GWAP系統之設計策略研究－以ESP為例

Abstract

「人智運算」在電腦科學中是一種創新的概念。其概念目前是有許多事情是人能夠輕易解決而電腦做不到的事情。尤其在2004年ESP遊戲被推出之後開啟Games With A Purpose (GWAP)這塊新的領域，其目的是在玩家進行遊戲的時候也在幫系統標記圖片。這些圖片的標記能夠用在影像辨識和影像搜尋上，而且圖片標記的正確性和圖片標記產生的速率都非常高。ESP遊戲的成功使得人智運算和GWAP變成大家所矚目的焦點。GWAP是讓玩家在進行遊戲時不知不覺地幫系統進行運算，並產生對系統有用的資料。因為玩家覺得遊戲很好玩，所以玩家會一直沉迷或是狂熱地進行遊戲，並且幫系統進行運算產生有用的資料。本論文是探討如何增加GWAP系統的效能，此處針對ESP遊戲進行分析，並定義測量系統效能的函數，且提出能夠增加系統效能的圖片選擇演算法Optimal Puzzle Selection Algorithm (OPSA)。為了驗證OPSA確實能夠增加ESP遊戲的效能，實作實驗系統ESP Lite進行驗證。在二個月的實驗過程中，實際證實OPSA是能夠增加系統效能的圖片選擇演算法，並且發現OPSA的行為和玩家的行為有著相當大的關係。從此研究可以發現有策略地運行ESP遊戲系統就能增加系統的效能；同理其他的GWAP系統只要有策略地運行也同樣能增加系統的效能。“Human Computation” is an innovative concept in computer science. The idea is there is a lot of things that human can easy do that computers can not yet do. In 2004, the ESP game has been launched and it created an emerging field, Games With A Purpose (GWAP), in computer science. The objective of GWAP is creating difficult metadata when players are playing the game voluntarily. This thesis presents how to enhance GWAP systems. In this thesis, we use the ESP game as an example and propose a puzzle selection algorithm, Optimal Puzzle Selection Algorithm (OPSA), to enhance the ESP game system. For validating our proposed algorithm, OPSA, which actually enhances the ESP game system in real world, we implement a quasi ESP game, called ESP Lite. During a two-month experiment, we investigated the inner properties of the OPSA, and verified that the OPSA scheme achieves the best system gain for the ESP game system. The results of this thesis also confirm that GWAP systems are more efficient if they are designed and played with strategies.