Home

APPROVAL SHEET

1. eee Drow MsStam based Dear o ais bara a ale we eb ee t 5 3 3 WordNet based Search 0 0 200008 5 3 4 Synonym based Search Exhaustive 5 3 5 Synonym based Search Greedy 205 5 3 6 Experiments and Results 6 XDF Output XML Instance 6 1 XML Instance Generation 10 11 13 15 16 16 18 20 22 23 24 29 90 32 32 34 30 30 36 37 40 41 43 AY 111 T Experiment on Supplier Discovery el Desion or Experiment sad ee eed Sok oe he KER ee a 1 2 Experimental Results 8 Conclusions and Future Works Sel CORCIU ONS aa a e A A de de de A Oso SHUTS WOKS e a A amp eR He amp a ee eee A User Manual A 1 Basic Form Interface 0 2 00 000000845 A 1 1 Navigation Interface e urraca Ae eo AZ Comain NOE d 64 4 oe de A Sh hd GE e S Alia Data Mpat Node edad E he She A 1 4 Data Input Node List Type A 1 5 Form Component Remove Button LLO ELO OSION y ace a a ca ar BAH A 1 6 1 Search Existing Form Components A 1 6 2 Define New Form Components A2 Basiclnio Form Page ars ok cares Sn a OR ee Be we A 2 1 BasicInfo Nodes s soe sit r ane 000000000000 A3 Service Form Page s sm wo aona a Oe ew a we eee A A 3 1 Service Nodes ici ad o da a ee wey AA EDE e w Bote Ge Bh we OE we or eR A See ed Ded wl A 4 1 Raw Data of Supplier Capability Profiles
2. cally viewed as labeled trees Each node of these trees represents a data element or an attribute by a label of English word or concatenation of words or their abbrevia tions Although XML is not a formal semantic model its structure and the English words for the labels contain rich semantic information Many XML matching ap proaches have been proposed 18 19 most of which analyze the similarity between these labeled tress based on their semantic and structural information Applications 20 21 have been developed to generate XML schema 7 based web forms to capture users information and produce XML instances as output However these applications support limited XML schema features And the web forms they generated are fixed Rein 22 proposed an application of dynamically generating a web form based on XML schema and producing XML instance as output It supports more XML schema features and allows users to dynamically add and remove items from the base form However Rein did not take into account the extensibility which is crucial in the information collection process Lacking of extensibility web form may not be able to capture the information that is specific to users domain Therefore we need more flexible and dynamic architecture to collect require ments and capabilities in a better structured way to support meaningful semantic analysis Chapter 3 XDF Overview This chapter describes logic flow of XDF system and
3. Library amp ee as evolving importing A 24 a existing form components a e A 2 2 create new form components 3 Generate XML j Customer requests Supplier profile instances i h XML instance XML instances 4 Store XML instances to repository E SS e A SA 16 CNC lt Mach P Y y 5 Discover Suppliers by ES ooo A n Atacan Eldorado Gun Driller lt Ya ar XML Matching e on ae From gt lt T0 gt 1 250 lt T0 gt Sa y dr cie an ed a an s sc sl a A unDrilingCape e a Figure 3 1 Logic Flow of XDF schemas in the repository At the beginning all the XML schemas including domain ontology core component and pre defined manufacturing schemas in the repository are parsed and then transformed into form components T hese form components will be rendered as form components that build the base form Step two Users input their data in this base form When the base form can not cover users information i e user can not find a place to input his her data users can extend the base form by searching existing form components and plugging them into the base form An intelligent search engine is provided for users to search existing form components This intelligent search engine combines the benefits of n gram based searching and WordNet based searching It tolerates typo and can find semantically similar words or phrases Because different users may describe their capabilities using
4. gt lt xs sequence gt lt xs complexType gt lt xs element gt lt xs group name EquipmentType gt lt xs choice gt lt xs element name ProductFinishingEquipment type T EquipmentGroup minOccurs 0 gt lt xs element name GrindingAndHoningEquipment type T EquipmentGroup minOccurs 0 gt lt xs element name ManualMillingEquipments type T EquipmentGroup minOccurs 0 gt lt xs element name ManualTurningEquipments type T_EquipmentGroup minOccurs 0 gt lt xs element name SawingSandingEquipments type T EquipmentGroup minOccurs 0 gt lt xs element name SheetMetalWeldingEquipment type T EquipmentGroup minOccurs 0 gt lt xs element name DrillingTappingEquipment type T EquipmentGroup minOccurs 0 gt lt xs element name ScrewMachineProducts type T EquipmentGroup minOccurs 0 gt lt xs element name GunDrilling type T EquipmentGroup minOccurs 0 gt lt xs element ref WeldingMachines minOccurs 0 gt lt xs element ref MachiningCenters minOccurs 0 gt lt xs element ref MillingEquipment minOccurs 0 gt lt xs choice gt lt xs group gt Y DrillingEquipment EquipmentName Y group EquipmentType please select one item Add GunDrilling DrillingTappingEquipment GrindingAndHoningEquipment GunDrilling MachiningCenters ManualMillingEquipments ManualTurningEquipments MillingEquipment ProductFinishingEquipment SawingSandingEquipments ScrewMac
5. lt hasIndustryFocus gt lt hasCertification gt lt ISO9001 gt lt hasCertification gt lt hasCertification gt lt UserDefinedCertification gt AS9100B lt UserDefinedCertification gt lt hasCertification gt lt Extension gt lt AdditioinalCapability gt Largest user of Omax Waterjets in North America lt AdditioinalCapability gt lt AdditioinalCapability gt 5 axis machining capability lt AdditioinalCapability gt lt AdditioinalCapability gt Six Sigma Black Belt Quality Manager lt AdditioinalCapability gt lt AdditioinalCapability gt Full time tooling engineer on staff lt AdditioinalCapability gt lt AdditioinalCapability gt In house Magnetic particle inspections lt AdditioinalCapability gt lt AdditioinalCapability gt Machining experience with composite and exotic materials lt AdditioinalCapability gt lt AdditioinalCapability gt Design support and engineering services lt AdditioinalCapability gt lt AdditioinalCapability gt Assembly and Kitting lt AdditioinalCapability gt lt Extension gt lt Supplier gt V lt Service name gt lt name gt Vertical CNC Milling lt name gt lt SpecilityGroup gt lt Specility gt Short run production lt Specility gt lt Specility gt Quantitive production lt Specility gt lt Extension gt lt Prototype gt prototype production lt Prototype gt lt SpecilityGroup gt lt FacilityGroup gt lt EquipmentGroup gt lt Addon gt lt Extension gt lt ManualMil
6. 2 2 A 4 2 BasicInfo Page 0 a Bue a ec A Oe we eee RAS Service Pace Se bu Sew GO AEA DAES Se OOO AAA AME Instance cees p 6 40 09 be RE ee AE ww A OS Bibliography AQ AQ ol 59 59 54 313 50 50 IS IS 59 59 59 60 62 69 66 67 68 12 2 13 14 70 19 4 1 9 1 9 2 5 9 1 1 3 1 3 2 3 9 3 4 4 1 4 2 4 3 4 4 4 0 4 6 Al 4 8 4 9 4 10 4 11 4 12 9 1 9 2 9 9 9 4 5 9 6 1 List of Tables Corresponding Relationship Between Wrapper Classes Schema Java Classes and Schema Constructs e Comparison of Search Methods with Synonymous Version of Search Key WOL ao e a Hh ae A AAA A Comparison of Search Methods with Typo Version of Search Keywords Comparison of Computing Time 0 Performances of Search Engines 0 0 0000088 List of Figures Logia FlOW wor ADE x es eee has es a e ee SO ee BOE ee Architecture ol XDE sees ke ra en ROG ain AO OSE a Architecture of Business Logic Layer 0 Architecture of Presentation Layer 0 Form Components Generation Flow Class Diagram of Schema Library 0 Wrapper Classes ms he e e A ss de Baw e e bw ELA Class Diagram of Wrapper Classes Simple Type Form Component with multiple occurrence Simple Type Form Component One Occurrence Complex Type Form Component with Simple Content Complex T
7. but allows element to carry attributes L7 e complexContent allows the content of a element to contain other elements or groups It also allows element to carry attributes e any specifies that any well formed XML is permissible in a content model It is also called any element This any element will be transformed to a Wildcard form component to allow users to extend the base form This transformation will be covered in Chapter 5 e group represents a group of schema constructs which can be element any element or group group can be only contained in complexContent 4 1 2 Schema Library Before generating form components XDF system transforms all schema con structs into a middleware called Schema Library Schema Library is a Java Class the instance of which contains a collection of Java objects each of which represents a schema construct explained in section 4 1 1 Figure 4 2 illustrates the class diagram of Schema Library We call classes in Figure 4 2 schema Java classes the instance of which are called schema Java objects Each schema Java object represents a XAML schema construct and as illus trated in Figure 4 2 the class diagram of schema Java objects maintains the same structure as that of schema constructs specified by XML schema Language That is every XMLSchema object contains multiple Element objects each of which has either a Simple Type object or a ComplexType object A Simple Type object may con ta
8. structure diagram as that of schema Java classes Combining the corresponding relationship between schema Java classes and schema constructs we obtain the cor responding relationship between wrapper classes schema Java classes and schema constructs presented in Table 4 1 Table 4 1 Classes and Schema Constructs Wrapper Class Element Wrapper Simple Type Wrapper RestrictionSimple Type Wrapper ListSimpleTypeWrapper UnionSimpleTypeWrapper Complex TypeWrapper SimpleContent Wrapper ComplexContent Wrapper ModelGroup Wrapper Group Wrapper Wildcard Wrapper Corresponding Relationship Between Wrapper Schema Java Class Element SimpleType RestrictionSimple Type ListSimpleType UnionSimpleType Complex Type SimpleContent ComplexContent ModelGroup Group Wildcard Classes Schema Java Schema Constructs elemen simple Type restriction list union complex Type simpleContent complerContent group any ModelGroup corresponds to un named group in XML schema language It groups elements so that these elements can be used to build the content model 21 of complex type A group of elements can be constrained to appear in the same order as they are declared by sequence schema construct Alternatively they can be constrained by choice construct so that only one of these elements may appear The third option for constraining elements in a model group is to permit all elements in the group appear
9. 2x Lle N L k Simle k E e E k 5 1 For instance in order to compute the similarity between string ShipAddress and string ShipTo we first obtain two word lists Ship Address and Ship To The two string have one shared word which is Ship Therefore the similarity score is 2 x 1 24 2 0 5 The search result is a list of form components ranked by their similarities by 5 1 The keyword based search is simple and fast but it does not utilize the semantics of words for synonyms or semantically similar words In addition it is unable to match with words in variation forms e g capabilities vs capability or containing typos 5 3 2 n gram based Search The similarity between two strings can be also measured by counting the number of the occurrences of different n grams 28 29 i e the substrings of length n in the two strings The more similar the strings are the more n gram they will have in common T he similarity can be defined as _ 2x Iden Ngram e Ngram k e Ngram e Ngram k 5 2 where Ngram e and Ngram k are the sets of items in the n grams of e and k and Iden Ngram e Ngram k is the number of n grams shared by e and k The denominator indicates the total number of n grams in the two N gram sets 36 n grams can be used with various length For experiments we use trigram n 3 Take the word shipment as an example bi grams 7s s
10. 33 A WordNet Available at http wordnet princeton edu 34 A D Lin An Information theoretic definition of similarity In Proc of the 15th International Conference on Machine Learning 1998 pp 296 304 135 E Ukkonen Approximate string matching with q grams and maximal matches Theoretical Computer Science 1992 pp 191 211 36 F Ameri and D Dutta An upper ontology for manufacturing service descrip tion ASME 2006 International Design Engineering Technical Conferences amp Computers and Information in Engineering Conference Philadelphia Septem ber 10 13 2006 37 L Rijsbergen C V Van Rijsbergen Information Retrieval 2nd Edition Butter worth London Boston 1979 Sl 38 J Shin N Ivizic J Kim F Ameri C McArthur S De Flitch and T Scac chitti An experimental evaluation platform for state of the art manufacturing supplier discovery methods Electronic Commerce Research and Applications 82
