A Domain Specific Language for Describing Facebook Interaction Graphs

因大量臉書用戶留下不少行為紀錄，吸引研究人員蒐集臉書資料，從中挖掘具有價值的潛在資訊與研究成果，並用 Graph 點線關係圖視覺化呈現成果。然而透過臉書 Graph API 蒐集資料的方法，自 2.0 版起漸漸提高了限速限制，造成蒐集資料開始是一個問題。雖有研者人員提出應對方法，卻也產生用戶識別度弱化問題。故本研究動機是提出一套工具能有效協助研究人員進行臉書資料蒐集與建立關係圖以利進行分析與研究，在進行相關臉書蒐集資料與社會網絡與社群網絡分析文獻探討後，實作一種基於爬蟲技術，蒐集具用戶識別度之臉書資料；以及提出描述臉書互動行為的關係圖領域專屬語言；實作對應之使用者操作介面工具，讓使用者透過圖形化操作方式描述定義臉書粉絲頁、使用者、Hashtag、留言及回覆與貼文之間互動行為建立關係圖。經過本研究實驗設計與驗證，證明蒐集資料具識別度、產生的關係圖是正確以及問卷調查指出 75% 受試人員認同本系統可快速建立關係圖。未來可以結合爬蟲與關係圖領域專屬語言線上即時建立關係圖，以及導入更多社群分析功能，以利研究人員可以在本系統上進行更深入的分析與研究。The huge amount of digital footprints of Facebook users have become a good research resource and we have seen many good results developed by collecting data from Facebook and visualizing it to node-link diagram. However, there are more and more rate limits in Facebook Graph API since version 2.0. Although one could overcome the restriction of rate limits by creating more Facebook App resources, yet this makes user identification become cumbersome in analyzing the retrieved data. Therefore, the motivation of this research is to provide a set of tools to assist researchers in collecting recognizable Facebook user data with a crawler and propose a domain specific language (DSL) to build the relation graph by describing Facebook interactions. With our crawler and the DSL tool, we are able to gather unique user data successfully. The experimental results show that we can construct node-link diagram conveniently, and 75% of the surveyed subjects agreed that our tool could be helpful for building graph