real time, finds this to store the concrete server of this example.
(3) knowledge management module.The main knowledge that comes from application of handling changes request and the Version Control service is provided.When the increase example or the upper strata structure of knowledge change in the knowledge model, all need to be undertaken the renewal of knowledge data base, and knowledge version management database is charged in the variation of the structure of knowledge by this module.
(4) data management module.This module mainly is responsible for the non-knowledge data in the update system.
(5) backup management module.This module major control was backed up knowledge data or non-knowledge data in the regular hour.
In addition, the logic control layer also provides following knowledge manipulation interface for Application Middleware:
(1) model persistence interface.JENA semantic net middleware can be saved to the whole axioms in the knowledge model (OWL document) among the knowledge data base group by this interface.
(2) knowledge acquisition interface.JENA semantic net middleware can extract knowledge data by this interface.
(3) information management interface.JENA semantic net middleware can carry out version management to knowledge data by this interface.
(4) data management interface.The data management middleware can manage non-knowledge data by this interface.
As long as 3, the intelligent use middleware layer comprises JENA semantic net middleware and two assemblies of data management middleware.JENA semantic net middleware is responsible for the knowledge acquisition request that the response application logic proposes, and carries out operations such as knowledge query and management.The data management middleware is mainly used in the non-knowledge data of management.
The ecommerce knowledge warehouse system that the method according to this invention is set up experimentizes, and its experiment condition and result are as follows:
The method that adopts distributed multimachine to be equipped with is mutually carried out the knowledge model storage, uses 10 DELL R300 servers (CPU: to strong 4 nuclears 3323, RAM:4G) altogether.Wherein 5 is the knowledge base distributed storage server, and softwares such as MYSQL2.0, JENA have been installed on it, and 5 station servers are standby host in addition.Fig. 8 provides the performance data of example of the present invention.As shown in Figure 8, the horizontal ordinate in the broken line graph is represented example quantity, and ordinate represents to retrieve the time cost of an example, and n represents the number of servers of example distributed store.In the performance test, each example comprises 5 axioms, and the experimental result among the figure is the mean value of following 100 test results of specified conditions.Wherein n is 1 o'clock, and the time cost of the axiom of an example of retrieval is 38.87ms-1045.78ms, and n is 5 o'clock, and the time cost of the axiom of an example of retrieval is 10.12ms-604.12ms.By experimental result as seen, provide the database server of knowledge distributed storage service many more, example recall precision is high more.Use 5 database servers to carry out the data retrieval efficiency requirements that knowledge store has enough satisfied e-commerce system among the present invention.
Above embodiment only is used to illustrate the present invention; and be not limitation of the present invention; the those of ordinary skill in relevant technologies field; under the situation that does not break away from the spirit and scope of the present invention; can also make various variations and modification; therefore all technical schemes that are equal to also belong to category of the present invention, and scope of patent protection of the present invention should be defined by the claims.
Claims (9)
1. a layered network knowledge model modeling method is characterized in that, may further comprise the steps:
S1: the ken of determining to want modeling;
S2: described ken is set up knowledge model by following formula, and described knowledge model KM comprises top layer knowledge model u (k), middle level knowledge model m (k) and bottom knowledge model l (k), and D represents ken,
KM＝f(D，u(k)∪m(k)∪l(k))
Described u (k) comprises the notion among the D, the sub-notion among m (k) the definition u (k), and l (k) comprises the example of described notion and sub-notion;
S3: utilize RACER reasoning from logic instrument that described knowledge model is verified, for the knowledge model that does not satisfy the logical condition requirement, need make amendment makes it can be by the checking of RACER.
2. layered network knowledge model modeling method as claimed in claim 1 is characterized in that, top layer knowledge model u (k) modeling comprises step among the described step S2:
Determine notion in the top layer knowledge model according to ken;
Judge whether to be notion specified limit and the note in the top layer knowledge model then direct if desired specified limit and note;
The attribute of the notion in the definition top layer knowledge model carries out the Attribute domain of attribute and codomain relatedly simultaneously, and described attribute comprises concept attribute and data attribute;
Judge whether to be described attribute specified limit, if desired then directly defined attribute restriction;
Utilize RACER reasoning from logic instrument that described top layer knowledge model is verified, for the top layer knowledge model that does not satisfy the logical condition requirement, need make amendment makes it can be by the checking of RACER.