11. 92 when all 16 key words were used This demonstrates that the extensibility of XDF allows us to collect additional information beyond what is defined in the ontology leading to a more accurate discovery Additional exper iments with different customer requirements and different human experts showed similar results 52 Chapter 8 Conclusions and Future Works 8 1 Conclusions The objective of the research reported in this thesis is to design and implement a form architecture that is flexible enough to capture a variety of customer require ments and supplier capabilities and generate XML instances from user inputted data These XML instances can be analyzed by advanced XML based matching algorithm to match suppliers and customers in the supplier discover process We implemented eXtensible Dynamic Form XDF to facilitate the process of collecting manufacturing capability information of suppliers and requirements of customers XDF is constructed by a collection of form components that are defined by several XML schemas including domain ontology schemas core com ponent pre defined manufacturing schemas and user defined schema XDF allows users to extend the base form by creating their own form components Thus it helps to better capture users domain specific information XDF also provides users with a synonym based search engine to search existing form components Users can dynamically add existing form components and ne
12. Schema language is a complex maze of constructs that overlap each other Completely covering features of XML schema is a tremendous amount of work and out of the scope of this thesis Instead we cover the most used schema constructs of XML schema 7 including 16 element is either complex Type or simpleType If it is complexType it can contain sub elements and carry attributes If it is simple Type it can only contain build in simple types 7 such as boolean integer date and string or derivation version of build in simple types sumpleType only allows its corresponding element to contain build in simple types attribute provides additional information on its corresponding element A attribute can only be simple Type restriction type allows deriving a new simple Type by restricting an existing simple Type list type is comprised of sequences of build in simple types and consequently the parts of a sequence themselves are meaningful union type enables the value of an element or attribute be one or more in stances of one type drawn from the union of multiple build in types and list types Build in type list type and the union type described above are collec tively called simple types complex Type allows its corresponding element to contain sub elements and carry attributes Its content model is either simpleContent or complexContent simpleContent restricts the content of a element to simple type data
13. XDF architecture XDF starts with a base form generated based on several XML schemas It allows users suppliers or customers to extend the base form of XDF by searching the existing form components Also users can freely create their own form components by using their domain specific concepts structures and semantics and then add them to the base form These user created form components will be stored as user defined schemas and they can be searched and reused by other users or form components later User inputed data on the eXtensible Dynamic Form will be automatically transformed the into XML instances which can be analyzed by advanced supplier discovery methods 3 1 Logic Flow of XDF Figure 3 1 illustrates the logic flow of XDF system The logic flow of XDF consists of four steps step 1 2 3 4 The algorithms for semantics based XML instance matching for the last step discovering suppliers will not be covered in this thesis However we will in Chapter 8 present the experimental results on comparison of three supplier discover methods Step one The base form is automatically generated by XDF system from segenes Domain ontology schema Pre defined 1 Generate a 2 Collect supplier Schema form base form a pan PP a capability profiles by component j P oc e dynamically extending the form 3 AAA schema STD Core components schema User defined schema
14. gt lt xs element gt y Y Address street city state zipCode Figure 4 8 a form component transformed from a Address element Figure 4 8 illustrates the situation where a form component is transformed from a element with complexT ype the content type of which is complexContent The sub elements of Address element are constrained by sequence construct The resulting Address form component contains four sub form components street city state and zipCode each of which is transformed from a sub element of Address element The order in which the four sub form components appear in 26 the Address form component is the same as the order their corresponding sub elements were declared 2 When sub elements of the wrapped element is constrained by choice con struct ModelGroup Wrapper will put all sub form components into a list box Only one sub form component can be chosen from the list box and rendered as the child of the super form component lt xs element name hasCertification gt lt xs complexType gt lt xs choice gt lt xs element ref ASs9000 gt lt xs element ref ASs9100 gt lt xs element ref ISO9002 gt lt xs element ref ISO9001 gt lt xs element ref ISO14000 gt lt xs element ref MIL gt lt xs element ref UserbDefinedcertification gt lt xs choice gt lt xs complexType gt lt xs element gt Y hasCertification
15. of 902 form components with the top 10 highest similarity scores If the sample form component exists in a search result we call it a hit Hit rate is the ratio of the number of hits to the number of search queries i e 8 We compare the average hit rates of search for the eight samples Table 5 1 shows the comparison results of the four search methods For search results of the synonymous version of search keys WordNet based and synonym based both exhaustive and greedy methods have the best hit rates i e 100 This is to be expected because they both utilize semantics of word synonyms On the other hand the search results of the typo version show n gram based and synonym based methods are the better i e 87 5 hit rate This is because WordNet based and keyword based search cannot deal with the similarity of typo words whereas n gram based method can compute the similarity between 43 characters of words by edit distance 35 Overall the synonym based search method has the best performance in terms of hit rate and average computing time Table 5 1 Comparison of Search Methods with Synonymous Version of Search Keywords Synonymous Version of Search Keywords Matching Method Hit Rate Average Computing Time millisec Keyword based 0 44 N gram based 25 45 WordNet based 100 20216 Synonym based exhaustive 100 2339 Synonym based greedy 100 2620 Table 5 2 Comparison of Search Methods with Typo Version of Searc
16. of the requirements for the degree of Master of Science 2011 Advisory Committee Professor Yun Peng Chair Advisor Professor Charles Nicholas Professor Yelena Yesha Copyright by Yan Kang 2011 Table of Contents List of Tables List of Figures 1 Introduction 1 1 Challenges in Supplier Discovery 1 2 Contributions of the Thesis otras eae RA Ra A 1 3 Organization of the Thesis es oe eye dd A di oe 2 Background and Related Works 3 XDF Overview del bogie Blow Or DP sica AO Dr 6 2 ER 4 amp ee 4 Gea ne are he ee eA ES OZ Data layer sitas etree oe ee oS aS Bob BH o22 business Losi laver se s eoi sorp eea ee eae ee Gere 3 2 3 Presentation Layer 00 4 XDF Generation 4 1 Form Components Generation 0 0 000004 AIT AME Schema sax ae eea ay hs Bh ae He ee ee a ee eww ow d AN 2 Enea MIDES w oi Be bd SE Ow ee 4 1 3 Wrapper Classes 4 6 6 42 a6 bed od ea wd od eae fea os Asa Fomm Components sos e a e io EA eS Ee Gro ee a 4 1 4 1 Form Component with Simple Type 4 1 4 2 Form Component with Complex Type 4 1 4 3 Wildcard Form Component 4 2 Base Form Generation 5 XDF Extension Dell SOVVCTICW a ie deo heist wee Bs san ee BS aan lov dpe ce ee he he he es 5 2 User defined Form Component 0 0 00040 5 3 Form Component Searching Methods 5 3 1 Keyword based Search
17. the number of search keywords and M is the average number of synonyms for each search keyword Figure 5 5 represents the same result as that of Table 5 3 This experiment demonstrates that greedy synonym based search method has the best performance in terms of computing time Table 5 3 Comparison of Computing Time Synonym based exhaustive Synonym based greedy WordNet based Two words 2 5 2 6 20 2 Three words 32 2 of 21 1 Four words 210 8 A 23 8 45 250 200 150 100 50 synonym based exhaustive synonym based greedy WordNet based two words three words four words Figure 5 5 Comparison of Computing Time 46 Chapter 6 XDF Output XML Instance XDF will output user inputted data as XML instances which can be either supplier capabilities XML instance or customer requirements XML instance These XML instances will be taken as input for the process of supplier discovery 6 1 XML Instance Generation As explained in Chapter 4 XDF maintains the structures and constraints defined in XML schemas In other words XDF is a web form representation of XML schema which provides a convenient way for users inputting data that conform to rules defined in XML schema Therefore the XML instances generated from user inputted data adhere to XML schema automatically Figure 6 1 illustrates an example of how a form component is encoded as a XML instance In this example the for
18. 34 defined form component is simple type it can only contain value of simple types e g strings and integers If it is complexType it can add arbitrary number of other form components as its sub structure through the ways as mentioned in section 5 1 In the case of Figure 5 2 the user defined a form component that is simpleType Therefore an input text box was created for user inputting simple type values 5 3 Form Component Searching Methods To extend the base form users can search the existing form components from Schema Library XDF search engine finds form components whose names are closely matched with the search keyword For the best quality and performance of search we have investigated several search methods keyword based n gram based WordNet based and synonym based methods In this section we describe and com pare these four different search methods We denote search keyword as string k and the name of form component to be compared in the search as string e 5 3 1 Keyword based Search Keyword based search is the simplest way to find the existing form compo nents It is based on exact string matching method 27 We assume that the two string k and e are either single word or concatenated words These String are tok enized to two sets of words denoted as L k and L e respectively For example if k ShipAddress then L k Ship Address The similarity between k and e is defined as 30
19. 4 7 shows a form component that is transformed from a complexType element BuildingMaterial The content type of BuildingMaterial element is sumpleContent that carries an attribute named ID The simpleContent Wrapper rendered this ID attribute as a text box in the BuildingMaterial form compo nent When the content type of the wrapped element is complerContentType the wrapped element may contain sub elements which indicates that the form compo nent to be rendered we call it super form component will contain one or more other form components we call them sub form components as its sub structure In this case ModelGroup Wrapper will be delegated to render these sub form components 29 There are three different situations when ModelGroup Wrapper rendering sub form components 1 When sub elements of the wrapped element is constrained by sequence construct ModelGroup Wrapper will render sub form components in the same order as these sub elements were declared lt xs element name Address gt lt xs complexType gt lt xs complexContent gt lt xs restriction base xs anyType gt lt xs sequence gt lt xs element name street type xs short gt lt xs element name city type xs string gt lt xs element name state type xs string gt lt xs element name zipcode type xs string gt lt xs sequence gt lt xs restriction gt lt xs complexContent gt lt xs complexType
