語意關聯式的新圖文整合－以陽明山國家公園研究報告為例

Abstract

以往GIS領域在如何增進使用者介面親和力有許多的切入面向，自然語言即為其中一種，但自然語言缺乏正規的表達方式，常造成關鍵字搜尋與語意式查詢間的語意落差，導致查詢行為受到限制。 本研究嘗試援引「語意網」的技術於地理知識搜尋領域，並分成建立語意網、定義關係、定義反應行為、串聯、聚焦及擴充六階段，透過Java為系統撰寫語言，結合中研院詞性標註系統、Protege及ArcGIS等工具，建構一基於語意網的圖文整合平台，期能於閱讀文件時適時提供空間元素對應的GIS資料，並經由文件斷詞、語意推論及圖台展示等過程，改善現有空間知識獲取的方式，最後再以五種模擬情境驗證系統，分析語意推論引擎介入前後影響搜尋結果的差異，實驗結果證明推論引擎可有效提升符合語意的搜尋結果滿意度。 本研究之成果，首先為調整現存之語意網建立與應用方法，加以正式化後用於空間知識搜尋領域，及本研究所建立之語意推論模組，可作為發展增加使用者語意彈性的搜尋引擎所使用，最後再將提出的概念落實於實務中，藉以改良現有之圖文整合架構。由新架構獲得的搜尋結果更能滿足人類的語意搜尋求，使閱讀的方式除了心象圖的產生之外，更有實際客觀的資料可加以輔助及驗證。

There have been many approaches to using natural languages to make user interfaces of GIS more friendly. However, the lack of systematic expressions in natural languages frequently causes semantic gaps between users’ keyword searches and semantic inquires and thus inhibits functions of inquiry. This paper applies 'semantic web' technology to search for geospatial knowledge, with the application involving the following six steps: establishment of semantic webs, definition of relationships, definition of the behavior of reactions, connection of multiple semantic nets, focus of the inference, extension the semantic net. This paper uses the Java language to combine tools such as Taiwan Academia Sinica's 'Chinese Parsing System', Protege and ArcGIS to create a semantic framework to integrate documentations and GIS, providing GIS data relevant to the document in use. In addition, Chinese Parsing Module, Semantic Matching Engine, and GIS display are used to improve existing methods of obtaining geospatial knowledge. Finally, this paper uses five scenarios to test the system mentioned above and analyze the effect of the Semantic Matching Engine in improving the performance of search. The results show the Semantic Matching Engine is significantly effective in this respect. This paper aims to make adjustment in existing establishment and application of semantic webs to specifically make them suitable for use in search for geospatial information. In addition, the Semantic Matching Engine developed in this paper can be used in search engines for increasing users’ semantic flexibility. The concepts in this paper are expected to be practically used to improve existing frameworks of GIS information integration to reach better performance in semantic searches.