自然語言式GIS查詢介面－以大安區餐飲服務為例

Abstract

GIS的使用者已逐漸擴展到一般大眾，是以使用者介面的方便性與親和性日益重要。目前一般大眾會使用到GIS的時機多半是上網瀏覽電子地圖網站，作主題地圖或者是道路、特定地標以及地址查詢。綜觀這些較為大眾使用的查詢介面，其提供查詢的格式固定且無彈性，使用者必須按照系統定義好的特定欄位，輸入關鍵字後方可執行查詢；且無法讓使用者以空間物件間的空間關係作查詢。是以，為增進使用者在使用GIS查詢介面時的便利性，本研究擬以自然語言切入，設計建構一個可處理空間關係的自然語言式GIS查詢介面，以增進使用者作查詢時的彈性與便利性。 本研究中，首先透過文獻回顧了解空間查詢問題之主要類型及在文字中空間敘述的型式，並整理目前中文自然語言處理方式之主要類型。在研究方法部分，實際調查一般日常生活中之空間查詢語句，整理語句並抽取出其主要之構成元素、歸納出數種主要查詢語句之類型。自然語言處理部分則採辭典法做斷詞，並根據所整理出之語句訂定其後續處理之語意語法規則。 系統實作部分，採大安區餐飲服務為示範案例。本研究所提出之查詢介面，使用者可直接輸入自然語言空間查詢語句對系統執行查詢，系統接收所輸入之語句後，內部進行斷詞處理、語意處理以及空間查詢處理後，會直接回覆使用者查詢出的地圖資料及相關訊息。此查詢介面其應用範圍亦不限定於餐飲服務之查詢，可抽換資料庫應用在不同種類的生活資訊查詢上。在系統測試部份，30個測試案例有73%（22個案例）可以完全斷詞並正確執行後續查詢。即本研究所建構之查詢介面，不僅提升了查詢上的便利性，亦在正確率上有一定之水準

The users of GIS gradually expand to the general people, so that the convenience of the user interface gets more and more important. Nowadays, people who use GIS mostly browse map website then execute spatial query. However, these map websites which are the most popularly used GIS query interfaces are not flexible for using. People who want to use the query interface must follow the predefined specific query type, and they can’t query by spatial relations. In order to increase the convenience and flexibility of spatial query, therefore, the nature language based approach was applied to be the guideline of the query interface design, and we develop and construct a natural based query interface which can be queried by spatial relations. In this research, first, we reviewed main types of spatial query questions, spatial referencing, and Chinese natural language processing. Second, We collected everyday spatial query sentences, extracted the main composition elements and concluded several kinds of main query sentence type. Finally, in natural language processing, we process word segmentation by dictionary approach and set the processing semantic and syntax rules. We take a case study on Food and Beverage service of TaiAn District in Taipei. Users can type natural language sentence and then execute spatial query. After executing word segmentation, semantic and syntax processing, the system will response user the answer by map and message on screen. If the system’s database changed, then the interface not only can be used in food and beverage service but also can be used in other kinds of service. The system was evaluated by using 30 examples and got 73% success in word segmentation and executing query. Therefore, it not only can improve the flexibly and convenience of spatial query but also has high performance in accuracy.