20. A 16 CNC x m 1 Extended Item y i gt sgt drup i FaciiityGroup S gt hasMaterial List l P me le Interfaces for extending the base form Figure 4 12 Base Form 30 As illustrated in Figure 4 12 The base form of XDF is generated from a collection of base form components The user inputted a Drilling service and its related data The user also extended the base form by adding an user defined form component named Material ol Chapter 5 XDF Extension 5 1 Overview XDF allows users to extend the base form of XDF by either searching the existing form components or creating their own form components When the base form can not cover the users information i e user can not find a place to input his her data the user can extend the base form by searching existing form compo nents from Schema Library and plugging them into the base form The searching function of XDF is performed by an intelligent search engine which combines the benefits of N gram based searching and WordNet based searching It tolerates typo and word variations and can match semantically similar words or phrases Because different users may describe their capabilities using different terms structures and semantics users may not be able to find a appropriate existing form component to fit their data When that happens users can freely create their own form components and then add them to the base form Each form component incl
21. APPROVAL SHEET Title of Thesis EXTENSIBLE DYNAMIC FORM FOR SUPPLIER DISCOVERY Name of Candidate Yan Kang Master of Science Computer Science 2011 Thesis and Abstract Approved Dr Yun Peng Professor Department of Computer Science and Electrical Engineering Date Approved Curriculum Vitae Name Yan Kang Permanent Address 4816 Grand Bend Drive Degree and date to be conferred Master of Science in Computer Science May 2011 Place of Birth Chongqing China Secondary Education Chongqing Number One Middle School Chongqing China Collegiate institutions attended University of Maryland Baltimore County M S Computer Science 2011 Chongqing Technology and Business University B S Computer Science 2007 Major Computer Science Professional publications Industry Oriented Bank Risk Early Warning Evaluation Based on Self Adaptive RBFNN and Uniform Design Method Yan Kang Shi Ying Kang proceedings of 2008 International Conference of Risk Management and Engi neering Management ICRMEM 2008 pp121 125 2008 Professional positions held Software Designer Chongqing New Century electric Co Ltd Dec 2008 May 2009 Lecturer Chongqing ZhengDa Software Polytechnic College Sept 2007 Oct 2008 ABSTRACT Title of Thesis EXTENSIBLE DYNAMIC FORM FOR SUPPLIER DISCOVERY Yan Kang Master of Science Computer Science 2011 Thesis directed by Dr Yun Peng Professor Department of Computer Sc
22. UserDefinedCertification Add UserDefinedCertification please select one item AS9000 AS9100 115014000 1509001 509002 MIL UserDefinedCertification UserDefinedCertification Figure 4 9 a form component transformed from a hasCerfitication element Figure 4 9 illustrates the situation where a form component is transformed from a element with complexType the content type of which is complexContent The sub elements of hasCertification element are constrained by choice construct The resulting hasCertification form component contains a list of sub form compo nents Only one sub form component can be chosen from the list and added to the hasCertification form component In this case UserDefinedCertification sub 2 form component was chosen form the list 3 The third situation is when the wrapped element contains a group The group can be either named group or un named group Either way a Group Wrapper will be delegated to render the group as form components Group Wrapper renders form components the same way as ModelGroup Wrapper does except that the ren dered form components have a group name Figure 4 10 illustrates that a group was rendered as a group EquipmentType form component lt xs element name DrillingEquipment gt lt xs complexType gt lt xs sequence gt lt xs element name EquipmentName type xs string gt lt xs group ref EquipmentType
23. W3 org Resource Description Framework RDF Concepts and Abstract Syn tax Available at http www w3 org TR 2004 REC rdf concepts 20040210 W3 org OWL Web Ontology Language Guide Available at http www w3 org T R 2004 REC owl guide 20040210 B Rensmann H Weigand Z Zhao V Dignum F Dignum and M Hiel As sessing the value of mediators in collaborative business networks Establishing the Foundation of Collaborative Networks Springer pp 155 162 2007 S Colucci T D Noia E D Sciascio F M Donini M Mongiello Concept abduction and contraction for semantic based discovery of matches and nego tiation spaces in an e marketplace Electronic Commerce Research and Appli cations vol 4 Summer 2005 pp 345 361 M Grieger Electronic marketplaces A literature review and a call for supply chain management research European Journal of Operational Research vol 144 2003 pp 280 294 Aikins Janice R Brooks W Clancey et al 1981 Natural Language Process ing Systems In the Handbook of Artificial Intelligence Vol I ed Barr Avron T9 14 15 A 16 o 17 18 A 119 A and Edward A Feigen baum pp 283 321 Stanford Los Altos CA HeurisTech Press William Kaufmann Inc R Baeza Yates Challenges in the Interaction of Information Retrieval and Natural Language Processing Computational Linguistics and Intelligent Text Processing In Comp
24. Y Extension e Extended Item Figure 4 11 A Wildcard form component As illustrated in Figure 4 11 Wildcard form component is transformed from an Extended Item element that takes an any element as its sub element This 29 Wildcard form component is rendered as an Extended Item button Once user clicks the Extended Item button XDF shows a form component search window Then the user can either search the existing form components with some keywords or create new form components 4 2 Base Form Generation Base form is an online web form that is for suppliers or customers inputing their basic information and general capabilities or requirements The base form is generated from a collection of base form components such as BasicInfo Service and Extension Base form provides users with interfaces to extend the base form if necess ary eXtensible Dynamic Form Add This item GENERATE XML INSTANCE SAVE DYNAMIC FORM RESTORE DYNAMIC FORM Base form components OOOO Y Basic Information Y Drilling Service material search Basic Info asMaterial name Drilling aterialGroup Y Service List PackingMaterial BuildingMaterial Drilling Service edit delete Descnption string Vial V User Extension asMaterialinSpindleOD Y Keywords as MaterialinSpindleL enat Material edit delete RetumableMaterialindicator a J J SeriallD Add Keyword Keyword zircaloy i Keyword Star S
25. acilityGroup gt EquipmentGSroup List gt Equipment List P Extension Add EquipmentGroups 69 Y EquipmentGroup List Add EquipmentGroup Y group EquipmentType RARA AAN rr rr rar Add Equipment Addon gt Extension Add Equipments Y Equipment List Add Equipment name type Description string Brand NumOfEquipments Addon gt Capability gt Extension The Equipment form component includes several input fields as follows name equipment name type equipment type or classification e g Milling 10 Description equipment textual description NumOfEquipments the number of equipments that the facility has Addon the additional parts features of equipment e g coolant addon for CNC machine Capability max min dimension capability supported by equipment The sub components are as follows Y Capability Y MaxDimension gt Diameter gt x gt y gt z gt Length gt Depth gt Weight gt Wide gt Thickness gt Extension e hasMaterial The Service page also includes the hasMaterial form component to specify the materials that the service can handle with The list of material includes Inconel Hastelloy Magnesium Polymer Alu minum1200 Aluminum2030 StainlessSteel Titanium Aluminum2011 Alu minum2014 AlloySteel Aluminum Copper Bronze Nickel CarbonSteel Ca stlron Brass Metal Plastic and Rubber If y
26. an the words themselves In this approach we only tokenize the search keyword string k to a set of words denoted L k For each word in L k a set of synonyms can be obtained from WordNet We choose one synonym for each word from L k as its alternative and concatenate them into a single string concatenated word The set of all concatenated words is denoted as C For example if L k Shipment Address the synonym sets for words Shipment and Address in L K are S Send Delivery and Sp Destination Location Reference respectively All possible concatenated words form a set C 40 SendDestination SendLocation SendReference DeliveryDestination Delivery Location DeliveryReference For each concatenated word in set C we compare it to string e and calculate the similarity score by n gram method stated in section 5 3 2 This synonym based method combines the benefits of n gram based and WordNet based methods It not only works well for matching semantically similar concepts and synonyms but also identifies matches considering word variations and typos It is relatively faster than WordNet based searching method when the search keyword set L k is small However there is a serious problem with this exhaustive synonym based method That is if you have N words in L k and each word has M synonyms the set C will contain MY number of concatenated words In other words this method becomes very ine
27. appropriate supplier discovery methods two factors should be considered First the supplier profiles should be captured in a better structured and a machine interpretable format so that search function can better identify the relationships between supplier profiles and customer queries and discover more relevant suppliers Second the form ar chitecture should provide users with advanced search engine or other approaches to guide them input their information precisely The form architecture should also provide functionality that allows users suppliers or customers to extend the form by considering their own terms and structures of contents Work reported in this thesis is aimed at addressing these and other related issues 1 2 Contributions of the Thesis In this work we propose an innovative form architecture called eXtensible Dy namic Form XDF to facilitate the process of collect function in supplier discovery XDF provides suppliers or customers flexibilities to extend the base form by either searching the existing form components or creating their own form com ponents An intelligent search engine is provided for suppliers or customers to search existing form components User created form components will be stored as user defined schema in repository and they can be searched and reused by other users or form components later 1 3 Organization of the Thesis The remainder of the Thesis is organized as follows Chapt
28. ccurrence option is as following 61 Y Address street city zipCode An example of the form component added by Multiple occurrence option is as following Y Address List Add Address street city zipCode street city zipCode You can click Add Address to add additional Address form components A 1 6 2 Define New Form Components If you cannot find any proper form component and want to create new form components click MddasNewcomponent The following dialog box will be shown to con firm the process 62 Search new form components The new form component is a container node It can either contain multiple data input node or be extended to contain other contain nodes In following Choose your form component type dialog you can choose your form component to be simple type or extended type Simple type user created form component can only contain multiple data input nodes with the same primitive type while extended type user created form component can contain other user created form components or searched existing form components Choose your form component type Simple type this form component only contains data input nodes The following is to choose the data type for these data input nodes and the occurence of these data input nodes Type MinOccurs MaxOccurs decimal 1 1 O Extended type You can further extend this form component oK There are five different da
