探勘群眾資料以用於開發旅遊應用的拍照和遊記素材推薦系統

由於行動網路與智慧型手機的普及且快速的發展，越來越多人使用智慧型手機取代數位相機，在旅遊過程將所見所聞拍照紀錄，並利用行動網路分享到社群網站。因此，網路社群所累積大量的照片和關連的標籤與位置資訊，讓資料探勘方法可能由海量的群眾照片資訊，探索使用者對於地標風景區的喜好，以及拍照地點，藉此分析地標的熱門旅遊區域，與行動App 相關服務整合。多數人也喜歡透過撰寫部落格來記錄或分享旅遊心得，但是需要花上許多時間來整理相片、搜尋相關資源。基於以上動機，本論文設計了四個子系統，適地性搜尋引擎 (Location-Based Search Engine, LBSE) 根據位置對應的地圖區塊事先建立索引，可藉由使用者提供的目前位置，立刻回覆附近相關物件資訊。標籤推薦系統 (Tag Recommendation System, TRS) 依據位置提供適合標註於照片的標籤清單。地標區域識別系統 (Landmark Area Recognition System, LARS) 依據位置判斷使用者位於哪些地標的範圍內，告知使用者是否為拍攝地標的好位置，可依照地標來推薦標籤以增加TRS 推薦標籤的準確度。旅遊部落格產生器 (Travel Blog Generator, TBG) 依據使用者標註的相片標籤，推薦相關的標題或描述，並且針對內容敘述的相關關鍵字，自動提供智慧標籤和相關參考連結。本論文整合四個子系統實現於iTravel 平台，並提出實驗評估，從實驗觀察中改善系統服務。

關鍵字

標籤推薦；部落格產生器；適地性搜尋引擎；地標範圍識別；標籤關聯探勘；文字解析

並列摘要

As the growth of smartphones and mobile networks, more and more users carry smartphones instead of digital cameras on their trip. Users are familiar to take photos with smartphones, and share photos to friends on social networks. Therefore, massive number of photos with location information and related tags are accumulated to be a large data set for applying data mining methods to explore user’s preferences on landmarks and photographic locations so that extracted information for popular landmarks can be integrated with mobile applications (Apps). Many users also like to write travel blogs on blog websites, but tedious works for collecting citations and thinking words spend users a lot of time. Based on aforementioned motivation, we develop four subsystems to integrate into a complete backend system for travel applications. First, Location-Based Search Engine (LBSE) creates indexes for geographic objects by partitioning the world map into blocks with BlockID estimated from the GPS position. Then, Tag Recommendation System (TRS) recommends significant photo tags by given the location information. Landmark Area Recognition System (LARS) identifies the area of a landmark based on crowdsourcing photos and tags obtained from social sites. Finally, Travel Blog Generator (TBG) recommends title, annotation and description for a tagged photo and provides citation links for some written keywords while writing travel blogs. This smart tags service is useful for blogger to write articles efficiently. These subsystems are implemented and integrated into the iTravel platform. We also do several experiments to evaluate the performance of proposed method or system.