建构CMMI知识地图.ppt

《建构CMMI知识地图.ppt》由会员分享，可在线阅读，更多相关《建构CMMI知识地图.ppt（74页珍藏版）》请在三一文库上搜索。

1、建構建構 CMMI知識地圖知識地圖 李健興 長榮大學資訊管理系副教授 兼資訊工程系籌備處主任 2004/11/17 1 Outline oIntroduction oThe Structure of Ontology oOntology-based Knowledge Management System oOntology Construction oCMMI Ontology oCMMI Assistant Tools oCMMI Ontology Extraction oFuture Works 2 Introduction 3 Ontology (知識地圖 ) oThe ontology

2、 is a collection of key concepts and their interrelationships collectively providing an abstract view of an application domain. oAn ontology is a formal, explicit specification of a shared conceptualization. nConceptualization nExplicit nFormal 4 Ontology (知識地圖 ) oOntologyexplicit formal specificati

3、ons of the terms in the domain and relations among them. oAn ontology contains a hierarchy of concepts within a domain and describes each concepts property through an attribute-value mechanism. oRelations between concepts describe additional logical sentence. 5 Ontology (知識地圖 ) oThe main application

4、 areas of ontology technology nKnowledge management nWeb commerce nElectronic business nDatabase design nNatural language processing nMulti agent system n 6 飛機 航空公司 : 班機號碼 : 時間: 速度: 價格: 交通資訊 台北台南台北高雄台北澎湖 自行開車 路線: 時間: 火車 班次: 車種: 時間: 速度: 價格: 搭巴士 巴士公司 : 路線: 時間: 價格: Example 搭船 船公司 : 路線: 時間: 價格: 7 Ontolo

5、gy Example 發佈、 表示 導致、造成 、帶來 氣象 影 響 向、往 帶來、引進 氣象報導氣象百科天文. . . . . . 寒流颱風降雨. . . . . . . . . . . . Relation Association 中央氣象局 /氣象 局 型態：預報人員 、 天氣圖 表示、警告、評 估 颱風 編號 :*(Neu)號 中心位置： :*(Nc)(Ncd)(Neu)(Nf ) 強度 :輕度颱風 型態 :暴風圈 來襲、形成、登陸 降雨 降雨量 *(Neu)公釐 累積雨量 *(Neu)公 釐 種類 :大雨、陣雨、 大雷雨、豪雨 、豪大雨 型態：雨量、打雷 發生、襲擊、增加 移動方向

6、方向 :東方、南方 西北方、東 南方 移動、靠近、前 進 氣流 型態 :西南氣流、 冷氣流 接近、影響、流 動 農林漁牧業 型態 :漁港、農田 、農作物、 魚貨量 避風、休耕 地區 區域 :山區、平地、 台灣、中部、 東半部 各縣市：台北市、 台 南縣 海域 :東海、南海 海岸 :西海岸、沙岸 呈現、滯留、徘徊 災害 型態 :水災、旱象、 土石流、山崩 、洪水、房屋 倒塌、河水暴 漲、落石、雷 擊、霜害 來襲、形成、登陸 氣壓 型態 :副熱帶高氣 壓、熱帶性 低氣壓 增強為、逼近 發 生 導 致 造 成 民眾 /人民 型態 :人數 注意、受困 提 醒 時間 型態 :最近、昨日 、 今日、白天、

7、 午後 根據、開始 8 DAML+OIL format 9 Characteristics of Ontology oFormal Semantics oConsensus of terms oMachine readable and processable oModel of real world oDomain specific 10 Reasons to Develop Ontologies oTo share common understanding of the structure of information among people or software agents. oTo

8、 enable reuse of domain knowledge. oTo make domain assumptions explicit. oTo separate domain knowledge from the operational knowledge. oTo analyze domain knowledge. 11 Process of Developing an Ontology oDeveloping an ontology includes: nDetermine the domain and scope of the ontology. nConsider reusi

9、ng existing ontologies. nEnumerate important terms in the ontology. nDefine classes in the ontology and arrange the classes in a taxonomic (subclass-superclass) hierarchy. nDefine attribute and describe allowed values for these attribute. nFill in the values for attribute for instance. 12 Ontology L

10、earning Process 13 The Structure of Ontology 14 The three-layered object- oriented ontology Domain Category 1Category 2Category 3Category k Concept 1 Attributes 1 Operations 1 Concept 2 Attributes 2 Operations 2 Concept 3 Attributes 3 Operations 3 Concept n Attributes n Operations n Concept 4 Attrib

11、utes 4 Operations 4 Concept 5 Attributes 5 Operations 5 Concept 6 Attributes 6 Operations 6 Concepts Set AssociationGeneralization Aggregation 15 The four-layered Object- Oriented Ontology Domain Category 1Category 2Category k Concept 3 Attributes 3 Operations 3 Concept 1 Attributes 1 Operations 1 C