29. ccurs equal to one and minOccurs also equal to one Figure 4 6 shows a form component that is transformed from a simple Type element Certification Its maxOccurs attribute and minOccurs attribute have the same value which is one Therefore There is only one Certification item in the list and user can not add remove items to from this list 4 1 4 2 Form Component with Complex Type When the wrapped element is complex Type the Element Wrapper will delegate complex Type Wrapper to render the form component A form component trans formed from element with complexType is called complex type form component The representation of a complex type form component is determined by the content type of the wrapped complex Type element 24 If the content type is simpleContent Type the form component will be rendered by SimpleContentWrapper SimpleContentWrapper renders a form component the same way as Simple Type Wrapper does except that it also renders attributes carried by the wrapped element lt xs element name BuildingMaterial gt lt xs complexType gt lt xs simpleContent gt lt xsrextension base xs string gt lt xstattribute name ID type xs string use optional gt lt xs extension gt lt xs simpleContent gt lt xs complexType gt lt xs element gt gt 5 Y BuildingMaterial ID Building Material Figure 4 7 a form component transformed from a BuildingMaterial element Figure
30. ch search engine was compared with the discovery results from human experts The comparison was measured by Recall Precision metrics 37 and normalized DCG normalized Discounted cumulative gain 38 This experiment uses Top n Precision Recall metric proposed by Kim 30 that shows the result of both Precision and Recall with n 30 1 2 Experimental Results Table 5 1 shows the performances of the four search results The measures are normalized to the 0 1 range The first two rows give the Recall when comparing the top 6 and 10 of the search results with the list generated by the human expert The recall value 0 50 on the upper left corner is read as among the top 6 suppliers found by the Keyword method 3 of them are also in the top 6 from the human experts list Table 7 1 Performances of Search Engines Metrics Keyword Ontology Partial XML Full XML Top 6 0 50 0 50 0 50 0 50 Top 10 0 60 0 50 0 60 0 60 nDCG 0 32 0 90 0 85 0 92 The results shown that for Recall metrics the ontology based search performed l slightly worse than both Key word based and our XML based search In terms of nDCG which takes into consideration of relative ranking of all 30 suppliers per formance of Keyword based search is significantly worse than the other two More interestingly although the XML based method performed worse than the ontology based search when giving the same input of 11 key words 0 85 vs 0 9 its perfor mance improved to 0
31. chema ate N schema Select at least one up to 20 categories to complete your 1 i F DA 1 Y Screw Machine Products amp Turned Parts Swiss 1 STD Co re U d fi d Machine Model Star SA 16 CNC a E components ser e ne i i als schema 2 Il Gun Drilling Machine Model Eldorado Gun Driller l l Capabilities Holes From 0 078 i ll To 11 250 Depth epth 15 XML instances repository Y Clamping Systems Bar Stock Shapes W Rounds 2 Squares ARRIN Y Hexagons Tran or _ hanes lt seuares Web Clampi Materials Services Y Steel Stainless Name zircaloy Figure 3 2 Architecture of XDF 3 2 1 Data Layer The Data Layer is a repository that stores all the XML schemas that define form components These include domain ontology schemas core component XML schemas pre defined manufacturing schemas and user defined XML schemas The domain ontology schema is generated based on ontology defined in 23 Core compo nents schemas come from Open Application Group s Integration OAGi 24 OAGi implements core components schemas with the purpose of increasing interoperability for enterprises and encourages all business languages to be based on same concepts These core component schemas define grammar rules key naming conventions and key common content Since core component schemas are mainly focusing on the 10 general concepts of e business we created pre defi
32. cs High temperature Copper PP la Extended ltern Plastic Extended ltern Extended Item Y Axis Axis 5 UserDefinedMaterial Add UserDefinedMaterial UserDefinedMaterial Forging Following the same steps we added all four services to the XDF 16 E EE E E GENERATE XML INSTANCE SAVE DYNAMIC FORM Y Basic Information Y Vertical CNC Milling Basic Info name Vertical CNC Milling Y Service List Vertical CNC Milling edit delete ali Water jet cuttng edit delete Horizontal CNC Milling edit delete gt Keywords CNC Turing edit delete gt SpecilityGroup gt Facility Group P hasMaterial List User Extension gt Extension A 44 XML Instance After inputing all the information We can click the GENERATE XML IN STANCE button to generate the XML instance of the inputted data lt Supplier gt lt name gt LT Enterpirse lt name gt lt url gt www itenterprises com lt url gt lt hasIndustryFocus gt lt AerospaceProductandPartsManufacturing gt lt hasIndustryFocus gt lt hasIndustryFocus gt lt 0ilAndGasFieldMachineryAndEquipmentManufacturing gt lt hasIndustryFocus gt lt hasIndustryFocus gt lt UserbefinedIndustry gt Alternative Energy lt UserbefinedIndustry gt lt hasIndustryFocus gt lt hasIndustryFocus gt lt UserDefinediIndustry gt Contract Manufacture lt UserbefinedIndustry gt lt hasIndustryFocus gt
33. different terms structures and semantics users may not be able to find a appropriate form component to fit their data Thus users can extend the base form by creating their own form components These user defined form components will be transformed into XML schema and saved in user defined schema file Step three and Step four When user finish inputing their data all the data will be transformed into XML instances and be stored to XML instance repository Step five XML instance will be used in the process of supplier discover This process will be conducted by matching customer requests XML instance and supplier profile capabilities XML instance To accomplish all these functionalities a three layered architecture was devel oped 3 2 XDF Architecture XDF employs a three layer architecture Presentation Layer Business Logic Layer and Data Layer As illustrated in Figure 3 2 the Data Layer is a repository that stores all the XML schema files and XML instance files the Business Logic Layer contains a Schema Library and it handles all the communication between Presentation Layer and Data Layer through a collection of web services the Pre sentation Layer is a web based dynamic form through which users can interact with the whole XDF system Presentation Layer Business Logic Layer Data Layer Schema i Form Web A XML sciema repository Components l a i eee Services Man Pre defined ontology s
34. ements service fetches the form elements from Schema Library that will be transformed into base form components by a collection of wrap pers These base form components will build the base form e SearchFormElements service searches and returns form elements that will be transformed into form components that user is looking for 12 e StoreUserDefinedFormComponent service stores user defined form components back to Schema Library as form elements These form elements will be saved as XML schema in user defined schema file e StoreXMLInstance service stores XML instance generated from user inputed data to XML instance repository The relationship between form component and form element and the role of form component in Presentation Layer will be discussed in Chapter 4 3 2 3 Presentation Layer Presentation Layer is the User Interface from where users can input data search and create form components As illustrated in Figure 3 4 Presentation Layer includes a collection of form components There are three types of form components base form components user defined form components and searched existing form components Base form components are used to build the base form of XDF They are generated when user load XDF website When users can not find a appropriate place in the base form to fill in their information they can search existing form components through an intelligent form component search engine to extend the base f
35. er investigate because of the complexity and ambiguity in natural language 14 To overcome the problems of these approaches several knowledge based ap proaches have been developed for the manufacturing domain 15 16 17 The search capabilities can be enhanced by utilizing manufacturing knowledge based on the formal semantic representations e g Ontology Most of them employ ontolo gies to capture and represent semantic information The ontology should be shared and agreed upon by both suppliers and buyers often called shared ontology However ontology based approaches often lead to many challenges due to immaturity of technologies in semantic representation measuring and reasoning We employ three reasons First developing and maintaining a single shared ontology is time consuming and expensive because all participants should keep understanding all the concepts and semantics Second some information is unlikely to be captured by the shared ontology because it could be too specific and make the shared ontology too complex Third there are no tools to easily capture the ontological information from the textual descriptions Another approach to enhance the search capabilities is to utilize XML data representations which are widely used in the e business industry to represent the structured information XML data representations are XML instances of a XML schema that defines their structure content and semantics XML instance is typi
36. er 2 describes background of supplier discovery and the related works The architecture and logic flow of XDF is described in Chapter 3 Chapter 4 explains in details how the base form of XDF is generated Chapter 5 explains XDF extension and discusses four form component search methods A comparison experiment on the performance of these four search methods is provided Chapter 6 explains XML instance creation Chapter 7 measures the performance of XDF by comparing XML based match making algorithm to some other supplier discovery approaches Conclusions and future work are outlined in chapter 8 Chapter 2 Background and Related Works E marketplaces are a new business model which is developing rapidly in todays dynamic markets Typically e marketplaces have three roles provision of institu tional infrastructure supplier discovery by matching customers s requirements and suppliers capabilities and facilitating the transaction 10 In this work we focus on the supplier discovery role especially the data collection in supplier discovery Most of the approaches proposed 11 12 for matching customers and suppliers in supplier discovery are based on similarity retrieval of textual description T hese approaches often ignore semantics contained in the textual descriptions Though a few natural language processing NLP technologies have been developed to an alyze the meaning of textual descriptions 13 Their practicability needs furth
37. er extend the DrillingMachine form component An example is as follow 64 Y DrillingMachine Extended Item Y hasMaterial List Add hasMaterial Plastic H4 Plastic string Y Quantity Quantity decimal A 2 BasicInfo Form Page The initial form page of eXtensible Dynamic Form is BasicInfo page as shown below eXtensible Dynamic Form Add This Item GENERATE XML INSTANCE SAVE DYNAMIC FORM RESTORE DYNAMIC FORM Y Basic Information Y Basic Info Basic Info search name gt Service List gt User Extension url P hasindustryFocus List gt hasProductFocus List gt hasCertification List Extended Item You can input your basic profile including supplier name web site URL industry and product focus categories certification and more into BasicInfo form page The initial form page cannot be removed Each data node is described in the next subsection 69 A 2 1 BasicInfo Nodes e name suppliers name e url suppliers website URL e hasIndustryFocus manufacturing industry types that supplier has focused on You can add multiple industry focus data node Each data node provides a list of industry types that is obtained from the manufacturing ontology You should choose the most relevant industry type of what supplier has been involved in If you want to input additional information then you can click _Extended Item to extend the form Y hasindus
38. ery and Sy Destination Location Reference respectively For the word Deliver in L e The highest n gram similarity score is with word Delivery in synonym set S1 which is 0 66 For word Location the highest n gram similarity score is with word Location in synonym set S2 which is 1 And since L k and L e have the same length the length penalty factor is 1 Therefore the similarity score between string k and string e is 1 x 5 x 0 66 1 0 83 In the next section we will assess the performance of the five form component search methods in terms of hit rate and computing time through the experiments we have conducted 42 5 3 6 Experiments and Results We have conducted experiments with limited scope to assess the performance of these four search methods using eight sample form components from the Schema Library to generate search keys We consider these samples as the correct expected search results when searched by keywords generated from them Then we generate the arbitrary search keywords based on the names in the samples There are two ver sions of search keywords a synonymous version and a typo version For example for a form component named DespatchDestination we generate two search keywords a synonymous version ShipmentAddress and a typo version ShpmetAddres Using these two versions of keywords we use each of the four different methods to search the Schema Library
