Tables with Nulls and Functional Dependencies: Explanation and Quality of Query Answers

Several applications today deal with tables that are the result of merging other tables coming from different sources (as when recording the results of collaborative work or when merging tables during data staging in data warehouses). Such tables usually have missing values (also called nulls) and/or contain data that do not respect given constraints (such as key constraints). In this paper we study the influence of nulls and/or inconsistent data on the answers to queries to such tables by (a) providing to the user explanations regarding the expected presence (or absence) of certain tuples in the answer and (b) by defining measures for assessing the quality of query answers.