應用 GOMS 於中文打字之績效預測

Abstract

由於中文文字的圖形特性，使得鍵盤式的中文輸入並非是最佳的人機溝通 方式，但它卻是目前最普遍的，是以有必要瞭解這項人機互動的行為。在 人機互動中，認知模式可幫助系統設計者瞭解人機介面設計對使用者的影 響，因此也成為人機介面評估、設計和人員訓練的輔助工具。基於認知模 式的重要性，本研究應用Card, et al. 於 1983 年所提出的 GOMS( Goals, Operators, Methods and Selection rules) 模式來發展中文輸 入作業的認知模式，並以此描述為基礎來預測專家(expert)輸入中文的時 間。為了驗證發展的模式之準確性，我們並設計了一個打字實驗。實驗的 分析乃針對個別受試者，預測其執行特定打字作業的時間。實驗結果顯示 ，中文打字時的各項活動（如知覺、認知、運動）有相當的複雜性；但是 本實驗所驗證的模式，仍能成功地預測專家執行打字作業的時間。 One critical issue in adopting computers in Taiwan is to design interfaces that enable users to easily input Chinese characters( Kanji). Kanji is inherently different from English. To input Kanji, a user must decompose a word into radicals, and input them through a traditional English keyboard on which the radicals are marked. It appears that such Kanji entry involves susbstantial cognitive(e.g., momory, motor, and perceptual) effort. Previously, Card et al. have proposed GOMS(Goals, Operators, Methods and Selection rules) to model cognition involved in English data entry. GOMS is a way of describing what the user needs to know and to do in order to perform computer-based tasks. The METT(GOMS Model of Expert Transcription Typing), an extension of GOMS for modeling nonsequential component processes, has been shown useful for predicting the time to execute English typing tasks. The purpose of this thesis is to study if METT can be applied to predict the performance of Kanji entry. Through such analysis, we hope to understand cognition involved in Kanji entry. An experiment was performed to validate the applicability of METT to model Kanji data entry behavior model. The research findings indicate that METT can be used to predict performance with acceptable accuracy. In addition, based on the research finding, a new model, MECT (GOMS Model of Expert Chinese Typing), is proposed. A preliminary analysis shows that MECT can predict performance with better accuracy than METT. However, this difference between two model's predicting power is not indicative of their respective theoretical soundness. Instead, it shows that more studies are needed to understand the complexity of cognition involved in Kanji entry.