12、oncept 2 Attributes 2 Operations 2 Concept n Attributes n Operations n Concept 4 Attributes 4 Operations 4 Class-layer Instance 3 Attributes 3 Operations 3 Instance 1 Attributes 1 Operations 1 Instance 2 Attributes 2 Operations 2 Instance m Attributes m Operations m Instance 4 Attributes 4 Operation

13、s 4 Instance-layer Association Generalization Aggregation Instance-of 16 The four-layered News Ontology (cont.) Domain Event 1Event kEvent 3Event 2 Candidate Chinese Terms Extended Concept Relation Category 1Category 2Category q Association Operation Templates Attributes Event Concept 1 Operation At

14、tributes Object Concept 1, m Operation Attributes Event Concept 1, 1 Operation Attributes Event Concept P Operation Attributes Event Concept P,1 Operation Attributes Object Concept P,n 17 The four-layered News Ontology 發佈、 表示 導致、造成 、帶來 氣象 影 響 向、往 遠離、移動 帶來 、 引進 氣象報導氣象百科天文 . . . . . . 寒流颱風降雨 . . . . .

15、 . . . . . . . Relation Association 表示、警告、評估 型態：預報人員、 天氣圖 中央氣象局 /氣象 局 來襲、形成、登陸 編號:*(Neu)號 中心位置： :*(Nc)(Ncd)(Neu)(Nf) 強度:輕度颱風 型態:暴風圈 颱風 發生、襲擊、增加 降雨量 *(Neu)公釐 累積雨量 *(Neu)公 釐 種類:大雨、陣雨、 大雷雨、豪雨 、豪大雨 型態：雨量、打雷 降雨 移動、靠近、前進 方向:東方、南方 西北方、東 南方 移動方向 接近、影響、流動 型態:西南氣流、 冷氣流 氣流 避風、休耕 型態:漁港、農田 、農作物、 魚貨量 農林漁牧業 呈現、滯留、

16、徘徊 區域:山區、平地、 台灣、中部、 東半部 各縣市：台北市、台 南縣 海域:東海、南海 海岸:西海岸、沙岸 地區 來襲、形成、登陸 型態:水災、旱象、 土石流、山崩 、洪水、房屋 倒塌、河水暴 漲、落石、雷 擊、霜害 災害 增強為、逼近 型態:副熱帶高氣 壓、熱帶性 低氣壓 氣壓 發 生 導 致 造 成 注意、受困 型態:人數 民眾/人民 提 醒 根據、開始 型態:最近、昨日 今日、白天 午後 時間 影 響 恢 復 出現、發生 18 Fuzzy Ontology (cont.) Domain Category 2 C : Concept A : Attribute O : Operatio

17、n Category 1Category 3Category k Class-layer C1;C1E1,C1E2,C1Ep AC11,AC12 ,AC1q1 Cm;CmE1,CmE2,CmEp ACm1,ACm2,ACmqm OCm1 ,OCm1,OCmqm C2;C2E1,C2E2,C2Ep AC21,AC22 ,AC2q2 C3;C3E1,C3E2,C3Ep AC31,AC32 ,AC3q3 C4;C4E1,C4E2,C4Ep AC41,AC42,AC4q4 OC41,OC41,OC4q4 C5;C5E1,C5E2,C5Ep AC51,AC52,AC5q5 OC51,OC51,OC5q5

18、 Association Aggregation Generalization Event E1Event E2Event E3Event Ep OC11 ,OC11,OC1q1OC21 ,OC21 ,OC2q2OC31 ,OC31 ,OC3q3 LBR LNR 19 Fuzzy Ontology 20 Ontology-based Knowledge Management System 21 CREDIT Research Center oLocated at National Cheng Kung University. oSupported by Walsin Lihwa Group.

19、(2001-2004) oContain three main research groups. oMore than 10 professors and 50 Ph.D or master students. 22 CREDIT KM System (cont.) oProcess Management nWorkflow BPM + Web service nCMMI (中小企業) nMobile Workflow oDocument Management nKnowledge Map nQ and A nFAQ nPersonalization nSemantic Search nKno

20、wledge Update 23 CREDIT KM System oMeeting Management nMeeting Scheduling nMeeting Notification nMeeting Follow-up oMessage Management nBBS nNotification nDirectory Service for Message Delivery 24 Enterprise Networking Resource Non-structured Data Internet/Intranet News/Documents XML-based E-documen

21、ts CMMI-based CREDIT K.M. System Personalized Service Ontology Repository Automatic Classification Service Document Abstraction Service Document Repository Workflow Service Intelligent Mobile Delivery Service On-line Tracking Service Personal Ontology End User Ontology Construction Service Meeting S

