以工作觀為基礎之知識支援模式與系統:工作相關知識遞送與分享

全文

(1)國立交通大學資訊管理研究所博士論文. 以工作觀為基礎之知識支援模式與系統: 工作相關知識遞送與分享 Task-based K-Support Model and System: Delivering and Sharing Task-relevant Knowledge. 研究生：吳怡瑾指導教授：劉敦仁. 中華民國九十五年一月 i.

(2) 國立交通大學資訊管理研究所博士論文. 以工作觀為基礎之知識支援模式與系統: 工作相關知識遞送與分享. Task-based K-Support Model and System: Delivering and Sharing Task-relevant Knowledge. 研究生：吳怡瑾研究指導委員會：陳彥良博士魏志平博士羅濟群楊千. 博士博士. 指導教授：劉敦仁. 博士. 中華民國九十五年一月. ii.

(3) 以工作觀為基礎之知識支援模式與系統: 工作相關知識遞送與分享 Task-based K-Support Model and System: Delivering and Sharing Task-relevant Knowledge. 研究生：吳怡瑾. Student：I-Chin Wu. 指導教授：劉敦仁. Advisor：Dr. Duen-Ren Liu. 國立交通大學資訊管理研究所博士論文. A Dissertation Submitted to Institute of Information Management College of Management National Chiao Tung University in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy in Information Management January 2006 Taipei, Taiwan, Republic of China 中華民國九十五年一月. iii.

(4) 博碩士論文授權書 (國科會科學技術資料中心版本 92.2.17). 本授權書所授權之論文為本人在＿國立交通大學＿＿＿大學(學院)＿資訊管理＿＿＿系所＿＿九十四＿＿＿＿學年度第＿一＿＿學期取得＿博＿＿士學位之論文。論文名稱：以工作觀為基礎之知識支援模式與系統：工作相關知識遞送與分享 1. □同意 ;不同意（政府機關重製上網）本人具有著作財產權之論文全文資料，授予行政院國家科學委員會科學技術資料中心、國家圖書館及本人畢業學校圖書館，得不限地域、時間與次數以微縮、光碟或數位化等各種方式重製後散布發行或上載網路。本論文為本人向經濟部智慧財產局申請專利（未申請者本條款請不予理會）的附件之一，申請文號為：＿＿＿＿＿＿，註明文號者請將全文資料延後半年再公開。 2.;同意 □不同意（圖書館影印）本人具有著作財產權之論文全文資料，授予教育部指定送繳之圖書館及本人畢業學校圖書館，為學術研究之目的以各種方法重製，或為上述目的再授權他人以各種方法重製，不限地域與時間，惟每人以一份為限。上述授權內容均無須訂立讓與及授權契約書。依本授權之發行權為非專屬性發行權利。依本授權所為之收錄、重製、發行及學術研發利用均為無償。上述同意與不同意之欄位若未鉤選，本人同意視同授權。指導教授姓名：. 劉敦仁. 研究生簽名：（親筆正楷）. 學號：8934502 （務必填寫）. 日期：民國九十五年一月十二日 1. 本授權書（得自 http://sticnet.stic.gov.tw/sticweb/html/theses/authorize.html 下載或至 http://www.stic.gov.tw 首頁右下方下載）請以黑筆撰寫並影印裝訂於書名頁之次頁。 2. 授權第一項者，請確認學校是否代收，若無者，請個別再寄論文一本至台北市(106-36)和平東路二段 106 號 1702 室國科會科學技術資料中心王淑貞。(本授權書諮詢電話:02- 27377746) 3. 本授權書於民國 85 年 4 月 10 日送請內政部著作權委員會(現為經濟部智慧財產局)修正定稿， 89.11.21 部份修正。 4. 本案依據教育部國家圖書館 85.4.19 台(85)圖編字第 712 號函辦理。. iv.

(5) Task-based K-Support Model and System: Delivering and Sharing Task-relevant Knowledge Student: I-Chin Wu. Advisor: Dr. Duen-Ren Liu Institute of Information Management National Chiao Tung University. ABSTRACT In task-based business environments, a pertinent issue in deploying knowledge management system (KMS) is providing task-relevant information (codified knowledge) to fulfill the information needs of knowledge workers. Historical codified knowledge, i.e. experiences and know-how extracted from previous task executions, provides valuable knowledge for knowledge workers to accomplish tasks successfully. Accordingly, a repository of structured and explicit knowledge, especially in document form, is a widely adopted codification-based strategy for managing knowledge in KMS. This work first discusses the issue of managing codified knowledge by building the task-oriented repository from the perspective of business task. To organize and manage task-relevant information, the repository is constructed with support from domain ontology (topic taxonomy) to effectively utilize codified knowledge. Thus, providing effective knowledge retrieval function to mitigate the difficulty of accessing knowledge items from the knowledge repository is a challenging work. Accordingly, a task-based knowledge support model is proposed to tackle the problem. The proposed model proactively delivers task-relevant codified knowledge and promotes knowledge sharing among knowledge workers in task-based business environments. A novel task-relevance assessment approach is proposed to identify the knowledge worker’s information needs on tasks, for brevity, task-needs. The proposed approach generates task profiles via the collaboration of knowledge workers to analyze the relevance of tasks and codified knowledge. The approach can. v.

(6) alleviate the problem of accessing needed knowledge items from vast amounts of codified knowledge. Moreover, an adaptive task-based profiling approach and a task peer-group analytical method are proposed to track workers’ dynamic task-needs and identify workers’ task-based peer-groups p. Knowledge workers can obtain task-relevant knowledge with the aid of task-based profiles and peer-groups. Furthermore, we seek to extend and refine our model to resolve long-term knowledge support problem. According to our empirical investigation, knowledge workers engaged in knowledge intensive task usually have different information needs during the long-term task performance. That is, another challenge of deploying KMS is to support task-relevant knowledge based on workers’ task-needs at different task progress, i.e., stages or milestones. Accordingly, we proposed a task-stage knowledge support model that incorporates the information-filtering model with the identification of worker’s task-stage. A correlation analysis method is proposed to identify a worker’s task-stage, and an ontology-based topic discovery method is proposed to determine a worker’s task-needs for specific topics of stage. Consequently, the system can be tailored to support long-term task performance. A task-based K-Support portal is developed to facilitate knowledge reuse and further to streamline task execution. The portal is grounded in a research institute to support the execution of knowledge-intensive task by stimulating the operation of knowledge delivering and sharing. Moreover, various experiments have been conducted to evaluate the proposed model. The experimental results reveal that the proposed model and system can provides knowledge support in task-based environments effectively. Keywords: Knowledge management system, Task-relevant knowledge, Codified knowledge, Task-relevance assessment, Adaptive task profile, Knowledge delivery, Knowledge sharing, Task-stage, K-Support portal. vi.

