資訊搜尋歷程之學習與任務表現關聯之實證研究

Abstract

本研究以國立故宮博物院「書畫典藏資料檢索系統」與「器物典藏資料檢索系統」為研究個案，邀請博物館與博物館網站之愛好者與潛在使用者作為研究對象，並設計三項模擬受試任務，進行故宮典藏系統的使用者行為研究與網站評估。研究目的旨在瞭解受試者於典藏系統檢索相關資源時的搜尋歷程之行為，以及透過Vakkari (2016)之搜尋即學習(Searching as learning, SaL)指標檢視其學習行為。研究根據任務表現結果區分每個任務表現相對佳與差之受試者群組，並進而透過Zero-Order State Transition Matrices (ZOSTs)，與滯後序列分析(Lag Sequential Analysis)分析方法探索不同群組之搜尋樣式。研究發現透過搜尋形成、資源選擇與資源互動之搜尋歷程所歸納之SaL指標可觀察受試者之學習表現進而展現在最後任務表現上，故SaL與任務表現有正向關係並可反應搜尋品質。研究進一步發現表現較佳之受試者有較多關鍵字使用技巧、展現較多占比與高品質的資源互動；表現較差之受試者，無法進行適當的關鍵字檢索，雖然部分受試者有高頻率的資源選擇行為，但透過SaL指標可發現最終沒有好的表現。研究藉由採用的分析方法以及學習指標可深入了解系統對使用者探索故宮文物的支援效益，最後並基於使用者行為角度提出系統改善之建議。This study aims to explore users’ search and learning behaviors via the use of Digital Archives of Antiques Search System of the National Palace Museum in Taiwan. The study invited special interest museum visitors to join three simulated tasks using one of the systems. The purpose of this research is to observe the users’ search behaviors during the information search process and then investigate how learning occurs via examining the search as learning (SaL) criteria proposed by Vakkari (2016). After analyzing participants’ search behavior and learning outcomes, we differentiated them into better and worse task-performance groups. We then adopted zero-order state transitions (ZOSTs) to and lag sequential analysis (LSA) approaches to derive the significant repeating patterns from search paths for each task. The results show that users’ final task performance can be observed by SaL learning criteria of search formulation, source selection and source interaction stages of the ISP. That is, the users’ task performance has a positive relationship with the numbers of met learning criteria, which also reflects the quality of the search outcome. The results show that users with better task performance knew how to search information by keywords with high-quality interaction behaviors within the system. Users with worse task performance cannot make proper keyword searches. In addition, some even have a high frequency of selection of source behaviors but do not lead to good results, which also reflected on the learning criteria. The aid of the proposed analyzing methods and the adopted SaL criteria can help in the understanding of the effects of the targeted system in supporting the users to search for collections of Chinese art and antiquities.We also provide suggestions for refining the systems from the perspectives of users’ search behaviors.