以「問題–解法」為基礎的軟體知識分類法

Abstract

軟體發展是腦力密集及知識密集的工作，在開發過程中，軟體工程師常需搜尋所需技術知識來加速其發展。由於目前技術知識庫的關鍵字缺乏良好制定法則，無法有效代表技術知識的意義，進而影響知識搜尋的範圍、精度與速度，對軟體生產力造成不良影嚮。

現有知識管理平台常見的知識分類法則大致可以分為四類，即Taxonomy、Faceted Classification、Case-Based Reasoning與Ontology。這些方法中，Taxonomy太過簡單故缺乏效率；Faceted classification會因知識數量增多而過於複雜，且難以對軟體知識定義有效率的面向；CBR則缺乏良好的分類架構；Ontology只做到知識的描述，而欠缺知識間的關聯性。故這些方法應用於軟體知識管理仍有許多不足之處。

本論文發現多數技術知識主要內容是針對特定議題提出解決方法，因此提出以「領域類別」、「議題」、「解法」、「技術」來代表一技術知識的性質，且發現此種表示法易於建構技術知識間的關聯性。進而本論文提出以「問題–解法」為基礎的軟體知識分類架構，以領域類別為基礎，在類別下有子議題，各議題下有許多論文提出不同的解法及各種技術解決此議題。本論文並依提出的知識分類方法，實作了一雛型（Prototype）知識庫，以驗證其可行性。

本文所提之新技術知識關鍵字可清楚表示論文之性質，且可依領域類別、議題、解法或技術有效且精確搜尋出所要的技術知識，較以前方法更為實用。

It is well convinced that software industry is highly knowledge- and labor- intensive. During software development process, software developers usually search for technical knowledge to shorten the development time and improve the software quality. However, technical knowledge doesn’t have good indexing keyword definition to stand for its main point. It will cause poor accuracy and efficiency of knowledge query. Consequently, it’s the obstacle toward productivity of software development.

Most popular knowledge classification techniques are Taxonomy, Faceted classification, Case-based reasoning, and Ontology. Among these techniques, Taxonomy is too simple to be effective. Faceted classification is too complicated when the amount of knowledge increases, it’s also difficult to define good facet for software knowledge. Case-based reasoning lacks classification structure for knowledge domain. Ontology uses its model to describe knowledge, but there is no relationship between knowledge. Since above techniques is not designed for software knowledge, applying to software knowledge management has drawbacks.

We find that the purpose of technical knowledge is to propose an approach to an issue. Hence, technical knowledge could be represented by its category, issue, approach and technique. With this representation, it’s very easy to construct the relationship between knowledge. Accordingly, this thesis proposed a problem- approach-based software knowledge classification system. This system classify knowledge domain with hierarchy structure. Leaf-node categories have issues. Technical knowledge proposes approach to each issue, and uses certain techniques. Then we can represent technical knowledge effectively, and increase the accuracy and efficiency of software knowledge query. To prove the feasibility, we also implement a prototype system.