(7) 以工作觀為基礎之知識支援模式與系統: 工作相關知識遞送與分享研究生：吳怡瑾. 指導教授：劉敦仁博士國立交通大學資訊管理所. 摘要建構知識管理系統已是企業組織有效管理企業知識，獲取產業競爭優勢的重要策略。而企業主要是以工作為基礎來進行企業活動之運作與管理，組織人員執行各項工作以達成企業之營運目標。在以工作為基礎之企業環境，考量組織工作特性，設計適合的知識推薦機制，以提供組織人員工作相關之知識物件與資訊，是建構知識管理系統之重要議題。一般而言，在各類知識物件中，文件為將知識外顯化的重要方式之一；此外，文件除提供豐富之資訊並且也是增加速度最為可觀之知識物件。因此，企業若能將各式知識物件以結構化方式存放至知識庫並使之外顯化，勢必能有效保存與提供組織知識資產。本研究主要設計以工作為基礎的主題分類架構(task domain ontology)…，並引入模糊分類方法，將企業內的各項知識物件與工作，配合該主題分類架構加以分類與整理。此外，為支援知識工作者克服執行工作中所遭遇之困難，本研究提出以工作為基礎之知識支援模式，預期達到有效知識彙集、遞送與分享之目的。本研究首先提出系統化的工作相關知識評估機制，透過工作者間之協同合作以支援其資訊需求，並整合工作相關知識評估機制於知識支援系統中，以協助組織人員透過工作特徵檔擷取工作所需的知識。該工作相關評估機制，分析工作與知識物件之相關性並建置工作特徵檔(task profile)，以協助組織人員透過工作特徵檔擷取工作所需的知識物件，預期協助知識工作者從大量知識物件中有效獲取工作相關知識，克服工作執行中所遭遇之困難。在此基礎之上，我們更藉由知識工作者資訊回饋過程，提出修正工作特徵檔之方法外，依該工作特徵檔，提出工作社群網路分析與建構方法，並探討與評估知識工作者之間互動. vii.

(8) 所構成的社群網路如何促成知識遞送與分享。研究內容主要包括：(1)提出適性化的工作特徵模式，藉由工作相關回饋機制修正工作特徵檔，以描述知識工作者之動態性工作資訊需求；(2)提出工作同好群組分析法，依據工作者特徵檔分析知識工作者資訊需求之相似性，並建立工作社群網路。在我們後續的研究中，發現工作者對於知識密集性工作之資訊需求是動態的，會隨著時間與環境而演化改變。因此，有效之知識支援需提供適性化機制以依據工作者之動態需求提供相關知識；此外，工作之執行，常需逐步執行階段性任務以完成工作，而不同階段有不同之工作資訊需求。根據組織工作特性而由系統主動提供工作相關知識的相關研究並未考慮工作之階段性；因此，本研究進一步改良先前知識支援模式，提出工作階段性為基礎之工作相關知識支援模式與系統架構。研究內容主要包括：（1）根據知識工作者不同時間點的工作特徵檔，運用相關係數分析法，偵測工作者目前之工作階段；(2)以組織之工作主題分類架構為基礎，分析知識工作者於工作執行中之主題變換情形，以判別知識工作者現階段資訊需求主題；(3)該模式依據作者之工作階段與需求主題之變換，進而調整其資訊需求特徵檔，提供符合工作階段性之相關知識。本文並依所設計之知識支援模式而設計實驗，以驗證方法於提供知識支援之有效性。此外，並以物件導向方式實作以工作為基礎之知識支援系統，建構協同合作之工作環境，以提供有效的工作相關知識遞送與分享。該系統落實在一研究單位，藉由使用者滿意度回饋以評估系統之有效性。研究結果顯示該知識支援模式與系統能有效達成知識遞送並促進組織成員之知識分享。. 關鍵字：知識管理系統、工作相關知識、編撰知識、工作相關評估、適性化工作特徵檔、知識遞送、知識分享、工作階段、知識支援平台. viii.

(9) 誌謝. 論文即將付梓，揮別博士生涯之際，心中有所感動、不捨與責任。當初所持有的理想，雖有未盡之憾，但學生仍會持有夢想，堅定樂觀向前邁進。博士班期間，感謝指導教授劉敦仁老師，對怡瑾論文悉心的指導，並提醒我適時的沉澱，培養我獨立研究的態度。感謝口試委員羅濟群老師、楊千老師、陳彥良老師以及魏志平老師與指導教授，於口試期間對論文的建議與指正，以及對怡瑾生涯發展的關懷，謝謝老師們。此外，感謝亦師亦友的蕭敏雄老師在語文對學生的啟迪、感謝曾國雄老師不吝惜的指導與慈愛的笑容、洪永城老師、吳誠文老師對我的鼓勵與關懷。每個擦身而過的緣分或深或淺，都給我無比的力量，怡瑾銘記在心。特別感謝博士班期間，嘉源與 K-support 的朋友們昆學、北晨、韋孝，在研究對我的勉勵與協助，並分擔我許多實驗室事務，謝謝你們的善解，願我們都能以開懷的氣度面對一切可能。感謝籌尹學姊、民新與聰洲學長、孟蓉、秋婷、文彥、政龍、之怡、春鋒、栩嘉、皇志…等可敬可愛的朋友們，大方的分享生活與研究的點點滴滴；多才多藝的博士班同學們昶瑞…等；教育學程的朋友心玫、金鳳…等；曾門的朋友宜中學長…等；以及交大的朋友們，謝謝你們貼心的關懷，幫助我渡過生活與研究的瓶頸。對於陪我走過十年寒暑的男友俊佑，內心有無以言喻的感動，喜歡與你ㄧ起發現生活的無限可能與享受生命的喜悅!!最後，謹將此論文獻給我摯愛的外婆與家人，感謝您們賜與的福份，我會努力當一個可愛而堅強的孩子。. 吳怡瑾 (nancy) Phinally Done !!. ix. 于交大. 2005/1/12.

(10) Contents ABSTRACT. ................................................................ v. 誌謝. ............................................................... ix. Contents. ................................................................ x. List of Tables ............................................................ xiii List of Figures ...........................................................xiv Chapter 1. Introduction ............................................ 1. 1.1 Research background and motivation ...........................................1 1.2 Research objectives and tasks ...................................................... 3 1.3 Contributions ............................................................................... 4 1.4 Content organization.................................................................... 6. Chapter 2. Related Work ........................................... 8. 2.1 Knowledge management in task-based working. environment ... 8. 2.1.1 Knowledge management systems and information technology....... 8 2.1.2 Task-based knowledge retrieval........................................................ 9 2.1.3 Knowledge sharing in community of practices............................... 11 2.2 Text mining technique for codified knowledge management.......12 2.2.1 Information retrieval in vector space model................................... 12 2.2.2 Relevance feedback techniques ....................................................... 13 2.3 User modeling for information filtering......................................13 2.3.1 Researches of information filtering to support information needs 13 2.3.2 User modeling technique to support knowledge-intensive tasks .. 15. Chapter 3. Task-based Knowledge Support .............. 17. 3.1 Rationale to design task-based knowledge support ..................... 17 3.1.1 Task-based organizational environment......................................... 17 3.2 Framework of task-based knowledge support .............................18. Chapter 4 Task-Oriented Information Repository: Managing Codified Knowledge..................... 21 4.1 Task-oriented information repository.........................................21 4.1.1 Extracting task corpus..................................................................... 21 4.1.2 Task categorization model ............................................................... 22 4.2 Domain ontology formalization.................................................. 24. Chapter 5. Collaborative Task-Relevance Assessment26. 5.1 Preliminary concepts and term definition.................................. 26 x.

