『問題-解法』軟體知識庫之設計與實作

Abstract

軟體發展過程中，軟體工程師遇到問題時常在技術期刊或技術報告尋找解答，但一般期刊資料庫、技術報告庫或公司技術資料庫的技術文件大都依年代、作者或文章名稱等作分類，軟體工程師需花甚多時間才能找到所需文件；針對此缺點，謝祖望學長[6]提出「問題-解法」知識分類法，並以一雛形系統證實可快速找出相同問題及解法之技術資料，唯其雛型系統過於簡單，應用於私人知識庫、技術報告庫及期刊知識庫時，仍有缺乏技術關鍵字逆向查詢功能、缺乏巨量資料處理功能、缺乏文件編碼機制等等缺點，本研究運用「問題-解法」知識分類法進行私人資料庫、技術報告庫及期刊資料庫之設計。

本研究首先依據「問題-解法」知識分類機制及技術文件數量提出一個十欄之64-bit知識編碼機制，可同時用於書本及技術論文之編碼，進而提出關鍵字逆向索引機制以快速找出相同解法或技術之論文或技術報告，最後提出一個適用於巨量文件資料之分散式儲存機制，運用這些機制分別提出技術報告庫、私人知識庫、期刊知識庫及全文資料庫之系統架構，並實作一技術報告庫雛形系統，藉以驗證上述機制之實用性，證實「問題-解法」知識分類法可供軟體知識師依問題、技術或解法快速找出所需技術文件，且可與現有知識搜尋機制並存。

During software development, software developers often search solutions from books, technical reports, and papers when they encounter problem. Technical Documents in journal databases, technical report database, and private knowledge database are classified by publish date, author name, or title. It often spends much time to find the right documents. Because of this drawback, Tzu-Wang Hsieh[6] proposed the Problem-Approach-based Software Knowledge Classification. He also implemented a prototype system which shows that documents solving the same issue or using the same approach can be found very quickly.

Due to the prototype system is too simple, when it is applied to private knowledge database, technical report database, and journal database, it has some drawbacks. First, the system lacks supporting for inverted search using keyword. Second, it lacks the ability of processing huge amount of documentations. Finally, it lacks encoding mechanism for documentations and keywords. This research uses the Problem-Approach-based software knowledge classification to design system architectures for private knowledge database, technical report database, and journal database.

A 64-bits code with ten fields based on the Problem-Approach-based software knowledge Classification technique is proposed to represent the ID of a technical document which can be a technical paper, technical report, or book chapter. Base on the encoding mechanism, we propose an inverted index mechanism to store technical documents with the same approach or technique. Finally, we propose a distributed knowledge management mechanism for huge amount of documents. With these mechanisms, System architectures for technical report database, private knowledge database, journal database, and full-text database are proposed. A prototype system for technical report database is also implemented. With the prototype system, software developer can quickly and precisely find needed technical documents by using issue, approach, or technique keywords. The conventional search mechanism, such as by author name, year, keyword, etc, is also implemented in the prototype system.