39. fficient when the search keyword set L k is large Therefore For efficiency purpose we adopt a greedy version of this method to compute the similarity between string k and e 5 3 5 Synonym based Search Greedy The greedy synonym based search method goes as follows 1 we tokenize k and e to L k 01 03 0 p and L e 41 82 tg 2 For each word v in L k a set of synonyms are drawn from WordNet denoted as S Si1 512 Sijs 3 For each word u in L e find the synonym s in all syn onym sets 5 with highest n gram similarity score between ug and s recorded as Score u The set of scores for all words in L e form ScoreSet Score uq q Al 1 2 L e The average scores in ScoreSet is the similarity subject to a penalty for length difference between L k and L e When L k gt L e The similarity is computed as _ q _ s L e EDD Lo n Sim e k 1 ZOO x LO x 2 Score ug 5 4 where the first factor of the formula is the length penalty Note that When L k lt L e the highest L k scores will be selected from ScoreSet from Score Set The similarity is computed as _ q _ s L e E A Le Sim e k 1 MATO xX LO x 2 Score u 5 5 For a concrete example assuming L k Shipment Address and L e Deliver Location and the synonym sets for words Shipment and Address in L K are S Send Deliv
40. h hi ip pm me en nt t tri grams s sh shi hip ipm pme men ent nt t quad grams 4458 4 sh shi ship hipm ipme pmen ment ent nt tHFF When we use tri grams the similarity between word shipment and ship is 2 x 4 10 6 0 5 n gram based searching solves the problem of typo input that Keyword based searching faces Because every string is decomposed into small parts so any errors that are presented affects only a limited number of n grams leaving the rest intact However n gram based search has its limitations it cannot match semantically similar concepts e g synonyms 5 3 3 WordNet based Search Both keyword based and n gram based methods use string based similarity metrics WordNet based search on the other hand employs semantics of words to enhance string based metrics Kim 30 proposed a WordNet based approach to measure the semantic similarity between two strings This approach has four steps 1 Pre processing Similar to keyword based search this approach first tok enizes the string k and string e to two sets of words L k and L e respectively Thus the string to string matching problem is reduced to word to word matching 37 problem as illustrated in Figure 5 3 ShipmentAddress Shipment Address t tN AZ ft string to string matching ay word to word matching Nb DespatchDestination Despatch Destination Figure 5 3 St
41. h Keywords Typo Version of Search Keywords Matching Method Hit Rate Average Computing Time millisec Keyword based 12 5 25 N gram based 87 5 43 WordNet based 12 5 4345 Synonym based exhaustive 87 5 1840 Synonym based greedy 87 5 2016 However the experiment conducted above only considers search queries with one and two search words For fair performance comparison We conducted the experiment to compare computing time of these methods considering L k gt 2 In this experiment we only compare the computing time of WordNet based exhaustive 44 synonym based and greedy synonym based methods since they have best overall performance in terms of hit rates and they are computationally expensive Table 5 3 shows the comparison of computing time for these three search meth ods Each computing time shown in Table 5 3 is the average computing time of eight search queries From the first row of Table 5 3 we can see that the computing times for exhaustive synonym based and greedy synonym based search methods are almost the same This is because the concatenated word set of exhaustive synonym based method is relatively small when only two search keywords were inputted However the computing time of exhaustive synonym based search method increases signif icantly when the number of search keywords is beyond two This is reasonable because the algorithm complexity for exhaustive synonym based search method is O M where N is
42. h keywords that represents the meaning of what you want to add These search keywords should be written as a string in the camel case or inputted separately by a white space If you input search keywords as a string in camel case these words should be joined without spaces with each words 59 initial letter capitalized within the compound and the first letter is either upper or lower caseas in PurchaseOrder shippingAddress or xPath If you want to search the existing form components then click Search put ton A synonym based search algorithm is currently used to find the matched form components The detailed process will be described in A 1 6 1 If you cannot find any proper form components and you want to create new form components using your input label then click 4435 New component button The detailed process will be described in A 1 6 2 If you want to return to your form then click Close button The dialog will be closed A 1 6 1 Search Existing Form Components If you click the Search button then the system starts to search the existing form components which matched to your input label An example for the search result of the input label Address is as follows Search new form components Address Search Add as New component Close Items from http www emarketplaces com pre defined templates Add Address Preview items from urn oasis names specification ubl schema xsd CommonBasicCompone
43. h word WordNet produces a group of synonym sets Each synonym set is an equivalence class of words sharing the same meaning within an ontology Figure 5 4 shows that for word Shipment two synonym sets are drawn from WordNet For word Despatch four synonym sets are drawn from WordNet Synonym set m 4 a 4 rd Shipment Despatch Dispatch Freight Payload Cargo Expedition Expeditiousness Despatch Figure 5 4 Synonym sets for word Shipment and Despatch 99 The similarity between two words say A and B is computed by averaging the similarity scores of all combinations of synonym sets each of which comes from the synonym sets group of word A and synonym sets group of word B respectively Kim s method 30 incorporates a form of information content based measure 34 in computing the similarity between two synonym sets 4 The final step is to run the model and compute the results WordNet based method works well for matching semantically similar concepts and synonyms However it is relatively slow due to the search of a large lexical database of WordNet and it cannot deal with the problem of word variations and typos that n gram based search can To address these limitations of WordNet based method we designed a Synonym based search algorithm 5 3 4 Synonym based Search Exhaustive This method extends the n gram method by using the synonyms of search keywords rather th
44. hemas need to be integrated into current XDF system 54 Appendix A User Manual This manual is intended for new users with little or no experience using dy namic form interface The goal of this document is to give a broad overview of the main functions of the dynamic form and step by step instructions about how to collect supplier profiles using the dynamic form interface Our system provides more flexible ways for suppliers to register their information using better structured customizable and better machine interpretable form architecture called eXtensible dynamic form XDF The XDF can customize the existing base form based on a collection of reusable form components The form components are modular so they can be easily attached or detached to from the existing base form Each form com ponent is tree like structured and may contain other form component as children There are four types of form components ontology components core compo nents pre defined components and user defined components They are transformed from domain ontology schema core component schema pre defined manufacturing schema and user defined schema respectively Ontology component provides a collection of manufacturing terminologies and the semantically structured form components which are related to the manufacturing ontology 313 Core component provides a collection of core form components which could widely appear in man
45. hineProducts SheetMetalWeldingEquipment WeldingMachines Figure 4 10 a form component transformed from a DrillingEquipment element 28 4 1 4 3 Wildcard Form Component Wildcard form component is transformed from any element As suggested by XML Schema best practices 26 Placing an any element at the end of complexCon tent of a element is a good way of adding extensibility to XML schema Employed the same concept Wildcard form component is utilized to extend the base form The extension of the base form will be discussed in Chapter 5 lt xs element name Service gt lt xs complexType gt lt xs sequence gt lt xs element name name type xs string gt lt xs element ref cbc Description minOccurs 0 gt lt xs element ref Keywords minOccurs 0 gt lt xs element ref SpecilityGroup gt lt xs element ref FacilityGroup gt lt xs element ref ont hasMaterial minOccurs 0 maxOccur lt xs element ref Extended Item minOccurs 0 gt lt xs sequence gt lt xs tattribute name n lt xs complexType gt type xs string use required gt lt xs element name Extended Item type T Extension gt lt xs complexType name T Extension gt lt aque z lt xs complexType gt Y Drilling Service a L name Drilling Service Description string Find a new Element a Search Add as New tem Close gt SpecilityGroup Py A gt FacilityGroup A P hasMaterial List y e
46. ience and Electrical Engineering Discovery of suppliers supplier discovery is essential for building a flexible network of suppliers in a supply chain The first step for supplier discovery is to collect manufacturing capabilities of suppliers and requirements of customers In traditional e marketplaces online form interfaces are typically used to collect the requirements and capabilities However those forms are mostly lack of flexibility to capture a variety of requirements and capabilities in a structured way In this the sis we propose new innovative form architecture called eXtensible Dynamic Form XDF to facilitate data collection process of supplier discovery This architecture provides several key innovations including 1 architecture for users suppliers or cus tomers to create new structure of form for their own contents 2 an synonym based intelligent search engine facilitating users to reuse the existing form components 3 hierarchical representation of the requirements and capabilities as XML instances Experimental results demonstrate that the proposed architecture is valuable for fa cilitating the supplier discovery process Keywords eXtensible Dynamic Form Supplier Discovery Data Collection Synonym based Search EXTENSIBLE DYNAMIC FORM FOR SUPPLIER DISCOVERY by Yan Kang Thesis submitted to the Faculty of the Graduate School of the University of Maryland Baltimore County in partial fulfillment
47. ill delegate Simple Type Wrapper to render the form compo nent When its wrapped element is simple Type The rendered form component is a list of text boxes items for user inputting data If the value of maxOccurs attribute of the wrapped element is bigger than the value of minOccurs attribute user can dynamically add remove items to from a list Otherwise the number of items in the list is fixed A form component transformed from element with simple Type is called simple type form component lt xs element ref IndustryFocus minOccurs 1 maxOccurs unbounded gt lt xs element name IndustryFocus type xs string gt Y IndustryFocus IndustryFocus IndustryFocus IndustryFocus x mnam Figure 4 5 a form component transformed from a Industry Focus element with maxOccurs equal to unbound and minOccurs equal to one Figure 4 5 shows a form component that is transformed from a simpleType element Industry Focus Its maxOccurs attribute is equal to unbound and 23 minOccurs equal to one T herefore user can add arbitrary number of IndustryFo cus items to this list but must left at least one IndustryFocus item in the list lt xs element ref Certification minOccurs 1 maxOccurs 1 gt lt xs element name Certification type xs string gt Y Certification Certification Figure 4 6 a form component transformed from a Certification ele ment with maxO