(11) 5.1.1 Preliminary concepts ....................................................................... 26 5.1.2 Term definition................................................................................. 28 5.2 Process of task-relevance assessment ........................................ 29 5.3 Collaborative task-relevance assessment ................................... 32 5.3.1 Phase 1-Identifying referring tasks based on category assessment32 5.3.2 Phase2-Assessing the relevance of referring tasks ........................ 36 5.4 Task-based K-Support based on assessment ...........................................38 5.4.1 Task profile generation by relevance feedback technique ............. 38 5.4.2. K-Support: task-based knowledge retrieval ................................. 39. 5.5 Experimental setup .....................................................................41 5.5.1 Overview of experiments ................................................................. 41 5.5.2 Data, participants and evaluation metrics ..................................... 43 5.5.3 Parameter selection ......................................................................... 46 5.6 Experimental results and implications........................................47 5.6.1 Experiment one: effect on fuzzy linguistic assessment.................. 47 5.6.2 Experiment two: effect on two-phase relevance assessment ......... 50 5.6.3 Experiment three: effect on collaborative assessment................... 52 5.7 Discussions ............................................................................... 53. Chapter 6 Disseminating and Sharing Task-relevant Knowledge .................................................. 55 6.1 Overview of K-Support model......................................................55 6.2 Adaptive task-based profiling approach ......................................57 6.2.1 Profile modeling and structuring .................................................... 57 6.2.2 Profile adaptation based on feedback analysis............................... 58 6.2.3 K-Delivery: Delivering codified knowledge proactively ................. 61 6.3 Peer-group analytical model ...................................................... 62 6.3.1 Establishing a user-user similarity matrix .................................... 62 6.3.2 Identifying task-based peer-groups................................................. 64 6.3.3 K-Sharing: Knowledge support from peer-group ........................... 65 6.4 Task-based K-Support portal ...................................................... 66 6.5 Experimental setup..................................................................... 66 6.5.1 Overview of experiments ................................................................. 66 6.5.2 Data, participants and evaluation metrics ..................................... 67 6.6 Experimental results and implications....................................... 69 6.6.1 Novelty of knowledge support ......................................................... 69 6.6.2 Quality of knowledge support.......................................................... 70 6.7 Discussions ............................................................................... 70. xi.

(12) Chapter 7. Task-Stage Knowledge Support .............. 72. 7.1 Task-needs evolution pattern modeling ......................................72 7.1.1 Task-stage Knowledge Support Module.......................................... 72 7.1.2 Task-needs evolution pattern modeling.......................................... 74 7.2 Changes of task-stage ..................................................................76 7.2.1 Stage identification process............................................................. 76 7.2.2 Sample analysis ............................................................................... 79 7.3 K-Support based on task-stage and task-needs topics ................. 80 7.3.1 Determination task-needs on topics................................................ 80 7.4 Knowledge support based on task-stage ..................................... 82 7.4.1 Profile adaptation ............................................................................ 82 7.4.2 Knowledge support .......................................................................... 83 7.5 Experimental setup .................................................................... 83 7.5.1 Overview of experiments ................................................................. 84 7.5.2 Data, participants and evaluation metrics ..................................... 86 7.6 Experimental results and implications....................................... 88 7.6.1 Experiment one: effect on task-stage identification....................... 88 7.6.2 Experiment two: effect on discovery of task-needs topics.............. 91. Chapter 8. K-Support System .................................. 96. 8.1 System architecture ................................................................... 96 8.2 System demonstration and scenario descriptions ...................... 98 8.2.1 K-Processing: Task relevant information processing ..................... 98 8.2.2 K-Assessment: Identifying task-relevant knowledge..................... 98 8.2.3 K-Delivery: Delivering codified knowledge proactively ................. 99 8.2.4 K-Sharing: Knowledge support from peer-group ......................... 100 8.3 Discussions ..............................................................................102. Chapter 9 9.1 Summary. Conclusions and Future Works ............. 104 ..............................................................................104. 9.2 Future works.............................................................................105. References. ............................................................ 108. Appendix A. Basic Concepts..................................... 115 A.1 Fuzzy Linguistic and Fuzzy Number........................................... 115 A.2 Fuzzy Relations ......................................................................... 116. Appendix B. Details of System Evaluation ............... 117. xii.

(13) List of Tables Table 1. Process of collaborative task-relevance assessment ................ 30 Table 2. Corresponding fuzzy numbers of linguistic term set by different evaluators............................................................................... 33 Table 3. Assess the relevance of executing task to categories ................ 34 Table 4. Relevant degree between tasks and categories ........................ 36 Table 5. Assessment on the relevance of positive referring tasks to the executing task......................................................................... 37 Table 6. Six selected executing tasks (on-going tasks) .......................... 44 Table 7. Result of knowledge support for task retrieval (B-RA vs.F-RA)48 Table 8. Result of knowledge support for task retrieval by ten novices. 49 Table 9.. Knowledge support for task-retrieval (2-F-RA versus F-RA) . 50. Table 10. Knowledge support for document retrieval (2-F-RA versus F-RA).....................................................................................................51 Table 11. Results of knowledge support ................................................ 53 Table 12. Users’ perceptions of information novelty ............................. 69 Table 13. Users’ perceptions of information quality.............................. 70 Table 14. Task stage identification rule................................................. 78 Table 15. Summation of Bit, j / Bi f, j in TRTWs of Trans3 ........................... 81 k. l. Table 16. Parameters adjustment across task-stages ............................ 82 Table 17. Experiments description........................................................ 85 Table 18. Result of knowledge support by stages (top-30 document support) ................................................................................ 89 Table 19. Result of knowledge support by stages under various top-N ...91 Table 20. Result of knowledge support by stages (experimental two) ... 93 Table 21. Average of likert scale value form system evaluation (Higher is better, range=1-6) ................................................................103 Table 22. Viewed / relevant of supporting items .................................. 117 Table 23. Irrelevant/ Normal Ratings of viewed items......................... 117. xiii.

(14) List of Figures Fig. 1. Task-based knowledge support .................................................... 7 Fig. 2. Framework of task-based knowledge support .............................19 Fig. 3. Example of domain ontology...................................................... 25 Fig. 4. Task-relevant knowledge source (explicit or tacit knowledge source) ......................................................................................41 Fig. 5. Experimental procedure ............................................................ 43 Fig. 6. Average recall-precision curves for experienced users and novices (B-RA vs. F-RA) ........................................................................ 49 Fig. 7. Average recall-precision curves for experienced users and novices (F-RA vs. 2-F-RA) ......................................................................51 Fig. 8. Interface of knowledge sharing (α=0.9) ..................................... 62 Fig. 9. Inferring similarity relationships based on workers’ task-needs 65 Fig. 10. Task need evolution module..................................................... 73 Fig. 11. Changes of task stages............................................................... 79 Fig. 12. Generality and specificity indicators ........................................ 80 Fig. 13. Result of knowledge support by averaging stages (performance value in y axes) ......................................................................... 88 Fig. 14. Result of knowledge support by averaging stages under various top-N ........................................................................................ 90 Fig. 15.. Result of knowledge support by averaging stages ................... 93. Fig. 16. Result of knowledge support by averaging stages under various top-N ........................................................................................ 95 Fig 17. System architecture ................................................................... 96 Fig. 18. Interface of K-Processing ......................................................... 99 Fig. 19. Interface of two-phase assessment ......................................... 100 Fig. 20. Interface of knowledge delivery .............................................. 101 Fig. 21. Six Degrees of relevance feedback ........................................... 101. xiv.

