查詢結果分析
相關文獻
- A Fuzzy Document Retrieval System based on Concept Networks and Cluster Analysis
- Research and Application of Fuzzy Sets Theory on Manufacturing Process
- 銀行授信決策應用類神經網路之研究--抵押貸款之實證研究
- 長期性資產購買或租賃模糊投資決策
- 模糊不迷糊
- 臺灣地區大豆、高粱供給對風險反應及其種植面積變動之預測--模糊集合理論之應用
- 地籍圖之線條重建
- 規則庫控制法在混合式隔振系統之應用
- Classification of Water Masses and Sound-Scattering Layer Biomass in the Waters off Northeastern Taiwan Using a Fuzzy Clustering Method
- 應用模糊集合理論於中壢電離層探測儀自動化電離圖判讀
頁籤選單縮合
題名 | A Fuzzy Document Retrieval System based on Concept Networks and Cluster Analysis=一個以概念網路及聚類分析為基礎的模糊文件擷取系統 |
---|---|
作者姓名(中文) | 林娟娟; 曾修宜; 陳培敏; |
作者姓名(外文) | Lin, Chuan-chuan; Tseng, Shou-yi; Chen, Pei-min; |
書刊名 | 東吳經濟商學學報 |
卷期 | 25 1999.06[民88.06] |
頁次 | 頁39-60 |
分類號 | 494.59 |
語文 | eng |
關鍵詞 | 文件擷取; 模糊集合; 聚類分析; 概念網路; Document retrieval; Fuzzy set; Cluster analysis; Concept network; |
中文摘要 | 本文提出一個模糊文件擷取系統,此系統利用概念網路來描述文件間的關連度,並依此值之強弱將文件分成不同的聚類。而這關連度是以模糊數值來描述,並把文件間的關連度以概念矩陣表示。再運用模糊集合理論,將概念矩陣作運算,找出其等價關係矩陣,以充分表達出兩兩概念或兩兩文件之間的關係。 我們的系統分為兩個子系統:一為文件分類、一為文件擷取,這兩個子系統分別在兩個階段時執行。第一階段是離線作業,文件分類子系統以概念矩陣作為輸入資料,進行文件聚類分析,將系統內全部文件分為若干群組,第二階段則為線上查詢作業,文件擷取子系統根據使用者線上鍵入之資料(與查詢文件相關之模糊向量),先搜尋出相關的文件群組,再進一步找出最符合查詢的文件。 相較於沒有作聚類處理之文件擷取,本系統實驗的結果顯示,在精確度降低1%以內,查詢所需的時間可以節省25-30%,如此可以大幅縮減文件擷取的時間。 |
英文摘要 | A fuzzy document retrieval system based on concept networks is proposed, where documents are categorized into clusters depending on the degree of relevance among them. In this paper, a concept network, used as knowledge, depicts the relevant relationships, which is expressed by fuzzy values, among concepts and is represented by a concept matrix. The implicit relevant relationships between concepts are derived by the matrix operations of fuzzy logic and transitive closure on the concept matrix. The system model is divided into two phases: document clustering and document retrieving. The first phase is an off-line job which figures out the document clusters based on the concept matrix. In the second phase, documents are retrieved on-line from related clusters according to user's queries. The proposed model is more efficient than the previous research due to the fact that, instead of all documents, the retrieving operation is performed only on the limited amount of documents sieved by clustering documents. Our experiment shows that, comparing to the nonclustered document retrieval, there is a 25-30% of time saving with the cost of decreasing the precision rate less than 1% in the document clustering approach. |
本系統之摘要資訊系依該期刊論文摘要之資訊為主。