48. in a RestrictionSimpleType object ListSimpleType object or a UnionSimple Type 18 Figure 4 2 Class Diagram of Schema Library object Complex Type object contains either SimpleContent object or ComplexCon tent object Both SimpleContent object and ComplexContent object can have zero or more Attribute objects ComplexContent object contains at least one instance of Compositor interface which must be a ModelGroup object ModelGroup object contains multiple instances of Particle interface that can be Element object Wald card object Group object and or another ModelGroup object Wildcard object is the representation of any element in Schema Library Summarily Schema Library represents XML schemas and their contained schema constructs in terms of Java objects After generating these schema Java objects XDF transfers them to the Pre sentation Layer A collection of wrapper objects will transform these schema Java 19 objects to form components 4 1 3 Wrapper Classes The instances of Wrapper classes i e Wrapper objects in the presentation layer are responsible to wrap the schema Java objects and transform them into form components Figure 4 3 shows all the Wrapper classes and Figure 4 4 illustrates the instance level relationship between Wrapper classes Figure 4 3 Wrapper Classes Figure 4 4 Class Diagram of Wrapper Classes 20 As illustrated in Figure 4 2 and Figure 4 4 wrapper classes keep the same
49. iness Logic Layer 4 Get base form elements Web d gt Services Schema Java l l P Objects may F Generator Search form elements Web Sch ma je gt Services aa Library l Store user defined b form components l f sA gt XML Schema l Generator Components Web Services gene rate Search Create l a Select at leas up to 20 categor a to complete your t j n Y Mica afpera Pars smss XML instance Machine Model Star SA 16 CNC gt fap lt SwissScrewMachine gt lt MachineModel gt Star SA 16 CNC lt Mach 2 W Gun Drilling lt SwissScrewMachine gt Store XML Machine Model Eldorado Gun Orilier r nsform lt GunDrillingCapabilities gt Capabilities lt MachineModel gt Eldorado Gun Driller lt oe q ol co e un oe instances Web To 1 250 e lt From gt 0 078 lt From gt lt To gt 1 250 lt To gt Depth 18 lt Holes gt i pet Services Y Clamping Systems lt BarStockShapes gt Bar Shape lt Rounds gt lt Squares Y unds lt BarStockShapes gt Y Squares i Clamping gt Y ao GunDrillingCapabilities Figure 3 3 Architecture of Business Logic Layer is responsible for transforming Schema Java objects into XML schemas Business Logic Layer includes four web services GetBaseFormElements ser vice SearchFormElements service StoreUserDefinedFormComponent service and StoreX MLInstance service e GetBaseFormEl
50. ity Largest user of Omax Wal AdditionalCapability 5 axis machining capabilit AdditionalCapability Six sigma Black Belt Qua AdditionalCapability Full time tooling engineer AdditionalCapability In house Magnetic particle AdditionalCapability Machining experience witt AdditionalCapability Design support and engine AdditionalCapability Assembly and kitting A 4 3 Service Page There are four services We take Vertical CNC Milling as an example We first inputted the name of the service and two specilities of this service short run production and quantitive production Then we inputted equipments and capabil ities of Vertical CNC Milling Finally materials used in this service were inputted 14 Y Vertical CNC Milling name Vertical CNC Milling Description OT string gt Keywords Y SpecilityGroup Y group Specility Type Prototype W Prototype prototype production Add Specility Specility Short run production x Specility Quantitive production x gt Extension Y EquipmentGroup List Add EquipmentGroup Y group EquipmentType ManualMillingEquipments 5 Y Equipment List Add Equipment CITI 19 Y Capability Y MaxDimension P Diameter vx xX 165 Y Extension Y hasMaterial List Add hasMaterial Extended ltern Elida FA Extended ltern AlloySteel y Extended ltern Y Characteristics Characteristi
51. lingEquipments gt lt Equipment gt lt name gt Vertical CNC Milling lt name gt lt ManualMillingEquipments gt lt EquipmentGroup gt lt FacilityGroup gt lt hasMaterial gt lt Stainlesssteel gt lt hasMaterial gt lt hasMaterial gt lt Carbonsteel gt lt hasMaterial gt lt hasMaterial gt lt AlloySteel gt lt Characteristics gt High temperature lt Characteristics gt lt Alloysteel gt lt hasMaterial gt lt hasMaterial gt lt Copper gt lt hasMaterial gt lt hasMaterial gt lt Plastic gt lt hasMaterial gt lt hasMaterial gt lt UserDefinedMaterial gt Forging lt UserDefinedMaterial gt lt hasMaterial gt lt service gt 18 Bibliography 11 M Christopher and H Peck Building the resilient supply chain Interna A tional Journal of Logistics Management vol 15 no 2 2004 pp 1 14 Thomas Publishing Company http www thomasnet com MFG com Corporate http www mfg com en Engineering Search and Supplier Catalogs http www globalspec com C S Li Y C Chang and J R Smith An e marketplace infrastructure for information Intelligent Multimedia Video and Speech Processing In Proc of International Symposium on 2001 pp 182 185 W3 org Extensible Markup Language XML 1 0 Fifth Edition Available at http www w3 org TR xml W3 org XML Schema Part 0 Primer Second Edition Available at http www ws org TR xmlschema 0
52. m component BasicInfo is encoded as a XML instance with an element named Supplier The sub elements of this Supplier element have the same hierarchical structure with the sub form components of the Ba sicInfo form component Chapter 7 will assess the performance of XDF by comparing XML based matching algorithm to some other supplier discovery approaches This XML based matching algorithm proposed by Kim 30 takes two XML instances generated by A Y Basic Info name AccuTrex Products Inc url http www accutrex com gt hasindustryFocus List P hasProductFocus List Y hasCertification List Add hasCertification Add UserDefinedCertification e son1 XML instance lt Supplier gt lt name gt AccuTrex Products Inc lt name gt lt url gt http www accutrex com lt url gt lt hasCertification gt lt Userbefinedcertification gt IS09001 lt UserbDefinedcertification gt lt hasCertification gt lt Supplier gt Figure 6 1 Saving Dynamic Form as XML instances XDF as input and computes the semantic based similarity between them 48 Chapter 7 Experiment on Supplier Discovery In this Chapter we assess the performance of XDF by comparing the search results based on XML instances generated by XDF to other supplier discovery ap proaches including a keyword based search and a ontology based search 7 1 Design of Experiment For experimental data we randomly choose 30 su
53. n Language MSDL an ontology created for for mal representation of manufacturing capability They also provide a quantitative measure to connect customer requirements and suppliers capabilities based on their semantic similarities The textual descriptions of the requirements and capabili ties are encoded into instances of MSDL The encoding process was performed by domain experts A total of 16 key words were extracted from the customer requirement as the input to the Keyword based search Ontology based search uses only 11 of the 16 key words the other 5 are not defined in the ontology For a fair comparison we ran two experiments for XML based method one creating requirement XDF form using all 16 key words the other using only the 11 key words defined in the ontology The results are called Full XML and Partial XML respectively The experiments were conducted as following Four search methods i e keyword based ontology based Partial X ML based and Full XML based were ex ecuted to discover suppliers whose capabilities satisfy customers requirements The result of each search engine is a ranked list of the 30 suppliers based on their similar ity scores To enhance the accuracy of the discovery result domain experts working 310 in the field of supplier discovery were requested to analyze the requirements and capabilities and manually produced a ranked list of matched discovered suppliers Then the discovery result of ea
54. ned manufacturing schemas that are specifically focus on manufacturing industry but not defined by the ontology All user defined form components will be stored as user defined schemas Users can integrate user defined form components with existing form components with the help of XDF s search engine The topic of search engine will be covered in Chapter 5 The repository also stores XML instances generated from user inputted data These XML instances will be taken as input in the process of supplier discovery In addition the repository stores a set of binary files which are not shown in Figure 3 1 These binary files store the whole eXtensible Dynamic Form for each user Users can retrieve their eXtensible Dynamic Forms next time they input new data Thus they do not have to start their work from the scratch 3 2 2 Business Logic Layer Business Logic layer is responsible for transferring information between the Presentation Layer and Data Layer As illustrated in Figure 3 1 Business Logic Layer contains a Schema Library and a collection of web services performs the function of transforming information between different representations Figure 3 2 shows the details on the Business Logic Layer The Schema Library is a collection of Java objects generated from XML schemas through a Schema Java Objects Generator In other words it is the repre sentation of XML schemas in the Business Logic Layer The XML Schema Generator 11 Bus
55. nt their own capabilities or requirements Therefore those fixed forms may not be able to capture users domain specific information Although some of those fixed forms provide users with search engines to find appropriate form components to put their domain specific information it did not solve the lack of flexibility problem of traditional fixed form First most of these search engines are based one keyword string matching methods therefore users with semantically similar capabilities or requirements but represented using different syntaxes may not be identified by simple keyword matching methods Second traditional forms do not provide users a way to create their own form components based on their domain specific concepts The second step is to find suppliers henceforth search function whose capa bilities are of the greatest relevance to requirements specified by customer Several approaches have been proposed to enhance the search function e g semantic based search They mostly reply on the structured data models such as XML 6 7 RDF 8 and OWL 9 The unstructured or semi structured information collected by traditional forms makes it difficult to use those advanced search approaches To enhance the search functions thus it is necessary to first enhance the form archi tecture that collects requirements and capabilities in a better structured way In order to collect information accurately and use the
56. nts 2 Add Reference Preview Add Description Preview Add CustomerReference Preview Items from urn oasis names specification ubl schema xsd CommonAggregateComponents 2 Add Address Preview Add RegistrationAddress Preview ii The results can be grouped by different types of form components ontology common core pre defined and user defined components Each result only shows the label of the form components prrrreressssreensensrenean Click Et to preview the form layout An example for preview is as follows Search new form components Address Search Add as New component Close Items from http www emarketplaces com pre defined templates Add Address E Click items from urn oasis names specification ubl schema xsd CommonBasicComponents 2 Add Refe epee EE EERIE ETE SO TI LE I TIT SUE LEE LS LOE OT NT SATS EE 0 Add Desc Add Cust street Items from ur city Add Addr Add Regi state a Ll HA sss zipCode EEE If you find the proper form component and want to add it in the current form then click 4 at the left of the form component label as follows Click Ada ddress Preview Each form component may occur more than once in the form If you want to add the form component only once choose Single occurrence option as follows _ Chooseing Occurence Single O Multiple OK Otherwise Multiple occurrence should be chosen An example of the form component added by Single o