(15) Chapter 1 Introduction 1.1 Research background and motivation Deploying knowledge management systems (KMS) is an important strategy for enterprises to effectively managing business knowledge and gaining competitive advantage. The operations and management activities of enterprises are mainly based on tasks, in which organizational workers perform various tasks to achieve business goals [1][22][24][26]. Moreover, organizations try to maximize the use of knowledge assets to increase an organization’s profitability and productivity with the support of contemporary knowledge management tools. KMS employs Information Technologies (IT), such as document management and workflow management to facilitate the access, reuse and sharing of knowledge assets within and across organizations [17][39]. That is, the critical role of Information Technologies (ITs) is to assist knowledge workers to reuse valuable knowledge assets to carry out business tasks successfully [6][17][46]. Generally, ITs focus on explicit and tacit dimensions in knowledge management activities [28][39]. The former, explicit knowledge management, is achieved by a codified approach. Intellectual content codified into explicit form can facilitate knowledge retrieval and reuse [89]. Knowledge repository, knowledge-based systems, and knowledge maps are the supports for knowledge storage, organization and dissemination. And a repository of structured and explicit knowledge, especially in document form, is a widely adopted codification-based strategy for managing knowledge in KMSs [17][81][89]. The latter, tacit knowledge management, puts emphasis on dialoging via social networks to facilitate knowledge sharing. Knowledge expert directories, yellow pages, communities of practices and talk rooms, support interpersonal communication for knowledge sharing [3][41]. Notably, empirical findings indicate that codifying intellectual content into a knowledge repository makes workers highly exploit existing organizational resources [29][49]. Accordingly, knowledge (information) retrieval is considered a core component to retrieve codified knowledge in KMS. An effective knowledge retrieval function can mitigate the difficulty of accessing knowledge items from a knowledge repository and support the operation of knowledge-intensive work in business environments [24][27].. 1.

(16) In task-based business environments, an important issue of deploying KMS is providing task-relevant information (codified knowledge) to fulfill the information needs of knowledge workers during task execution. That is, effective knowledge management relies on understanding workers’ information needs on tasks, for brevity, task-needs. Recently, the information retrieval (IR) technique coupled with workflow management systems (WfMS) was employed to support proactive delivery of task-specific knowledge according to the context of tasks within a process [1][2][23][24]. The KnowMore system maintains task specifications (profiles) to specify the process-context of tasks and associated knowledge items [1][2]. The Kabiria system supports knowledge-based document retrieval in office environments, allowing users to conduct document retrieval according to the operational context of task-associated procedures [15]. Context-aware delivery of task-specific knowledge thus can be facilitated based on the task specifications and current execution context of. the. process.. Furthermore,. a. process. meta-model. specifying. the. knowledge-in-context is integrated with workflow systems to capture and retrieve knowledge within a process context [44]. Although providing an appropriate view for designing task-based knowledge support, the above works focus on specifying the process-context of tasks to support context-aware or process-aware knowledge retrieval, rather than on a systematic approach to construct task profiles. Moreover, the adaptation of profiles to track workers’ dynamic information needs is not addressed. For complex and knowledge-intensive tasks, the collaboration among knowledge workers may arise around common goals, problems and interests. Accordingly, contemporary KMSs rely on an effective approach to construct a community of practice to promote knowledge sharing. A community of practice consists of people who share common needs of information; hence, a community of practice is an effective approach to promote knowledge creation, transfer and sharing within or across organizations [3][13][18][41]. The Milk system supports informal communication and knowledge sharing for knowledge workers performing tasks in different work practices [3]. OntoShare, an ontology-based KMS, models the interests of users and provides automatic knowledge sharing in communities of practice with the aid of profiles [18]. Although user profiles had been employed to stimulate knowledge disseminations in communities of practice, they did not. 2.

(17) consider the identification of peer-groups with similar task-needs to form communities in the task-based business environment. Furthermore, for knowledge-intensive tasks, such as research projects in academic institutions, and product development in R&D departments, it is more difficult to supply task-relevant knowledge during the progress of task execution. That is, works’ information needs on task, for brevity, task-needs, generally change during the long run of task performance. Thus, the issues of identifying and tracking workers’ current task-stages and task-needs topics, and adjusting their profiles during task performance deserve further exploration. To provide a more effective long-term knowledge support, we propose a task-stage knowledge support model that incorporates Information Filtering model with the identification of worker’s task-stage and task-needs topics.. 1.2 Research objectives and tasks This dissertation mainly investigates the issues related to delivering and sharing codified knowledge from the perspective of business task. Major research objectives are listed below. (1) Proactively delivering task-relevant knowledge to workers engaged in knowledge-intensive tasks. •. A task-relevance assessment approach is proposed to identify workers’ information needs on task.. •. A task-based knowledge support model is proposed to track and model workers’ dynamic information needs on task. The proposed model also promotes knowledge sharing among knowledge workers.. (2) Enhancing task-based knowledge support model to provide effective knowledge support at different task-stages •. Developing a task-stage knowledge support model to provide task-relevant knowledge according to workers’ dynamic task-needs at different task stages.. •. Also, employing user modeling technique to identify worker’s task-stage and task-needs topics of stages.. (3) Deploying a task-based K-Support portal to acquire, organize, and disseminate the organization’s knowledge resources from the aspect of task. 3.

(18) •. Providing a collaborative task-based workplace to facilitate knowledge retrieval and sharing among peer-groups.. •. Delivering and sharing task-relevant knowledge to fulfill the workers’ task-needs at various task-stages.. 1.3 Contributions The contribution of this dissertation is to achieve knowledge reuse and support from the perspective of knowledge-intensive task. That is, extracting, organizing, and disseminating relevant knowledge (codified knowledge) to fulfill the information needs of knowledge workers during task execution. This work first proposes a novel task-relevance assessment approach to identify the knowledge worker’s information needs on tasks. Rather than specifying task characteristics directly by knowledge workers, a systematic approach is desirable to create task profiles by analyzing retrieved documents and assessing the relevance among tasks. Note that historical task-related information items preserved in the knowledge repository, such as task descriptions and codified knowledge, are valuable knowledge assets to support task profile construction. The proposed approach generates task profiles by the collaboration of knowledge workers to analyze the relevance of tasks and codified knowledge. Task-based knowledge support is facilitated through providing knowledge workers relevant knowledge based on task profiles. Although this work does not consider the process-aspect and context awareness, as discussed in previously pilot studies [1][2][24][44], this approach can alleviate the problem of accessing needed knowledge items from vast amounts of codified knowledge. Furthermore, methods of the adaptation of profiles to track workers’ dynamic information needs are proposed in this work. The worker’s dynamic task-needs can be analyzed based on the changes of workers’ profiles during task performance. An adaptive task-based profiling approach is proposed to tackle worker’s dynamic information needs on tasks. A task profile describes the key features of a task and is the kernel for discovering and disseminating task-relevant information to knowledge workers. This approach models the worker’s task-needs based on feedback analysis, i.e. explicit or implicit feedback on knowledge items. In addition, this work not only considers the profiles of feedback items but also considers the profiles of relevant. 4.

