認識城市文化的一種方法是了解這個城市的人們閒暇時會做什麼。實際上，尋找當地活動和促銷活動是大多數人在旅行或搬到新城市時的普遍需求。然而，對於活動發佈者如政府、團體、公益組織或是小型企業而言，將此類消息傳遞給合適的人仍然是一個挑戰。雖然現今已有Facebook活動事件，以及其他售票網站、或是活動發佈網站提供了活動註冊功能以滿足此類活動信息傳遞的需求，但是大多數組織除了在自家的網站公佈消息之外，則會通過Facebook等社群網站上發布來傳遞訊息的，以期活動訊息可以接觸更多人。因此在本計畫中我們考慮使用信息技術從Facebook等社群媒體上的貼文來擷取活動事件、建立活動事件資料庫，以提供社交活動事件搜索服務、以及探索每個城市或區域的發展樣貌。我們預計應用Web抓取、Web資料擷取、以及活動名稱辨識和活動場所識別等自然語言處理技術，來完成活動事件資料庫的建置；同時為提供活動查詢功能，我們也必須提供活動位置的定位，以及對不同資料來源所獲得的活動事件進行去重複的偵測；最後我們也將對活動事件進行活動分類，進而分析每個城市或區域的發展型式，提供城鄉發展規劃的參考資訊。本計畫預計分成三年完成：第一年我們將專注於Web抓取模塊的建置，以不斷監視社交媒體(例如Facebook)、售票網站、以及各組織網站上的新活動貼文；並從下載的貼文中進行活動事件檢測和活動事件名稱識別。第二年的目標則在於活動場地識別和地理定位；由於活動場所可以是地址，興趣點(POI)或一塊區域(稱之為GPE或地理政治實體)，而且一條消息中可能提到了多個興趣點，因此活動場地識別也必須決定正確的地址及興趣點，再將地點的GPS位置定位於地圖上，以強化活動搜尋服務。最後第三年計畫則包括整合不同資料來源，進行重複事件檢測、活動事件的分類、以提供城市或區域的活動視覺化分析，提供城鄉發展規劃的參考資訊。 ;One way to explore a city is to know what people do on their leisure. In fact, looking for local events and promotions is a common need for most people during travel or moving to a new city. However, delivering such messages to the right persons is still a challenge for small business that do not sell tickets on high-end websites. Instead, most events are usually distributed by posting on social networking sites like Facebook. To fulfill such information need, we consider information technologies to extract events from 230K Facebook fan pages to build an event database and provide a social event search service. The technologies includes web scrapping and web data extraction as well as natural language processing for event names and venues recognition. The project is expected to be completed in three years: In the first year, we will focus on the web crawling modules to continuously monitor event-registration websites, ticket-selling websites, social media (such as Facebook) for new event posts, and download posts for event announcement detection and event name recognition. The goal of the second year is to recognize event venue and locate the venue on the map. Because event venues can be an address, point of interest (POI), or a geopolitical entity (GPE), and multiple POIs or locations may be mentioned in a message, we also need to determine the correct POI and locate the position on the map to strengthen the event search service. To integrate events extracted from different data sources and for the purpose of event analysis, in the third year we plan to deduplicate events in the event database and conduct events classification such that we could better provide event statistics for urban and rural development planning from different perspective (such as time, space and event categories).
