| 摘要: | 摘 要
地理網路(geospatial Web, GeoWeb)表示包含地理空間要素的網路資源，如地圖、地理編碼照片、供應地理資料的網路服務等。如同一般網路資源，地理網路資源散布整個網際網路，使得辨識及整合地理網路資源成為一個艱巨挑戰的任務。為了促進地理資源的發現及重複利用，我們需要地理網路爬蟲主動探索資源並建立一個地理網路搜尋引擎以供查詢及排序結果。然而，為了進一步增進資源的搜尋效能，我們認為理解地理網路資源的語意和網路使用者興趣以提供個人化的搜尋結果極其重要。因此，本研究首先設計地理網路資源的本體論(Ontology)，結合多個領域本體的概念來表示地理網路資源。本體論有助於對應與探索概念間的語意關係。地理網路資源本體論亦可幫助語意化(Semanticization)使用者搜尋歷史以建立使用者模型以代表使用者的偏好。最後，當搜尋引擎接收到使用者的查詢，該查詢則基於地理網路本體論和使用者模型進行個人化擴充，進而找到語意相似且符合使用者偏好的搜索結果。再透過比對地理網路資源之語意設計適當的排序機制，幫助使用者有效地找到所欲搜尋的和相關的地理網路資源。實驗分為三組使用者，第一組為五位訓練查詢數量不同的模擬使用者、第二組為五位訓練查詢數量相同的模擬使用者、第三組為十位不受限制的真實使用者，其正規化衡量搜尋引擎質量指標(Normalized Discounted Cumulative Gain, NDCG)值分別為0.97、0.97、及0.85。整體而言，實驗結果顯示所提出的地理網路個人化搜尋方法能夠有效地根據使用者偏好找尋並排序相關的地理網路資源。
;ABSTRACT
Geospatial Web (GeoWeb) represents the collection of Web resources that contain geospatial components, such as maps, geocoded images, Web services hosting geospatial data, etc. Similar to general Web resources, GeoWeb resources are scattered on Internet, where the identification and integration of GeoWeb resources become a challenging task. To facilitate geospatial resource discovery, we need GeoWeb Crawlers to proactively discover GeoWeb resources, and establish a GeoWeb search engine to provide querying and ranking mechanisms. However, to further improve resource discovery efficiency, we believe that understanding semantic meanings of GeoWeb resources as well as user interest is important to provide personalized search results. Therefore, in this study, we first design a GeoWeb resource ontology, which contains necessary concepts from many domain ontologies to represent a GeoWeb resource. This ontology helps map and discover semantic relationships between concepts. The GeoWeb resource ontology also helps semanticize a user’s search history to construct a user model representing important concepts to the user. Finally, when receiving a user query, the query is extended based on the GeoWeb ontology and the user model. In this case, semantic-similar and personalized search results can be found. By comparing with the semantic of GeoWeb resources, a proper ranking algorithm is designed to help users find targeted and related GeoWeb resources efficiently. In the result, two experiments of artificial users and one experiment of real users show that the proposed solution achieves 0.97, 0.97, and 0.85 Normalized Discounted Cumulative Gain (NDCG) values. Overall, the experimental result shows that the proposed solution is able to discover and rank relevant resources according to users’ interest. |