(19) topics in the domain ontology. Note that we refer the domain ontology as the taxonomy of topics in our task-based problem domain. Different from traditional information filtering techniques with user profile, which only considered the profile of feedback items, the profile adaptation approach considers both the profiles of related tasks and the profiles of relevant codified knowledge to adjust the task profile. For promoting knowledge sharing among workers, a task peer-group analytical method is proposed to identify task-based peer-groups according to workers’ profiles, namely, task interests. The main characteristic of this method is that a fuzzy inference procedure is employed to infer the implicit and transitive relationships of knowledge workers based on task-needs. The proposed method can infer the implicit relationship among workers; even they did not provide feedback on the same knowledge items. With the aid of task-based profiles and peer-groups, the proposed K-support portal can provide task-relevant knowledge and promote knowledge sharing among task-based peer-groups. Moreover, according to our empirical investigation, knowledge workers engaged in knowledge intensive tasks (e.g., research projects in academic organizations, project management in firms, etc.) have different information needs during the long-term task performance. The Vakkari studies (2000, 2003), which focus on a user’s information seeking activities during task performance (e.g., writing a proposal, completing a project, etc.), show that information needs vary according to different task stages. Therefore, we propose a knowledge support model based on task-stage to proactively deliver task-relevant knowledge. A correlation analysis method is proposed to identify a worker’s task-stage (e.g., pre-focus, focus formulation, and post-focus task stages), and an ontology-based topic discovery method is proposed to determine a worker’s task-needs topics of each stage. Consequently, the model can also be tailored to support long-term task performance. Finally, we develop a collaborative task-based K-support portal to facilitate knowledge reuse and to further promote knowledge sharing among peer-groups. The view of designing task-based knowledge support is the studies of context-aware or process-aware knowledge retrieval and knowledge delivery with the aid of user modeling. Details will be given in Section 3. Meanwhile, several experiments have been conducted to evaluate the effectiveness of the proposed knowledge support model 5.

(20) based on task or task-stage in terms of precision and recall. The empirical system evaluation is also conducted to examine the effectiveness of the proposed system in terms of novelty and quality metrics.. 1.4 Content organization Fig. 1 illustrates the whole view of this work and the remainder of this work is organized as follows. The literature review is given in Chapter 2. Chapter 3 addresses the rationale to design task-based knowledge support system and presents the framework of the proposed system. Note that the tasks and functions of each module given in Fig. 1 are described in this section. Chapter 4 introduces the process of building the task-oriented repository, as depicted in the block one (B1) of Fig.1. The repository is designed for organizing and managing task-relevant information. In addition, a task domain ontology is structured to organize and classify knowledge items based on tasks. The K-support model and methods to provide task-based knowledge support with the aid of profiling technique are given in Chapter 5, 6, and 7. Note that the associated experiments to evaluate the effectiveness of the proposed methods are also carried out. Chapter 5 presents the proposed task-relevance assessment approach to identify the worker’s information needs on tasks. The task-relevance assessment approach is designed to analyze the relevance of tasks and codified knowledge in the repository. Furthermore, a task profile is generated to support the proactive delivery of task-relevant knowledge. The assessment procedure is also given in the block two (B2) of Fig. 1. The lines with the numbers denote the assessment procedure. Next, Chapter 6 describes the proposed methods to disseminate and share task-relevant knowledge based on the generated profiles. The block three (B3) of Fig. 1 illustrates the main executed engines of Chapter 6 & 7. The user behavior tracker is an on-line module to capture workers’ dynamic behaviors, including access behaviors on the task-based domain ontology and documents. The task profile handler uses task-based profiling approach to adjust workers’ task profile to reflect workers’ current task-needs (information needs on the target task). The peer-group analyzer employs peer-group analytical method for identifying task-based peer-groups with similar task needs based on task profiles. Details will be addressed in Chapter 6. Chapter 7 extends the task-based knowledge support model to provide effective knowledge support at different task-stages The task-stage identifier and task-needs analyzer are within the block three (B3), which are responsible for tracking the evolution of a worker’s task-needs. Methods to. 6.

(21) identify worker’s task-stage and task-need topics of stages are presented in this chapter. Finally, the proposed K-Support portal with associated system evaluation is presented in Chapter 8. Conclusions and future works are discussed in Chapter 9.. Fig. 1. Task-based knowledge support. 7.

(22) Chapter 2 Related Work 2.1 Knowledge management in task-based working environment 2.1.1 Knowledge management systems and information technology Knowledge Management (KM) is a cycle, sometimes repeated process, which generally includes creation, management and sharing activities. [17][26][28][55][82]. Organizations deploy Knowledge Management Systems (KMS) to maximize the effectiveness of knowledge assets in increasing organizational profitability and productivity [30][55]. Contemporary KMS employs Information Technologies (IT), such as document management and workflow management to facilitate the access, reuse and sharing of knowledge assets within and across organizations [17][39]. Generally, information technologies (ITs) mainly focus on two dimensions, explicit and tacit dimensions, to support knowledge management activities [11][29][39]. The former is achieved by codified approach. Intellectual content codified into explicit form can facilitate knowledge retrieval and reuse [12][89]. Knowledge repository, knowledge-based system, knowledge maps are the like to support knowledge storage, organization and dissemination [29][39][89]. The latter put emphasize on dialoging via social networks to facilitate knowledge sharing. Knowledge expert directories, yellow pages, communities of practices and talk rooms are the like to support interpersonal communication to rapid knowledge sharing [3][13][39][41]. Several researches classified the knowledge management practices based on the two dimensions. According to Gray (2001a) empirical finding that the knowledge codified into knowledge repository make knowledge workers highly exposit existing resources within organization, whereas community of practices that provide informal personal communication can moderate to explore new possibility. Kankanhalli et al. (2003) pointed out those product-based firms in a high-volatility context are rely both codification and sharing approaches. Xerox, Microsoft, Hewlett-Packard are the examples. In summary, the critical role of ITs are to assist knowledge workers in fully and economic reusing valuable knowledge assets by decreasing the level of skills required in accomplishing the task successfully [28][39][49]. In addition, KMS with the aid of IT can assists workers in fully and economic reusing valuable 8.

(23) knowledge assets to accomplish the objective of task successfully.. 2.1.2 Task-based knowledge retrieval The repository of structured, explicit knowledge, especially document form, is a codified strategy to manage knowledge [17][29]. However, with the growing amount of information in organizational memories, KMSs face the challenge to help users find pertinent and needed information. The information can be delivered in a specific context of business environments. The information retrieval (IR) technique coupled with workflow management systems (WfMS) was employed to support proactively delivery of task-specific knowledge according to the context of tasks within a process. [1][24].. Furthermore,. a. process. meta-model. specifying. the. knowledge-in-context is integrated with workflow systems to capture and retrieve knowledge within a process context [44]. Despite the subtle difference among these works, they provide an appropriate view to achieve knowledge support based on tasks. Moreover, knowledge retrieval is also considered a core component in task-based business environment to access knowledge items in knowledge repository [24][27]. Herein, we categorized the task-based knowledge management work from two perspectives: one is knowledge delivery with the aid of user modeling and the other is context-based proactively delivery knowledge. The perspective is departure from the points of process complexity and knowledge intensiveness [21]. Based on the above points, for classes of business process are derived which are low (or high) business process and weak (or strong) knowledge intensity. In the following, the related works of task-based knowledge management will be given according to the classifying of business process. Task-based knowledge delivery with the aid of user modeling: This kind of knowledge management framework put emphasizes on codified (e.g., documents) knowledge retrieval and delivery in supporting workers' day-to-day tasks operation. Translating users’ information needs into compromised queries is not an easy work [75]. Most systems rely on Information Retrieval (IR) techniques to access organizational codified knowledge. The technique of Information Filtering (IF) with a profiling approach to model users’ information needs is an effective approach to proactive delivering relevant information to users. The technique has been widely. 9.

