Towards task models for embedded information retrieval

摘要: This paper investigates to what extent task-oriented user support based on plan recognition is feasible in a highly situation-driven domain like information retrieval (IR) and discusses requirements for appropriate task models. It argues that seeking tasks which are embedded some higher-level external context (e.g. travel planning) often exhibit procedural dependences; these dependences mainly due task; they can be exploited inferring the users' goals plans. While there clear need models IR account situational determinants of behaviour, required hybrid take both “planned” “situated” aspects. Empirical evidence points made reported from probabilistic analysis sessions with fact database experience plan-based state-based methods an experimental planning system