57. ogies com Ontology 1236208666 owl cs _ xf Extended Item Y Certification Certification Figure 5 1 illustrates an example of form extending More detail processes to create or search the form components will be described in the following subsections 99 5 2 User defined Form Component User defined form components are stored as user defined schemas in the repos itory They can be searched and reused by other users or forms later Figure 5 2 illustrates an example of form component creation operation Find a new Element Search Add as New tem Close Choose your term type amp Type MinOccurs decimal B 1 rea iS term as a OK ok A Y KeyFeature lo A Add KeyFeature lt xs schema xmlns xs http ww w3 org 2001 XMLschema gt Exs element name KeyFeature type xs string gt KeyFeature lt xS schema gt Figure 5 2 an example of form component creation In this example the user inputs a keyword key feature and clicks the Add as New Item XDF creates new form component with the name of KeyFeature that has a text input box and insert it into the current position of the form The generated form component is encoded and saved as a element with string simple type in the user defined schema As explained in section 4 2 a form component can be either simple type or complex type This is also true for user defined form components If the user
58. only once or not at all We do not support the third option in that we allow an element appears multiple times in a model group Group corresponds to named group in XML schema language The difference between un named group and named group is that named group can be declared globally and referenced by other element 4 1 4 Form Components Form components are the basic building blocks for constructing XDF In user s perspective of view a form component is a sub form that can be attached as sub structure to base form or other form components A form component is rendered by an ElementWrapper and its representation is determined by the Element object wrapped by this ElementWrapper In other words a form component is generated from a Element object stored in Schema Library and they have one to one relation ship Such a Element object is named form element Logically Schema Library is a Library that stores all form components As illustrated in section 4 1 2 and 4 1 3 Element object and element schema construct refer to the same concept with different representations Therefore with the purpose of explaining the relationship between form component and element schema construct and how a form component 22 is represented by a Element Wrapper we will replace the role of Element object in Element Wrapper with element schema construct in the following three sections 4 1 4 1 Form Component with Simple Type Element Wrapper w
59. orm Users are also allowed to create their own form components on the fly to fill in their domain specific information Details about how the Presentation Layer is constructed will be covered in Chapter 4 13 Presentation Layer Business Logic Layer i Base form Select at least one up to 20 categories to complete your a Screw Machine Products amp Turned Parts Swiss Machine Model Star SA 16 CNC J Add Forn gt Gun Drilling Machine Model Eldorado Gun Driller Capabilities Holes From 0 078 To 11 250 Depth 15 J Clamping Systems Bar Stock Shapes Rounds Squares Hexagons Materials Steel Stainless Name zircaloy A i i i Figure 3 4 Architecture of Presentation Layer 14 Chapter 4 XDF Generation In the traditional e marketplaces online web form interfaces are typically used for the supplier profiles collection Suppliers can input their data on the web form These inputted data will be sent to server for further processing Several electronic marketplaces e marketplaces such as Thomasnet 2 mfg com 3 and GlobalSpec 4 provide such web forms to collect suppliers information Typically these web forms are static web pages The limitation of such static web pages is that if some time later more suppliers information need to be captured we have to add new fields to the form by manually updating these
60. ou want to define new material 71 you can choose UserDefinedMaterial in the list as follows Y hasMaterial List Add hasMaterial UserDefinedMaterial Add UserDefinedMaterial x UserDefinedMaterial A 4 Example In this section we will give a concrete example of transforming text description of capability profile of LT Enterprise into XML instance vis XDF A 4 1 Raw Data of Supplier Capability Profiles Core services Vertical CNC Milling 165 x 55 x 24 Vertical Envelope 5 axis Horizontal CNC Milling CNC Turning Lengths to 120 0 Capacity to 18 0 Diameter Water jet Cutting Materials e Stainless Steel Carbon and Alloy Steels Aluminum High Temperature Alloys Copper Plastics e Castings Forging We have been satisfying clients since 1982 in Defense Aerospace Gas amp Oil Energy contract Manufacturing markets We have invested in the latest equipment technology which consists of Water jet CNC lathe vertical and horizontal machining up to five axe for prototype short run and production quantities e Largest user of Omax Waterjets in North America AS9100B amp ISO 9001 Registration e 5 axis machining capability Six Sigma Black Belt Quality Manager Full time tooling engineer on staff In house Magnetic particle inspections Machining experience with composite and exotic materials e Prototype to production capabilities e Design support and engineering se
61. ow to add a Drilling Service Service Giving a Name i After click the OK button the Grilling Service is added to the Service List on the right side of the base form When you click the hyperlink of Drilling Service the Drilling Service will be showing on the right side of the navigation sidebar Add GENERATE XML INSTANCE SAVE DYNAMIC FORM RESTORE DYNAMIC FORM Y Basic Information Y Drilling Service Basic Info y gt name Y Service List a edit delete Description So string gt User Extension x Keywords gt SpecilityGroup FacilityGroup P hasMaterial List Extended Item A 3 1 Service Nodes e name manufacturing service name e Description manufacturing service description e SpecilityGroup a group of supplier core specialties for the specific manufac turing services For example supplier may specialize in precision machining prototype production short run production quantitative production assem 68 bly testing and so on You can specify multiple specialties by choosing them from the given list or type in Y SpecilityGroup Y group SpecilityType Prototype Prototype Add Specility gt Extension e FacilityGroup a group of manufacturing facility for the specific manufacturing services You can choose the existing equipment list from EquipmentGroup or define new equipment using Equipment as follows Y F
62. ppliers from Thomasnet and collected their capability data in the form of textual description An artificially made customer requirements was created A human expert from DSN innovations Corp provided a similarity ranking of the 30 suppliers for the requirement We consider this human expert generated raking as the ground truth for the experiment Using XDF we encoded the textual descriptions of the requirements and capa bilities of the 30 suppliers and then generated their XML instances XML instances are analyzed by a semantic based XML instance matching algorithm proposed by Kim 30 This method takes two XML schemas or instances as input and computes the semantic based similarity between them We compare results of XML based method to two other approaches keyword based search and ontology based search For Google keyword based search we used Google custom search engine which allows user to create a customized search pref 49 erences e g search within specific website URL We built a sample web site that contains URL to multiple web pages each of which includes the textual capability description for one of the 30 suppliers Then we run the search of the web site using the google custom engine We choose multiple keywords of the given requirements For Ontology based approach we used a match making algorithm proposed by Ameri and Dutta 36 The proposed match making algorithm operates over Manufacturing Service Descriptio
63. ring to String Matching to Word to Word Matching 2 Modeling The word to word matching problem can be modeled as a max imum weighted bipartite graph matching problem 31 32 as follow Given an undirected graph G AN B E where A a4 do m and B b1 b2 b are two sets of vertices Each vertex in graph G represents a word E 11 12 e AX B is a set of edges each of which carries a weight Each edge e connects a vertex a in set A to a vertex b in set B and vertices in the same set can not be connected Each vertex in set A has a connection to every vertex in set B A matching M of graph G is a subset of E such that no two edges in M shares a common vertex The maximum weighted bipartite matching is a matching whose sum of the weights of the edges is the highest among all possible matchings The maximum weighted bipartite matching is formulated as an integer programming defines as bellow 38 Maxmize gt e EE Dd Subjectto A AN erl 5 3 gt j 1 Lij 1 V2 1 B vi 0 1 where xi is 1 1f e M otherwise is 0 3 Weights Computing The maximum weighted bipartite graph matching problem depends on the weights carried by each edges in E Weight on an edge is equal to the similarity score between two words connected by this edge In Kim s approach 30 the similarity between two words is computed based on their syn onym sets drawn from WordNet 33 For eac
64. rvices e Assembly and Kitting 12 We first extracted information from the text description of capability profile There are four core services Vertical CNC Milling Horizontal CNC Milling CNC Turning and Water jet cutting Eight materials have been used in the production process AS9100B and ISO 9001 certifications have been registered Capabilities of LT Enterprise are also described in this text description A 4 2 Basiclnfo Page In the Basic Info page we first inputted the name of the supplier that is LT Enterprise and the industries it focuses on Then we inputted the certifications LT Enterprise registered We extended the base form by adding several capabilities of LT Enterprise Y Basic Info name LT Eorernirse url www itenterprises con Y hasindustryFocus List Add hasindustryFocus AerospaceProductandPartsManufacturing Bl Extended Item x OilAndGasFieldMachineryAndEquipmentManufacturing H Extended Item x UserDefinedindustry B Add UserDefinedindustry d UserDefinedindustry Alternative Energy UserDefinedindustry rH Add UserDefinedindustry UserDefinedindustry Contract Manufacture d 13 Y hasCertification List Add hasCertification 1509001 o lx Extended Item UserDefinedCertification la Add UserDefinedCertification UserDefinedCertification AS91008 Y AdditionalCapability Add AdditionalCapability AdditionalCapabil
65. s users to input their data into the form Each container node can be expanded or collapsed by clicking or Y at the left of the node if it contains other nodes If a container node is designed for containing a node that might occur multiple times it will be indicated by the postfix List at the label Y hasindustryFocus List A 1 3 Data Input Node name A There are different types of data input nodes Decimal Integer Double Boolean DateTime and String However the current prototype system of dy namic form only provides a single interface as shown above You can input any text data into the input node Other data types will be supported later 08 A 1 4 Data Input Node List Type BallorRollerBearings B This is a special type of data input node to represent the list type of data You should choose one of them listed in the combo box It may contain other data nodes as children A 1 5 Form Component Remove Button A Each form component has a remove button You can easily remove the form components by clicking at the right of the form components A 16 Extension Each form component may include a Extended tem button which allows you to extend the form by adding the existing form components or creating new form components When you click the Extended ltem button a pop up dialog will be shown as follows Search new form components Search Add as New component Close First you should input searc
66. static pages and then recompiling and redeploying the application to the server In contrast we automatically generate web forms for XDF from XML schemas that defines the structure of the whole web form The benefit of using XML schema as basis for generating XDF is that XML documents are well designed for structured content and because of its widespread acceptance between applications companies and industries XML becomes an im portant part of any matching strategy of structured content In addition revision of XML schemas are much easier than revising the form directly therefore making the XDF form maintenance and revision much more efficient 15 4 1 Form Components Generation The eXtensible Dynamic Form system will parse XML schemas and transform all the schema constructs T in these schemas into a middleware called Schema Library The Schema Library is a collection of schema Java objects 25 each of which represents a schema construct These schema Java objects will be wrapped by wrapper objects in the presentation layer and transformed into form components which are the basic building blocks for constructing the base form Figure 4 1 shows the logic flow of form components generation Schema Library 4 Schema Java objects Schema Java objects k E N i form components is XML schema Wrappers Objects Generator Figure 4 1 Form Components Generation Flow 4 1 1 XML Schema XML
