以事件屬性及事件關聯交叉導向建構知識管理系統之研究

Abstract

『知識是能夠改變某些人或某些事務的資訊』– Peter F. Drucker 曾經因為資訊的取得不易，資訊不足是個令人困擾的重大問題，伴隨著資訊科技軟硬體的快速發展，資訊不足的問題逐漸變成了資訊過量，再加上近來知識經濟興起並逐漸蔓延，一九九○年代中期之後，知識管理這門學問，正以驚人的速度濫觴。 談到知識管理，最重要的便是知識收集和取用兩個構面。知識收集包含了知識物件的定義及知識物件的儲放，就傳統的應用而言，前者多是利用知識審查的方式進行所謂的知識認定；而後者，則多以關鍵字的方式連行分類儲放；在這個處理過程中，『人』扮演了一個決定性的角色，也因此，審查客觀性是一個相當不易克服的難題；什麼是知識？誰能決定何謂知識？在不同的程度、水準下，對知識的認知是否也相同？有了不同的認知程度，日後知識取用時賴以進行知識物件定位的關鍵字系統，又如何才能增進其精確性及可用性？ 至於知識取用的部份，目前較常見的作法，是針對知識物件的關鍵字或是一些基本屬性（如作者、發表日期、所屬單位…）進行搜尋定位的動作，抑或是利用全文檢索的功能進行大範圍的查詢，此種模式下，最大的問題便是知識物件的定位無法精確，往往查詢結果數量過多而造成選用上的費時與不便。此外，當欲進行關聯性查詢時（如知識物件形成的原因及其後續發展），單單以此種知識物件取用方式似乎不易達成。 綜合上述問題，本研究提出一個知識管理架構，透過事件屬性（知識物件的分類）與事件關聯（知識物件的因果關係）進行交叉定位，進而建構出一具靜態及動態情境的雙維度知識地圖，供使用者交叉查詢使用。在事件屬性上，利用一個類似專家系統的邏輯庫模式，藉由屬性間互斥或是獨立的關係建立所謂的基本邏輯（Basic logic），加上由使用者依需求型態而設定之外部邏輯（External logic），建構一習慣導向之知識樹（Knowledge tree），以簡化知識輸入時的分類動作；在事件關聯上，則是利用繼承的觀念，建構出具因果關係的事件流，除了提供『How』的查詢維度外（由屬性找出知識物件），多提供了一個『Why』的查詢維度（除了知識物件（事件）本身，尚可得知其來由及後續發展（事件流，Event tracking））。 最後，本研究將此模式試行於一家大型半導體公司P公司，提供研發類之專業知識管理使用。

It was only years ago that we still suffered from the insufficiency of information, as a result of poor-quality information acquisition method. Through the rapid development of software and hardware over the years, the previously known problem information insufficiency has been replaced by information abuse. The rise of economy scale also adds to the severity of the problem. In response to such transition, by the middle of 1990, the study of Knowledge Management has become so popular and spread quickly in many areas of research. Among the study of knowledge management, two areas have attracted the most attention, i.e. knowledge collection and access. Knowledge collection primarily deals with the definition and storage of the identified info object. By contrast, the former (or, definition) focus on knowledge recognition through knowledge auditing. The latter (or, storage), however, is often accomplished through so-called keywords. In the processing of knowledge collection, the human factor tends to play an important role. Maintaining of the objectivity over humanity hence becomes a major yet formidable issue: What exactly is knowledge? Who can decide what information qualifies as knowledge? Given different context and circumstance, how can we still maintain the same decision criterion over various information pieces? And how do we assure the accuracy and accessibility of keyword system for future application? On the other hand, knowledge access focuses on locating the target knowledge object, through searching either keywords or pre-defined basic attributes (e.g. author, publish date, department, etc.). Full-text indexing can be useful too, when the search scope becomes rather large. The primary challenge about knowledge access is the level of precision over the search result. Often the magnitude of the search result becomes so large that filtering through the result can be time consuming and rather inconvenient. Subsequently, such search method may seem inefficient when applying relational queries (for example, locating the background and event tracing of the identified info object) To overcome the issue mentioned above, this paper proposes an Knowledge Management Framework, which is realized through a two-dimensional knowledge map, capable of satisfying users with various types of cross-queries. Such framework is made possible through the method of cross-identification over event attribute (knowledge object category) and event relation (knowledge object history). The use of event attributes is enabled through an expert system-like logic base. In brief, a habit-oriented knowledge tree that comprises of Basic logic (describing the exclusiveness or independency of attributes) and External logic (defined by users based on types of requirement) was built to simplify the classification when entering attributes. In applying event relation, the concept of inheritance is utilized to construct the event tracking. As a result, in addition to the『How』 dimension (i.e. locating the knowledge object through attributes), an additional 『Why』 dimension can also be provided to users. Consequently, other than the knowledge object (or, event) itself, the background and event tracking of the object can also be acquired. The proposed method has been successfully applied to the R&D knowledge management of a large-scale semi-conductor company P.