3. layered network knowledge model modeling method as claimed in claim 2 is characterized in that, middle level knowledge model m (k) modeling comprises step among the described step S2:
The sub-notion of the notion in the definition top layer knowledge model;
Judge whether to be described sub-notion specified limit and note then direct if desired specified limit and note;
Define the attribute of described sub-notion, simultaneously the Attribute domain of attribute and codomain are carried out relatedly, described attribute comprises sub-concept attribute and data attribute;
Judge whether to be described attribute specified limit, if desired then directly defined attribute restriction;
Utilize RACER reasoning from logic instrument that described middle level knowledge model is verified, for the middle level knowledge model that does not satisfy the logical condition requirement, need make amendment makes it can be by the checking of RACER.
4. layered network knowledge model modeling method as claimed in claim 3 is characterized in that, end top layer knowledge model l (k) modeling comprises step among the described step S2:
Determine notion or sub-notion that example is affiliated;
Specify unique ID for described each example and represent this example;
Judge whether and to specify note for described example, then directly specify note if desired;
The attribute of described example institute's categorical conception or sub-notion is carried out instantiation, comprise Attribute domain instantiation and codomain instantiation;
Utilize RACER reasoning from logic instrument that described bottom knowledge model is verified, for the bottom knowledge model that does not satisfy the logical condition requirement, need make amendment makes it can be by the checking of RACER.
5. set up the method for knowledge warehouse system based on the arbitrary described knowledge model of claim 1-4 for one kind, it is characterized in that, may further comprise the steps:
Set up the knowledge data base group, described knowledge data base group comprises n platform knowledge data base server, is used for storing the notion of top layer knowledge model and middle level knowledge model and described notion corresponding example in the bottom knowledge model, and wherein n is a positive integer;
Set up non-knowledge data base, creation of knowledge example routing table in described non-knowledge data base is used for writing down the IP address of each example at knowledge data base group's knowledge data base server;
Set up knowledge edition data storehouse, described knowledge edition data storehouse be used for according to top layer knowledge model and middle level knowledge model variation control described data warehouse version.
6. the method for setting up knowledge warehouse system as claimed in claim 5, it is characterized in that, described foundation comprises the step that every knowledge data base server stores example amount threshold is set in the knowledge data base group step, when the storage of first station server reaches threshold value, then store newly-increased example into second station server.
7. the method for setting up knowledge warehouse system as claimed in claim 5 is characterized in that, described method also comprises sets up backup database, is used for described knowledge data base group's data are backed up.
8. the method for setting up knowledge warehouse system as claimed in claim 6 is characterized in that, described method also comprises the administration module of creation of knowledge database cluster, non-knowledge data base, knowledge edition data storehouse and backup database respectively.
