國立台灣科技大學 資訊工程系所
模糊類神經網路實驗室 研究論文
Fuzzy Neuron Laboratory Paper

93級畢業碩士 呂素芳 發表論文


工作流程搜尋 :以工作流程探勘為基礎的網頁排序

摘要

由於企業入口網站提供了大量的資訊,因此員工需要搜尋引擎來幫助他們找尋需要的資訊
。雖然目前一般的搜尋引擎已具備幫助員工找尋相關資料的能力,但由於企業資料增加得
太快或是分類的原因,因此遇上了一些問題。這些問題包含搜尋結果的精確度過低以及搜
尋系統缺乏重要的領域知識(Domain Knowledge)。處理這些問題的其中一個方法就是建構
特定領域搜尋引擎(Topic-Specific Search Engines)。藉由將專家所建立的產品同義詞
庫及對企業工作流程的分群技術於搜尋系統,除了能讓員工在瀏覽搜尋結果時能更瞭解企
業的相關產品及其工作此領域的知識外,搜尋結果也較符合員工的需求。
在此篇論文中,我們將焦點放在產品與工作流程的關係上。我們提出一個名為工作流程搜
尋的系統來幫助員工搜尋以產品及工作流程為主的網頁。其基本精神在於增加了工作流程
探勘(Workflow Mining)及產品同義詞庫(Product Thesaurus)建立的技術於搜尋系統。
藉由對產品工作流程探勘的分群技術,讓相同產品聚類在一起以及由專家所建立的同義詞 (
Product Thesaurus),用來定義企業的產品及其產品同義詞。我們並且加入工作流程成本
的概念來做為相關產品排序的依據。此系統主要的貢獻在於能幫助員工更有組織的找尋他
們所欲尋找的產品及工作流程,且同時提高搜尋結果的準確度。


Workflow Search: A Ranking Based on Workflow Mining

Abstract

Since vast and heterogeneours documents are availbale on the Internet and Intranet, staff require search engines to locate their required information. Keyword-based search engines return some valuable search results with respect to staff queries. However, with rapid growth of the web pages and classification of the web pages, keyword-based search engines face several problems. Here of these problems are lacking of domain knowledge, lacking of workflw and low precision of search results. An effective means of resolving these problems is to construct topic-specific search engines. With the domain knowledge provided by experts, the precision of search results can be increased.

This study, examines how "product" and "workflow" are related. A new search approach, i.e., Workflow Search, based on Workflow mining is proposed, and the classification in a way that helps staff search for related web pages. Our ranking method, calculates he costs of workflow that can provide a good workflow for staff. Our experimental results, indicate that our system is operable by workflow mining to determine how products and workflow are related. Second, the precision of search results in our system is higher than that in the TCB and the FB portal.