67. te types Decimal Integer Double Boolean Date Time and String Note that current dynamic form generates the same data input node regardless of those data types but it will be updated The occurrence can be defined as MinOccurs and MaxOccurs indicators The MinOccurs indicator specifies the minimum number of times a data input node can occur while the MaxOccurs indicator specifies the maximum number of times a data input node can occur The default values for MinOccurs and MaxOccurs are 1 and 1 respectively Note that 1 means unbounded meaning that there is no limitation to the occurrence of the data input node An example of the form 63 component created as data input node with multiple occurrences is as follows Y DrillingMachine Add DrillingMachine DrillingMachine DrillingMachine _x DrillingMachine x If you set the new form component as extended type choose the second option as follows Choose your form component type Simple type this form component only contains data input nodes The following is to choose the data type for these data input nodes O and the occurence of these data input nodes Type MinOccurs MaxOccurs decimal 1 1 O Extended type You can further extend this form component ok The newly generated DrillingMachine form component with extended type is as follow Y DrillingMachine x Extended Item sli Then you can click _Extended tem button to furth
68. tryFocus List Add hasIindustryFocus AerospaceProductandPartsManufacturing gt Extended Item X For other industry types that are not listed you can choose UserDefined In dustry in the list and then type the industry name in the form as shown below UserDefinedindust 2 lx Add UserDefinedindustry UserDefinedindustry e hasProductFocus the manufacturing product types that supplier has pro duced 66 Y hasProductFocus List Add hasProductFocus MachineToolsA4ndParts Extended ltem x Similar to hasIndustryFocus data node you can add multiple product types extend the form or input UserDefinedProduct in the form e hasCertification supplier quality certifications You can add multiple certifi cations extend the form or input UserDefinedCertification in the form Y hasCertification List Add hasCertification AS9000 AS59100 15014000 1509001 1509002 MIL UserDefinedCertification Y Ex e Extension you can add any other form components by searching the existing form components or creating new form components For details please go to section A 1 6 A 3 Service Form Page Multiple Service form pages can be added to the base form Each form page includes several data nodes name languagelD Description Keywords Special tyGroup FacilityGroup hasMaterial and Extension following is the screenshot 67 shows h
69. uding those in the base form has a Extended Item button as shown in Figure 5 1 The Extended Item button is actually a Wildcard form component transformed from any element as explained in section 4 2 3 The any element allows a element to contain any types of sub elements which allows form component to contain any other form components as its sub structure 32 Once the user clicks the Extended Item button XDF shows a form component search window T hen the user can either search the existing form components with some keywords or even create new form components Figure 5 1 shows the form search operation with the keyword certification XDF s search engine returns a list of form components User can preview each of the form components and insert one of them into the current position of the form In this case the Certification form components was added to the current form Y Keywords Add Keyword Find a new Element Keyword drill certification Search Add as New Item Close Preview defined templates Certification Add Extended Item Add hasCertification Preview Y Keywords 4 Items from um oasis names specification ubl schema xsd CommonAggregateComponents 2 Add DocumentDistribution Preview Add Keyword Keyword drill PS Items f um oasis names specification ubl schema xsd CommonBasicComponents 2 Addy CertificateType Preview toh from http www owl ontol
70. up gt User Extension gt hasMaterial List Through the list box at the top of the base form you can add Service and User Extension to the base Form All the Service and User Extension will be listed at the left side of the base Form for helping user navigate through different Service and User Extension When you click the hyperlink the corresponding web page will be showing on the right side of the navigation sidebar Service Add GENERATE XML INSTANCE SAVE DYNAMIC FORM RESTORE Basic Info e a A name Water jet cutting Y Service List _ e Vertical CNC Milling edit delete Description Water jet cui edit delete string Horizontal CNC Milling edit delete gt Keywords CNC Turing edit delete gt SpecilityGroup lr spe gt User Extension FacilityGroup gt hasMaterial List gt Extension ov delete Each page can be simply deleted by clicking button at the right side of each Service or User Extension Note that the initial basic information page i e BasicInfo cannot be deleted A 1 2 Container Node Y hasindustryFocus List XDF consists of two types of nodes container node and data input node The container node contains other nodes either other container nodes or data input nodes Container node has a text label indicating what the nodes it contains are all about Data input node has an input field which allow
71. utational Linguistics and Intelligent Text Processing vol 2945 2004 pp 445 456 B Kulvatunyou H Cho and Y J Son A semantic web service framework to support intelligent distributed manufacturing International Journal of Knowledge based and Intelligent Engineering Systems vol 9 2005 pp 107 wae J Jang B Jeong B Kulvatunyou J Chang and H Cho Discovering and integrating distributed manufacturing services with semantic manufacturing ca pability profiles International Journal of Computer Integrated Manufacturing vol 21 no 6 2008 pp 631 646 H Yu an Y Tao L Lilan and S Haiyang Research on manufacturing re source discovery based on ontology and QoS in manufacturing grid In Proc of the 2006 International Confe rence on Cyberworlds Lausanne Switzerland Nov 28 29 2006 E Rahmand and P A Bernstein A survey of approaches to automatic schema matching VLDB Journal vol 10 no 4 2001 pp 334 350 P Shvaiko and J Euzenat A survey of schema based matching approaches Journal on Data Semantics IV LNCS 3730 2005 pp 146 171 E Lai Mapping between HTML form and XML data Available at http www datamech com XMLForm 2006 O Chipara and A Slominski Xydra An automatic form gen erator for web services Extreme Computing Lab Available at http www extreme indiana edu xgws xydra R Raudjrv Dynamic Schema Based Web Forms Generation in Ja
72. va Master Thesis 2010 F Ameri and D Dutta A match making methodology for supply chain de ployment in distributed manufacturing environments Journal of Computing and Information Science in Engineering vol 8 no 1 2008 SU 24 The Open Application Group Open Application Group IntegrationSpecifica tion version 8 0 2002 25 Cay S Horstmann and Gary Cornell Core Java Volume I Fundamentals Prentice Hall Sun Microsystems Press 2008 26 David Stephenson XML Schema best practices Hewlett Packard Develop ment Company L P December 2004 27 C Charras and T Lecroq Handbook of Exact String Matching Algorithm College Publications February 2004 28 E Ukkonen Approximate string matching with q grams and maximal matches Theoretical Computer Science vol 92 no 1 1992 pp 191 211 29 es Grzegorz Kondrak N gram similarity and distance In Proc of International Conference on String Processing and Information Retrieval 2005 pp 115 126 30 J Kim Y Peng N Ivezic and J Shin An Optimization Approach for Semantic based XML Schema Matching International Journal of Trade Eco nomics and Finance vol 2 no 1 2011 31 A L Dulmage and N S Mendelsohn Coverings of bipartite graphs Canadian Journal of Mathematics vol 10 1958 pp 517 534 32 W B Douglas Introduction to Graph Theory Prentice Hall Chapter3 1999
73. wly created form components to extend the base form We compared four form component search methods and the result shows that the synonym based search method has the best overall perfor mance XDF transforms data inputted on the form into XML instances which will 59 be used in the supplier discovery process We compared three supplier searching methods including Keyword based search Ontology based search and XML based search XML based search takes XML instances generated from XDF as input The experimental results demonstrate that the eXtensible Dynamic Form is valuable for facilitating the supplier discovery process and in turn improving the search accuracy 3 2 Future Works There are several paths we would like to explore in the future First although synonym based search employed by XDF is faster than WordNet based search its efficiency will still suffer when the number of words inputted by the user is large More efficient algorithms need to be developed and integrated into the synonym based search method Second it is necessary to investigate how to maintain and utilize user defined schemas to improve the reusability of XDF The schema merg ing algorithms and social network techniques can be considered Finally Schemas on manufacture domain used in XDF are quite simple and thus they are inefficient to guide users to input useful information for large scale complex real world ap plications Richer manufacturing sc
74. y different circumstances of business information T his com ponent includes more common or general information than manufacturing domain specific information such as address party and id Pre defined component provides the basic form components defined by man ufacturing domain engineers User defined component provides the user defined form components which can be defined and shared by users A user can search and reuse the user defined form components defined by other users Or the user can create own form compo nents Our system will be updated soon to support the social management capability for users to create publish share and reuse their form components A 1 Basic Form Interface XDF helps you to construct a tree like data form structure using the reusable form components Each form component can be extended by attaching or detaching other form components as its children This section explains the usage of basic form interfaces for XDF A 1 1 Navigation Interface The navigation sidebar assists users to visit different parts of the XDF web site easily 50 GENERATE XML INSTANCE SAVE DYNAMIC FORM Y Basic Information Vertical CNC Milling Basic Info name Vertical CNC Milling Y Service List Vertical CNC Milling edit delete Description Water jet cuttng edit delete Horizontal CNC Milling edit delete gt Keywords CNC Turing edit delete P SpecilityGroup gt FacilityGro
75. ype Form Component with Complex Content Form Component with List carmona a ac de Form Component with Group Wildcard Form Component TASC OUI se arc Ge Jo os as Ge ee RT Example for Form Extending sele oe ee ewe ow eH aes Example for Form Component Creation String to String Matching to Word to Word Matching DVOOMVI Cbs Lar ee ae A GE bee ESS Comparison of Computing Time 004 XML Instance Generation 0 0 00 ee ee ee es 1V Chapter 1 Introduction 1 1 Challenges in Supplier Discovery In todays dynamic manufacturing industry discovery of manufacturing sup pliers henceforth supplier discovery is essential to build a flexible network of suppliers in supply chain 1 To utilize the supplier discovery several electronic marketplaces e marketplaces such as Thomasnet mfg com and GlobalSpec have been established In general supplier discovery function involves two steps The first step is to collect supplier capabilities and customer requirements henceforth collect function In the traditional e marketplaces online forms interfaces 2 3 4 are typically used for the collect function 5 However those forms are mostly fixed and pre defined so they are not flexible enough to capture a variety of requirements and capabilities in a structured way Different users suppliers or customers often use different terminologies structures and semantics to represe

APPROVAL SHEET

Contents

Download Pdf Manuals

Related Search

Related Contents