(24) used in the areas of Information Retrieval and Recommender Systems [31][52][58]. The profiling approach has also been addressed by some KMSs to enhance knowledge retrieval and further promote knowledge sharing among project-based or interesting groups [1][2][3][18]. Accordingly, the techniques of information filtering with intelligent agent-based architecture are commonly adopted in this type of framework to streamline the knowledge delivery from internal or external knowledge repositories [73][88]. Notably, a promising user modeling method, in which the system delivers the relevant information to the user profile is demanded in this type of knowledge support [7][71]. The idea of cooperative agent architecture has been proposed to achieve task-based Information filtering within work process [19]. Three types of cooperating agents: process agents, document warehouse agents and retrieval agents are designed for evaluating if the retrieved documents are relevant to the workers’ tasks at hand. Furthermore, a CodeBroker system is proposed for supporting software developer to reuse the organizational program components repository properly [88]. Similarly, the information filtering with user modeling and agent-based techniques are applied in the system for making delivered information relevant to the task-at-hand and personalized to the worker’s information needs. The task-based knowledge delivery with the aid of user modeling is quite suit applied in knowledge intensive task due to it has capability to model worker’s task needs and individual needs based on user modeling technique. The chief defect of this framework is that it generally cannot proper incorporate the contextual information of business task into the user profile. Context-based proactively knowledge delivery: The information can be delivered in a specific context of business environments. To this end, KMSs increasingly emphasize the organization of all the possible task-specific knowledge by supporting context-aware knowledge access and retrieval [1][5][44]. The Kabiria system supports knowledge-based document retrieval in office environments by allowing users to conduct document retrieval according to the operational context of task-associated procedures [15]. Furthermore, a process meta-model specifying the knowledge-in-context is integrated with workflow systems to capture and retrieve knowledge within a process context [44]. That is, context becomes an impartment component that can be utilized for improving the understanding of relevant knowledge of business task within the KMS. Recently, the knowledge context model. 10.

(25) is even proposed to support the collaborative work of virtual teams by utilizing the contextual information [4]. Furthermore, acquiring and disseminating role-relevant process views was considered in workflow environments [72]. Alvarado et al. (2004) also proposed acquiring and organizing corporate memory from the perspective of role/job position, in which an Organizational Memory is modeled by adopting UML/XML to specify the ontologies for organization positions, tasks, and application domains. The context-based knowledge delivery model is quite suit applied in knowledge intensive and non- routine task due to it has knowledge context model to capture or utilize the business process context for supporting task execution. Furthermore, it can even support the operation of business process with high process complexity. However, the kind of knowledge support model still lacks in learning capability to support real time context sensitive knowledge delivery till know. That is, besides understanding the work context of the given task, the model also needs to learn and response the worker’s task-needs in the real time.. 2.1.3 Knowledge sharing in community of practices For complex and knowledge-intensive tasks, the collaboration among knowledge workers may arise around common goals, problems and interests. Domain experts or experienced workers who hold valuable tacit knowledge play important roles in assisting knowledge workers to accomplish business tasks [51]. The ultimate goal of KM is to enable innovative activities by promoting collaboration or communication among knowledge workers in organizations [26][84]. Collaboration may take place in a formal group such as a business project or in an informal group such as a community of practice. A community of practice consists of people who share common needs of information; hence, a community of practice is an effective approach to promote knowledge creation, transfer and sharing within or across organizations [3][13][18][41]. Although user profiles had been employed to stimulate knowledge disseminations in communities of practice, they did not consider the identification of peer-groups with similar task-needs to form communities in task-based business environments.. 11.

(26) 2.2 Text mining technique for codified knowledge management 2.2.1 Information retrieval in vector space model The key contents of a codified knowledge item (document) can be represented as a feature vector of weighted terms in n-dimensional space, using a term weighting approach that considers term frequency, inverse document frequency and normalization factors [67]. The term transformation steps, including case folding, stemming, and stop word removing, are conducted during text pre-processing [7][60][65][83] The term weighting then is employed to extract the most discriminating terms [67]. Let d be a codified knowledge item (document), and let G d = <w(k1, d), w(k2, d), …, w(kn, d)> be the feature vector of d where w(ki, d) is the weight of a term ki that occurs in d. Notably, the weight of a term represents its degree of importance to represent the document (codified knowledge). The well-known tf-idf approach is often used for term (keyword) weighting. The approach assumes that terms with higher occurrence frequency in a document and occurring in fewer other documents are better discriminators to represent the document. Let the term frequency tf (ki , d ) be the occurrence frequency of term ki in d, and let the document frequency df (ki ) represent the number of documents that contain term ki. The importance of term ki to a document d is proportional to the term frequency and inversely proportional to the document frequency, which is expressed as Eq. 2.1.. w(ki , d ) =. 1. ∑ (tf (k , d ) × log( N i. df (ki )) ). 2. tf (ki , d ) × log. N df (ki ). (2.1). i. where N is the total the number of documents. Notably, the denominator in the right side of Eq. 1 is a normalization factor to normalize the weight of term. Similarity measure: The cosine formula is a widely used similarity measure to. assess the degree of similarity between two items x and y by computing the cosine of G. G. the angle between their corresponding feature vectors x and y , which is given by Eq. 2.2.. G G G G x• y sim( x, y ) = cosine( x , y ) = G G x y The degree of similarity is higher if the cosine similarity is close to 1.0.. 12. (2.2).

(27) Each document or query can be represented as a document or query feature G vector in a vector space model. Let d j represent a document vector of a document G dj and let q be a query vector of a query q. The similarity between a document dj and a query q, sim(dj, q), can be calculated as the cosine of the angle between the two G G G G vectors d j and q , namely cosine( d j , q ).. 2.2.2 Relevance feedback techniques Relevance feedback effectively improves search effectiveness through query reformulation. Various studies have demonstrated that relevance feedback applied in the vector model is an effective technique for information retrieval [63][68]. Eq. 2.3 and 2.4 illustrate two classical relevance feedback methods designed by Rocchio G (1971) and Ide (1971), respectively. A modified query vector q m is derived using the G relevance of documents (as feedback) to adjust the query vector q [7].. G. G. Standard_Rocchio: q m = α q + β. G. G. Ide_Dec_Hi: q m = α q + β. 1 Dr. ∑. ∑. G. d j −γ. ∀d j ∈Dr. 1 Dn. G. ∑. G. dj. (2.3). ∀d j ∈Dn. G. d j − γ max irrelevant ( d j ). (2.4). ∀d j ∈Dr. Where Dr denotes the set of relevant documents and Dn represents the set of irrelevant documents according to user judgment. Dr. and Dn. represent the. number of documents in the sets Dr and Dn. Meanwhile, α , β , γ are tuning constants. The function of maxirrelevant returns the most irrelevant document. The two methods produce similar results [7]. Most studies suggest that the information of relevant documents is more important than that of irrelevant documents [32][68].. 2.3 User modeling for information filtering 2.3.1 Researches of information filtering to support information needs Information retrieval and information filtering technologies applied in document management systems are generally the first pace of knowledge management initiatives, since textual data such as articles, reports, manual, know-how documents and so on are treated as valuable and explicit knowledge within organizations [55]. Information retrieval and information filtering are considered as the core techniques to achieve knowledge retrieval. In addition, information retrieval provides not only. 13.

