一個以Ontology為基礎的Web-Mining技術應用於供應鏈競爭分析之研究

A Case Study of the E-Logistics of Global 3C Manufacturing Enterprise

本研究的目的是開發一個以ontology為主的產業供應鏈知識探勘平台，並提出一種利用文件探勘 (text mining) 與XML文件技術整合的方法來提供企業在供應鏈競爭分析上的應用;為測試所開發的文件探勘演算法，本研究在文件探勘的對象是與產業活動相關的新聞語料庫為主，並以RosettaNet作為系統的重要內容來源，以進行系統模型架構的實現及個案研究。根據這個探勘演算法，應用網路探 勘 (Web Mining) 的技術從大量半結構性 (semi-structured ) 以及非結構性 (unstructured) 的網頁文件中萃取出文件內容中相關的概念與知識，以產生用來描述文件的資訊 (i.e. metadata) 與階層式的知識分類目錄架構，再將原有網頁轉化為XML文件存於這個分類演算法的Ontology架構為基礎的XML文件庫中。本研究將文件探勘的技術應用於分類目錄的自動建立與維護，並開發一個能達到自動知識分類目的之XML文件資料庫系統。本研究主要採用的研究是以Web content mining的方法為主，亦即以文件探勘 (Text mining) 技術針對存在於WWW網頁中的文件資訊內容加以分析處理，並運用類神經網 路機器學習的技術來實現。

In this research we propose a novel approach to develop a platform for discovering supply-chain competitive analysis on an ontology-based web-mining technique. Also, by integrating a text mining approach with a XML document technique, in the developed platform we provide a way to allow businesses tackle difficulties in knowledge management for the supply-chain related information. To testify the developed web-mining algorithm, in this research a corpus associated with industrial information collected from specific news web sites (e.g. CNA News), with the RosettaNet standard framework, is employed as the major information source for conducting system implementation and case study. By applying the developed web-mining algorithm, in this work we attempt to extract concepts and knowledge from a huge semi-structured and unstructured HTML-document collections. The extracted concepts and knowledge can then be used to produce metadata and ontology to describe the contents in the original web documents. As such, the original web documents can be transformed into XML documents and stored in the XML document database based upon the ontology based "knowledge template". The research applies a text-mining approach to automating the construction and maintenance of a concept-hierarchy, in order to establish a XML document database based on the extracted metadata and ontology. The approach for knowledge extraction in this research is mainly using a Web-content mining method. That is, the existing WWW pages can be analyzed to generate a set of metadata to describe their content and produce an ontology for the XML document database through a text-mining technique, incorporated with a neural-net machine learning method for implementation.