Conjunctive queries (CQs) fail to provide an answer when the pattern described by the query does not exactly match the data. CQs might thus be too restrictive as a querying mechanism when data is semistructured or incomplete. The semantic web therefore provides a formalism—known as (projected) well-designed pattern trees (pWDPTs)—that tackles this problem: pWDPTs allow us to formulate queries that match parts of the query over the data if available, but do not ignore answers of the remaining query otherwise. Here we abstract away the specifics of semantic web applications and study pWDPTs over arbitrary relational schemas. Since the language of pWDPTs subsumes CQs, their evaluation problem is intractable. We identify structural properties of pWDPTs that lead to (fixed-parameter) tractability of various variants of the evaluation problem. We also show that checking if a pWDPT is equivalent to one in our tractable class is in 2EXPTIME. As a corollary, we obtain fixed-parameter tractability of evaluation for pWDPTs with such good behavior. Our techniques also allow us to develop a theory of approximations for pWDPTs.

中文翻译：

---

设计良好的带有投影的模式树的高效评估和静态分析

当查询描述的模式与数据不完全匹配时，连接查询 (CQ) 无法提供答案。因此，当数据是半结构化或不完整时，CQ 作为查询机制可能过于严格。因此，语义网提供了一种形式主义——被称为（预计）精心设计的模式树(pWDPTs)——解决了这个问题：pWDPTs 允许我们制定查询，如果可用，我们可以在数据上匹配部分查询，但不要忽略其余查询的答案。在这里，我们抽象出语义 Web 应用程序的细节，并在任意关系模式上研究 pWDPT。由于 pWDPT 的语言包含 CQ，因此它们的评估问题是棘手的。我们确定了导致评估问题的各种变体（固定参数）可处理性的 pWDPT 的结构特性。我们还表明，在 2EXPTIME 中检查 pWDPT 是否等同于我们的易处理类中的一个。作为推论，我们获得了具有如此良好行为的 pWDPT 评估的固定参数易处理性。我们的技术还使我们能够开发 pWDPT 的近似理论。