(28) text processing technique, but also document classification technology to help organizations collect and process documents to achieve the goal of knowledge reuse [70]. With the aid of information filtering, it not only reduces the problem of information overloading but also provides relevant and needed information to users to accomplish their tasks. Information filtering (IF) systems are commonly personalized to support long-term information needs of a particular user or a group of users with long-term information needs [53][54][80]. IF systems are similar to conventional information retrieval (IR) systems. The IR system mainly focuses on facilitating user’s short-term information needs, e.g. generally expressed information needs in a single search session. However, the IF system relies on the support of the kernel technology of IR, but it puts emphasis on methods to maintain and learn user profiles to support long-term information services [7][9][80]. IF stresses on maintaining a promising user profile, in which the system delivers the relevant information to the user profile [7][71]. Various methods for learning user interests or preferences from text documents or Web pages have been proposed [7][8][10][52][53][54][58]. The well-known methods in Information Retrieval or Information Theory are modified and then employed to model user’s dynamically changed interests, for example, Rocchio algorithm, information gain theory, Bayesian classifier. Notably, all these learning algorithms require relevance feedback collection process, either explicit feedback (where system collects user linguistic ratings) or implicit feedback (where system monitors user access behavior). The IF system learns the users’ current task-needs from the feedback on the supported information, and updates the model for future information filtering. Such kind of learing method can maintain the user profiles once the system received the feedback; therefore, the learning method is regarded as the incremental learning technique. The IF technique is realized in many real-world applications, for example: e-mail-filtering systems [53], personalized online newspaper [10], adaptive Web page recommendation service [8], and on-line academic research paper recommendation [52]. Accordingly, IF technology is acknowledged to be an effective way to reduce the information overload and provide personalized information [30][35]. Although IF systems provide proper profiling method to learn user’s dynamic needs/interests; however, most of existing systems do not consider. 14.

(29) integrating the user’s information needs with the progresses of task performance. A promising profile modeling approach considering the characteristics of task stage and user’s current information needs is more demanded in task-based business environments.. 2.3.2 User modeling technique to support knowledge-intensive tasks The characteristic of knowledge retrieval activity in working environment is that the worker’s information needs is associated with the executing task at hand. Meanwhile, a knowledge-intensive task consists of levels of progressively smaller subtasks to achieve the main task goal. That is, when the worker confronts with the task, there is a gap between the worker’s knowledge about the task and the perceived requirements of tasks. The gap is the information need and results in information seeking activities [14]. Generally, a worker uses documents to understand a task, solve the encountered problem, or result in another search behavior for finding a solution. Accordingly, several empirical studies focus on how documents are selected and used by workers during task performance. A well-known longitude project has been conducted to investigate a cognitive model of document use during a research project [78][79]. The study models document use as a decision-making process where decisions may occur at three points or stages during a research project, which are selecting, reading, and citing. Several empirical studies concentrated on discovering and analyzing the growth in students’ or scholars’ understanding of their own assigned tasks during conducting an actual research project [43][50][76][77][78][79]. The Kuhlthau’s study (1993) [43]is to observe people involved in information seeking over a period of time. Six stages were identified in his empirical study from the students’ description of their experience; these stages match the phases in the process of construction. The Vakkari (2000) studies concentrated on the user’s information seeking activities during the progress of task performance (e.g. writing a proposal, completing a project and the like). The Vakkari study is based on the Kuhlthau’s model to connect the research of information seeking activities to the pre-focus, focus-forming and post-focus stages of the process [76][77]. The empirical studies reveal that users’ information needs will vary at different task stage. For example, the types of information needs may vary from general information to specific information, and the choice of search terms is varied from broader terms to related terms. That is, a worker’s information needs 15.

(30) and information-seeking processes depend on worker’s progresses of task performance, or task stages, specifically. The characteristic of knowledge retrieval activity in working environment is that the worker’s information needs are associated with the executing task at hand. Meanwhile, a knowledge-intensive task consists of levels of progressively smaller subtasks to achieve the main task goal. Therefore, the concept of task stage in information seeking studied can support this work for providing task-relevant knowledge more precisely. And a promise knowledge support model to reflect workers’ current task-needs and task-stage is a critical issue deserved exploration.. 16.

(31) Chapter 3 Task-based Knowledge Support 3.1 Rationale to design task-based knowledge support The. proposed. work. focuses. on. providing. knowledge. support. for. knowledge-intensive tasks within organizations. Examples of knowledge-intensive tasks include thesis works and research projects in academic organizations, project management in firms, research work and product development in R&D departments, and the like. In such task-based environments, reusing knowledge assets extracted from historical task executions is the key to providing effective knowledge support for conducting tasks. Historical codified knowledge, i.e. experiences and know-how extracted from previous task executions, provides valuable knowledge for conducting tasks. For example, effective project management can benefit from KMS by referring similar projects to acquire best practice, lessons learned, working experiences, or knowledge resources. Research task innovation is generally based on previous research achievements. A knowledge repository that preserves the experience and knowledge of previous work (research task) is important to provide effective knowledge support for research tasks. However, with the increasing amount of information in the organizational memory (OM), contemporary KMS faces challenge to assist organizations acquire, organize and manage knowledge. Thus, delivering relevant historical codified knowledge to workers for accomplishing tasks at hand is also a challenging work deserves exploration. This work sought to tackle the challenges from the perspective of business task.. 3.1.1 Task-based organizational environment “Mary is a new worker of an industry analyzer in a project management institution. She is assigned to a survey task, “the opportunities of sensor network in healthcare”, and need to write a proposal. Since Mary is a novice of sensor network, she faced the problem to understand the assigned task. She wants to find task-related expert or colleague to solve the encountered problem or guide him to the right direction while understanding the perceived task. Unfortunately, workers who have relevant knowledge are busy for the business projects. Hence, Mary comes up with the idea to find the possible solutions from the document management system or information. 17.

(32) repository in the organization. However, tremendous amount of data frustrated Mary. That is, it is hard for Mary to have a clear view of information structure or taxonomy of the document management system or information repository.” The situation generally happens in the organization, especially in IT or MIS department of industry, or the industry analyst in project management institution. When a worker in an organization has information needs of the executing task, he/she might need the knowledge support to accomplish the task. Naturally, the worker may seek someone who has met this problem or has done similar experiences before. Otherwise, the worker may also try to find the relevant codified knowledge from the organizational repository. Thus, if knowledge resources in an organization are acquired, organized via the view of business tasks, workers could get more effective knowledge support.. 3.2 Framework of task-based knowledge support Figure 2 illustrates the system framework of the proposed task-based knowledge support based on profiles to facilitate task-based knowledge delivery and sharing. Participants include knowledge workers engaged in specific tasks and domain experts in specific subjects. The system comprises four main modules, namely task-oriented information repository, task profile handler, task-needs evolution, and task-oriented information service router. Task-oriented information repository. The task-oriented information repository is. designed for organizing and managing task relevant information. Building a proper repository to acquire and disseminate knowledge items is a key strategy for managing knowledge in the contemporary KMS. Information items indexed by proper concepts and categories can provide knowledge workers with meaningful access to organize intellectual content. Task-oriented repositories are constructed with support from category schema to effectively utilize codified knowledge. Such a repository stores codified knowledge corresponding to task execution, and contains three main databases, including the document-indexing database, task corpus, and task categorization database. The document-indexing database stores task relevant documents indexed using the inverted file approach. Meanwhile, the task corpus stores the key profile of each existing task. An existing-task is a historical task accomplished within the organization. Task corpus is used to describe the key. 18.