9. the method for setting up knowledge warehouse system as claimed in claim 6, it is characterized in that, described method also comprises creates the interface that is used for middleware visit knowledge data base group and non-knowledge data base, and described middleware comprises: JENA semantic net middleware and data management middleware.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201010113228A CN101799822A (en) | 2010-02-23 | 2010-02-23 | Method for modeling layered network knowledge model and method for establishing knowledge warehouse system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201010113228A CN101799822A (en) | 2010-02-23 | 2010-02-23 | Method for modeling layered network knowledge model and method for establishing knowledge warehouse system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN101799822A true CN101799822A (en) | 2010-08-11 |
Family
ID=42595500
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201010113228A Pending CN101799822A (en) | 2010-02-23 | 2010-02-23 | Method for modeling layered network knowledge model and method for establishing knowledge warehouse system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101799822A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104063429A (en) * | 2014-06-11 | 2014-09-24 | 深圳德协保税电子商务有限公司 | Predicting method for user behavior in e-commerce |
| CN105183817A (en) * | 2015-08-27 | 2015-12-23 | 北京时代焦点国际教育咨询有限责任公司 | Question bank modeling method and system based on domain-driven design |
| CN105550360A (en) * | 2015-12-31 | 2016-05-04 | 上海智臻智能网络科技股份有限公司 | Method and apparatus for optimizing abstract semantic library |
| CN105844335A (en) * | 2015-01-15 | 2016-08-10 | 克拉玛依红有软件有限责任公司 | Self-learning method based on 6W knowledge representation |
| CN107688600A (en) * | 2017-07-12 | 2018-02-13 | 百度在线网络技术（北京）有限公司 | Knowledge point method for digging and device |
-
2010
- 2010-02-23 CN CN201010113228A patent/CN101799822A/en active Pending
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104063429A (en) * | 2014-06-11 | 2014-09-24 | 深圳德协保税电子商务有限公司 | Predicting method for user behavior in e-commerce |
| CN105844335A (en) * | 2015-01-15 | 2016-08-10 | 克拉玛依红有软件有限责任公司 | Self-learning method based on 6W knowledge representation |
| CN105844335B (en) * | 2015-01-15 | 2018-04-20 | 克拉玛依红有软件有限责任公司 | A kind of self-learning method based on the 6W representations of knowledge |
| CN105183817A (en) * | 2015-08-27 | 2015-12-23 | 北京时代焦点国际教育咨询有限责任公司 | Question bank modeling method and system based on domain-driven design |
| CN105550360A (en) * | 2015-12-31 | 2016-05-04 | 上海智臻智能网络科技股份有限公司 | Method and apparatus for optimizing abstract semantic library |
| CN105550360B (en) * | 2015-12-31 | 2018-09-04 | 上海智臻智能网络科技股份有限公司 | Optimize the method and device in abstract semantics library |
| CN107688600A (en) * | 2017-07-12 | 2018-02-13 | 百度在线网络技术（北京）有限公司 | Knowledge point method for digging and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11327935B2 (en) | Intelligent data quality | |
| JP5598017B2 (en) | Judgment program, method and apparatus | |
| US8209284B1 (en) | Data propagation in a multi-shard database system | |
| US7886028B2 (en) | Method and system for system migration | |
| CN108388604A (en) | User right data administrator, method and computer readable storage medium | |
| US8108360B2 (en) | Database object update order determination | |
| CN106600302A (en) | Hadoop-based commodity recommendation system | |
| CN111858615B (en) | Database table generation method, system, computer system and readable storage medium | |
| US20120246110A1 (en) | Master Data Management in a Data Warehouse/Data Mart | |
| US11237749B2 (en) | System and method for backup data discrimination | |
| US10812551B1 (en) | Dynamic detection of data correlations based on realtime data | |
| CN102033938A (en) | Secondary mapping-based cluster dynamic expansion method | |
| US20080086478A1 (en) | Semantical partitioning of data | |
| US20150254577A1 (en) | System and methods for location based management of cloud platform data | |
| US8751543B2 (en) | Database view modeling using existing data model | |
| US9275125B1 (en) | System for organizing data from a plurality of users to create individual user profiles | |
| Gudivada et al. | Renaissance in database management: navigating the landscape of candidate systems | |
| CA3015566A1 (en) | Systems and methods for expediting rule-based data processing | |
| CN101799822A (en) | Method for modeling layered network knowledge model and method for establishing knowledge warehouse system | |
| CN112163024B (en) | Configuration information export and import method based on hierarchical association structure | |
| Constantinov et al. | Performing real-time social recommendations on a highly-available graph database cluster | |
| US10331627B2 (en) | Method and system for unified technological stack management for relational databases | |
| CN110515923B (en) | Data migration method and system between distributed databases | |
| CN102193975A (en) | Method for realizing transaction committing mechanism in online transaction of graphic database | |
| WO2016203479A1 (en) | A system and method providing database manipulation operations and management |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Open date: 20100811 |