22、cheduling Service Semantic Search Service CMMI Assistant Service 25 Semantic Search Service(cont.) oHuman-readable nHTML oMachine-readable nXML oMachine-understandable nSemantic Web with Ontology (RDF,DAML+OIL) 26 Semantic Search Service oKeyword-based search nSingle-word query nContext query nBoole

23、an query oConceptual search nConceptual query nNatural language query oSemantic search nOntology-reasoning query 27 Why Semantic Search? oMass information make user confused, current search engines are not good enough. (e.g. 腦科 v.s. 電腦科學) oQuality is more important than Quantity oSearch by “what the

24、y means“ not just “what they say“ oThe user who has no idea about domain terminologies cant find information easily. 28 XML file Repository Index Repository Personal Thesaurus Repository Ontology Repository CKIP Repository Repository WWW Information Retrieval Agent Indexing and Gathering statistics

25、Natural Language Processing Query Query Inference Query Personalization Query Results End User Parsing and Transforming formats Clustering Document Preprocessing Query processing Semantic Search Service Architecture 29 Personalized Service oMake a specific information service that can adapt to the b

26、ehavior of each user. oProvide a mechanism that can observe and analyze the browsing behavior of each user. oProduce a structure with personal custom and preferences for other services using. 30 User Used Behavior Functionalization End User WWW Personal Log Files Repository Browsed Contents Reposito

27、ry Log File Recording Engine Web Content Recording Engine Function of Browsing Frequency Function of Browsing Time Fuzzy Inference Preference Degree Content Conceptualization & Weighting Present Browsing Behavior Concepts and Weights of Browsed Content Sequence of User Browsing Behavior s1s2s3sn Use

28、r Behavior & Browsed Content Analysis Domain Ontology User Behavior & Browsing Content Analysis Additional Weighting of Related Sequence Concepts Domain Ontology Personal Ontology Personal Ontology 31 User Behavior Analysis oIn order to find out users favor tendency, the first job is analyzing the h

29、abitual behavior of reading. oConsider two features: reading time and reading frequency. oConsider reading time is related with content length, change the feature to 32 Browsing Frequency Browsing Time Content Length Personal Log Files Repository Browsed Contents Repository Feature Data Data Sorting

30、 Data Clustering Sorted Data Functionalization Function of Feature Data Feature Data Functionalization Feature Data Processing Frequency Feature Data Browsing Time Feature Data Feature Data Functionalization Function of Browsing Frequency Function of Browsing Time Personal Ontology 33 Question & Ans

31、wer System oQuestion analysis n5W1H owhat, who, when, where, why, and how. nIndirectly question & other oYesNo questionetc. oAnswer analysis nQuestion type o5W1H nDomain oDomain knowledge 34 Question & Answer System Question Ontology Question WhereWhat How Answer Ontology KM PMworkflow Q&A KM workfl

32、ow Ontology Q&A Search engine Ontology Domain Knowledge extraction & learning process Question Answering Subsystem Knowledge Extraction Subsystem Receive User query Documents Return Answer Ontology supervision Question & Answer Knowledge Base Answering processUser query process User 35 Question & An

33、swer Knowledge Base (cont.) oDomain ontology nObject-oriented ontology oQuestion ontology nThe knowledge of question domain nTo Classify and extract question oAnswer ontology nThe knowledge map of Q&A knowledge base 36 Question & Answer Knowledge Base oAlternation Rule nMorphological nLexical nSeman

34、tic oOntology supervision nOntology management nOntology inference 37 Internet e-News Retrieval Agent Fuzzy Inference Agent Chinese e-News Summar y Chinese e-News Ontology Chinese e-News Summary Repository Real-time e-News Repository e-News Repository G U I POS Tagger (CKIP) Chinese Term Filter Docu

35、ment Processing Agent OFEE Agent Extracted-Event Ontology PDA Cell Phone Notebook Event Ontology Filter Sentence Rule Base Sentence Generation Agent Summarization Agent Document Abstraction Service 38 Meeting Scheduling Service Meeting Host Meeting Scheduling Decision Support System (MSDSS) Group Ca

36、lendar Data Base (GCDB) Genetic Learning Agent (GLA) Fuzzy Inference Agent (FIA) Meeting Information Knowledge Base (MIKB) Personalized Knowledge Base (PKB) Evaluation Module Meeting Negotiation Agent (MNA) user names proper time with work priority Invitees Devices Cell Phone PDA Notebook Desk Compu