(33) Worker's Engaged in a Specific Task. User Behavior Tracker. Codified Knowledge and Human Resource Support. Text Processor & Information Extraction. Task-oriented Repository. Task-oriented Information Service Router. Task Profile Handler. Peer Group Analyzer. Task Needs Analyzer. Task Stage Identifier. Root. F2. F1 W (T 2 ). W (T 7 ). T7. T2. W( T 6 ). Domain Ontology Configuration. T6. Dynamic Task/Work Profile. kw1 ,…, kwn. Fig. 2. Framework of task-based knowledge support. subjects of an existing task, and is expressed as a feature vector of weighted terms. Section 4.1 details the extraction of task corpus of an existing task, which is derived by extracting the weighted terms from textual documents generated and accessed by the task. Moreover, the task categorization database records the relationships of existing tasks and categories, namely, the relevance degrees of existing tasks to categories. The task categorization database is used to support the operation of identifying referring tasks based on their similarity to the executing task derived using the relevance degrees of tasks to the categories. Moreover, tasks with similar subjects are grouped into fields. The repository is the knowledge base for task-based knowledge support. Details are discussed in Chapter 4. Task profile handler. The task profile handler provides mechanisms such as profile. creation, adjustment, integration and profile adaptation to conduct profile management. Two kinds of profiles, feature-based task profile and topic-based task profile, are maintained to model workers’ information needs on the target task at hand. y. Feature-based task profile describes the key features of a task and is the kernel. for discovering and disseminating task-relevant information to knowledge workers. y. Topic-based task profile models a worker’s information needs on the target. task, and is represented as a set of relevant tasks or fields of the target task with associated relevance degrees. Workers’ information needs may change during the progress on performing the target task. The user behavior tracker is an on-line module to capture workers’. 19.

(34) dynamic behaviors, including access behaviors on the task-based domain ontology and relevance. The profile handler uses an adaptive task-based profiling approach to adjust workers’ profiles. The peer-group analyzer employs a task-based peer-group analytical method to identify peer-groups with similar task needs (information needs on the target task) based on work profiles. Details will be addressed in Chapter 6. Task-needs evolution. The task-stage identifier and task-needs analyzer are within. this module, which are responsible for tracking the evolution of a worker’s task-needs. The profiles are employed as indicators for task-stage identifier and task-needs analyzer to model the worker’s task-needs of target task. Herein, worker’s task-needs are modeled as the topics nodes in domain ontology (DO) at different abstraction level which are relevant to the on-going task. The DO is a multi-level structure and each node in the DO represents a research topic in our application domain, as given in the Figure 3 of Chapter 4. The task-stage identifier is responsible for analyzing and determining worker’s task stage based on the changes of the task profile over time. The task-needs analyzer is responsible for tracking the worker’s access behavior over a period of time. The access behavior is analyzed based on the DO to discover worker’s task-needs on specific topics. Details are discussed in Chapter 7. Task-oriented information service router. The router helps knowledge workers. gather appropriate information from the task-oriented repository and task-based peer-groups. The router fetches task-relevant information according to the worker’s task profile. Moreover, each worker has his/her own view of task-relevant information, namely, personalized ontology, which is derived from his/her work profile on the target task and is organized according to the domain ontology. Knowledge sharing from other peer-group members is derived by retrieving each peer-group member’s personalized ontology. Details are addressed in Chapter 6.. 20.

(35) Chapter 4 Task-Oriented Information Repository: Managing Codified Knowledge 4.1 Task-oriented information repository To organize and manage task-relevant information, the repository is constructed with support from domain ontology (i.e., topic taxonomy) to effectively utilize codified knowledge. This session discusses the issue of managing codified knowledge with the support from category scheme. Categories representing the main subjects of organizations are defined to organize tasks and codified knowledge. Task corpus (feature vector of weighted terms) describing the key subjects of existing task can be constructed by extracting the weighted terms from textual documents. The task categorization database records the relevance degrees between existing tasks and categories based on the result the proposed task categorization model. The task categorization database is used to support the operation of identifying referring tasks based on their similarity to the executing task derived using the relevance degrees of tasks to the categories. Identifying a small subset of existing tasks as referring tasks can help knowledge workers conduct further task-relevance assessment without reviewing all existing tasks. This chapter illustrates two essential phases in constructing a task-oriented information repository: extracting task corpus from textual data gathered during task execution and deriving the relevance degrees between existing tasks and categories.. 4.1.1 Extracting task corpus The task corpus of a task tr is represented as a feature vector of weighted terms (keywords) derived by analyzing the set of documents generated and accessed by tr. G Each document dj is pre-processed and represented as a feature vector d j . The centroid approach is employed to derive the feature vector of a task by averaging the feature vectors of documents generated and accessed by the task. Let Dtr denote the set of documents that are generated and accessed by task tr. Furthermore, the task JG. corpus (feature vector) of task tr is defined as the centroid vector t r which is the vector obtained by averaging the feature vectors of documents in Dtr. Eq. 4.1 defines JG. JG. the centroid vector t r . The weight of a term ki in t r is represented as w(ki, tr).. 21.

(36) G 1 tr = Dt r. G d ∑ j. d j ∈ Dt r. (4.1). 4.1.2 Task categorization model Existing tasks are categorized based on fuzzy classification, and thus they may belong to more than one category. Fuzzy classification extends the traditional crisp classification notation to associate each object in every category with a membership function so that each object can belong to more than one category (Zadeh, 1965). The task categorization database records the relationships of existing tasks and categories, namely, the relevance degrees of each existing task to categories. The relevance degree between a task and a category indicates the strength that the task belongs to the category. The relevance degrees between categories and existing tasks are calculated based on the similarity measures between feature vectors of categories and existing tasks. The feature vector of a category is also expressed as a vector of weighted terms, which represents the main subjects of a category. The categorization procedure includes the step of deriving the feature vectors of categories and the step of deriving the relevance degrees between existing tasks and categories.. Deriving the feature vector of each category: Experts predefined a set of categories to represent the main subjects within the organizational domain, such as “Text Mining”, “Knowledge Management”, etc. The seed-based approach is then applied to generate the feature vectors of categories. Experts select some existing tasks which represent a category. The selected tasks are called the seed tasks of the category. Once the seed tasks have been decided, a centroid vector can be derived from the corpora (feature vectors) of the seed tasks to describe the category. The centroid vector of each category is derived by averaging the feature vectors of corresponding seed tasks. Let X denote a set of categories, X={c1, c2," , cm}, and let Tcj represent the set K of seed tasks of category cj. Let c jc be the centroid vector derived from the task K corpora (feature vectors) of seed tasks of cj. The centroid weight of term ki in c jc , K w(ki, c jc ) is derived as Eq. 4.2.. 22.