在傳統資訊系統開發過程中，需求分析通常被視為非常關鍵的步驟。但在Web-based的資訊系統開發中，需求分析卻很少被提及。這篇論文提要主要強調傳統資訊系統與Web資訊系統的差異，並提出一個社群導向系統的需求擷取架構。這個架構將需求擷取分成三個階段：初步分析、關鍵使用者需求分析以及一般使用者的反應。目前Web的開發技術以及社群設計的方法，通常只做初步分析，在系統開發過程中並不直接蒐集使用者的需求。本文提出的三階段架構強調關鍵使用者的重要性，並建議採用社會網路分析作為辦識關鍵使用者的工具。 關鍵使用者區分為資訊、溝通、交易以及娛樂四大類別。初步預試這四種不同類別的關鍵使用者是否會產生較多該類別的資訊需求，以及是否產生較多的需求總數。結果發現只有資訊的關鍵使用者產生較多的資訊需求以及總需求，而其他類別的關鍵使用者與產生需求數量上的關係都不顯著。而另一項有趣的發現是：20％資訊關鍵使用者可以產生大約80％的需求總數，與Pareto規則的預測相似。資訊關鍵使用者的意見是否可以代表全體的意見，預試的結果顯示80％以上的需求重要性評估，關鍵使用者與全體的意見沒有顯著差異。 實地研究採用則探究兩個線上社群，一為關係導向社群，另一則為興趣社群，用以探索使用者角色，涉入程度以及需求知覺間的關係。多變量共變數分析的結果顯示：去除涉入程度的影響後，使用者角色會顯著影響使用者對於需求的知覺，雖然影響的方式以及程度並不相同。高涉入的使用者對於需求的敏銳度，普遍比低涉入者需求要高，並不因為需求種類而有所不同，最後並討論管理的意涵以及日後的研究方向。 / Although the requirement analysis is generally considered a critical stage in traditional information systems development (ISD), but it does not get enough attention in most Web-based information systems development (WISD) or the emerging community-oriented design. The thesis highlighted the difference between ISD and WISD, and proposed a three-stage model of user requirements elicitation for community-oriented design. This model divides the requirements definition in three stages: initial analysis, key user requirements elicitation, and regular user responses. Most current WIS and community design methodologies consider only initial analysis or attempt to build common system architecture, and neglect actual users’ requirements. Key user input is emphasized in this model, and social network analysis (SNA) is proposed as a tool for identifying key users. The pilot study empirically tested the relationship between the number of key users and that of elicited requirements. The study applied SNA to identify key users (defined as their influence) in “information”, “purchase”, “communication” or “entertainment” networks, and then elicited their requirements of two WIS. The preliminary results demonstrated that the number of key users in “information” dimension was significantly correlated with the numbers of elicited “information” requirements and overall requirements. However, the number of key users in “purchase”, “communication” and “entertainment” dimension had no significant relationships with the number of the elicited requirements. The requirements collected from 20 percent of “key users” accounted for approximate 80 percent of total requirements, similar to the results predicted by Pareto’s rule. In addition, the representativeness of requirements from key users opinions was also tested. Two online communities were designed to explore the relationship between user roles, user involvement and users’ perception towards requirements. And the MANCOVA results showed that user role (with user involvement as a blocking variable) has a major impact on an individual’s perception towards requirements, though the difference varies in a certain way. User involvement has also a determining effect on a user’s perception toward each type of requirements. Managerial implications were also discussed.