37、ter IFA 39 The Architecture of Fuzzy Inference Agent 40 The Flow Chart of Genetic Learning Agent Initiation SelectionCrossover Mutation Current Population New Population Evaluation replace Personalized Knowledge Base (PKB) Meeting Information Knowledge Base (MIKB) Fuzzy Inference Agent (FIA) Start e

38、litism 41 工作 流程 部門 角色 文件 管理者使用者 新增、刪除 修改架構 新增、刪除 修改流程 組織 架構 人員 (人事部 ) 載入、更新 部門角色 簽核 發文 加簽 定義會簽 文件控管 與分析 定義職務 代理人 新增、刪除 修改角色權限 定義角色 權限 組織設計師組織設計師 流程設計師流程設計師 流程管理員流程管理員 工作管理員工作管理員 Workflow Engine 解析流程 記錄變動 執行多條 流程 logs DB Workflow Service 42 Ontology Construction 43 Automatic Construction of OO Ontolo

39、gy oUse object-oriented data model to represent ontologies. oFollow object-oriented analysis procedure to build ontologies. oApply natural language processing technology to extract key terms from documents. 44 Automatic Construction of OO Ontology oApply SOM clustering technology to find concepts an

40、d instances. oApply data mining technology and morphological analysis to extract attributes, operations, and associations of instances. oAggregate attributes, operations, and associations of instances to class. 45 Structure of Object-Oriented Ontology Domain Category 1Category 2Category k Concept 3

41、Attributes 3 Operations 3 Concept 1 Attributes 1 Operations 1 Concept 2 Attributes 2 Operations 2 Concept n Attributes n Operations n Concept 4 Attributes 4 Operations 4 Class-layer Instance 3 Attributes 3 Operations 3 Instance 1 Attributes 1 Operations 1 Instance 2 Attributes 2 Operations 2 Instanc

42、e m Attributes m Operations m Instance 4 Attributes 4 Operations 4 Instance-layer Association Generalization Aggregation Instance-of 46 Concepts Class and Instance 47 Specific Class News Documents (Training Data) Part-of-speech Tagger (CKIP) Nouns SetVerbs Set Chinese Electronic Dictionary Segmentat

43、ion Standard Dictionary Academia Sinica Balanced Corpus Term Analyzer Data Mining Concepts Clustering Processing Association Rule Result Chinese Electronic Dictionary Academia Sinica Balanced Corpus Concepts Construction Agent Operations Construction Agent Attributes Construction Agent Relations Con

44、struction Agent Ontology Construction Procedure Domain Ontology Concepts Set Domain Ontology Construction (I) Refining Tagging Stop Word Filter 48 Domain Ontology Construction(II) Episodes Document Pre-processing Domain Ontology Nouns Sentences Concepts Concept Clustering Episode Extraction Attribut

45、es, Operations, Associations Extraction DAML+OIL Format Special Domain Documents Chinese Dictionary Data Flow Control Flow 49 Feature Term Pre-processor Episode Net Repository Episode Net Extractor Episode Extractor Concept Extractor Attributes- Operation- Association Extractor Episodes Repository C

46、oncepts Repository Chinese Domain Ontology Ontology Construction Agent Domain Term Combination Processor HowNet New Chinese Term Repository Knowledge Base Part-Of-Speech Tagger Verbs Repository Nouns Repository Data Flow Chinese Documents Control Flow Domain Expert Domain Ontology Construction(III)

47、50 Episode Extractor (cont.) oAn episode is a partially ordered collection of events occurring together. 51 德國門將卡恩贏得本屆世足賽代表最佳球員的金球獎。 德國(Nc) 門將(Na) 卡恩(Nb) 贏得(VJ) 本(Nes) 屆(Nf) 世足賽(Nb) 代 表(Na) 最佳(A) 球員(Na) 的(DE) 金球獎(Nb)。(PERIODCATEGORY) (德國, Nc, 1) (門將, Na, 2) (卡恩, Nb, 3) (贏得, VJ, 4) (世足 賽, Nb, 5) (代表,

48、 Na, 6) (球員, Na, 7) (金球獎, Nb, 8) 德國(Nc)_門將(Na)_卡恩(Nb) Germany_keeper_Oliver Kahn 卡恩(Nb)_贏得(VJ)_金球獎(Nb) Oliver Kahn_took_Golden Ball POS Tagger Stop Word Filter Episode Extractor Episode Extractor oThe following shows an example of extraction of episode from a sentence. 52 CMMI Ontology 53 The defini

49、tion of CMMI oThe CMMI, Capability Maturity Model Integrated, is a model for improving organizations processes and ability to manage the development, acquisition, and maintenance of products of services. 54 Maturity Level 2 Process Area 1（Requirement Management） Process Area 2（Project Planning） Process Area 3（Project Monitoring and Control） Process Area 4（Supplier Agreement Management） Process Area 5（Measure