Research On The Usability Of Preference Query Results

The preference query(e.g.,reverse top-k query,reverse skyline query,etc.)is one of the important queries in the database community.Based on the user preferences,the preference query returns the query results,which match user preferences best,to users.The preference query has a wide range of applications such as multi-criteria decision making,personalized recommendation service,and so forth.Currently,queries only return the query results to users.If the query results are unexpected for the users,they may frustrate users.However,the database system neither gives explanations for the unexpected query results nor offers any suggestions on how to obtain the expected results for users.The users only can debug the queries by themselves,which is cumbersome and time-consuming.If the database system can provide such explanations and suggestions,it helps the users understand initial query better,and know how to change the query until the satisfactory results are found,hence improving the usability of the database.Towards this,the studies on the usability of query results have been explored,which can make database system easier to use.However,the usability of query results is query dependent,meaning that different queries require different solutions.In addition,existing efforts on the usability of query results mostly focus on the relational database query.Therefore,the existing techniques cannot effectively solve the problems of usability for preference query results.Motivated by these,in this article,we systematically explore the usability of preference query results.Note that,due to the space limitation,we aim at the reverse top-k query and the reverse skyline query,two representatives of prefer queries.The key contributions of the article are summarized as follows:(i)The causality and responsibility problems on reverse top-k queries and reverse skyline queries.If the query result contains the user's unexpected object(s),or the user's expected object(s)does not appear in the query result,users may eager to know what causes the appearance of the unexpected object(s)and/or what causes the absence of the expected object(s).To this end,we explore the causality and responsibility problems on reverse top-k query and reverse skyline query,where causality represents the causes for answers/non-answers to queries,and responsibility quantifies the effect of a cause,which reflects its influences on the answers/non-answers to queries.(ii)The why-not and why questions on reverse top-k queries.The study of causality and responsibility only offers the explanations to the users.Usually,users would also like to know how to get the expected query results.In view of this,we explore the why-not and why questions on reverse top-k queries.Specifically,the why-not questions aim to make the expected object(s)appear in the query results and the why questions try to remove the unexpected object(s)from the query results.(iii)The why-few and why-many questions on reverse top-k queries and reverse skyline queries.In real applications,the query may return too many or too few(even empty)answer objects to users.If the answer objects are too many,users may be overwhelmed by the large information;and if the answer objects are too few or even empty,there is no selection for users.Both of them are undesirable for users.If the database system can provide the suggestions on how to refine the original query to satisfy the query result cardinality constraint,it would offer users useful information for decision making.In light of this,we investigate the why-few and why-many questions on reverse top-k queries and reverse skyline queries.To be more specific,why-few questions aims to increase the answer objects and why-many questions strive to decrease the answer objects.(iv)The reverse top-k query result analysis system.Based on the above proposed techniques,we develop an interactive system to analyze the unexpected reverse top-k query results.According to the feedbacks from users,the system offers the explanations of the unexpected query results and the suggestions on how to obtain the expected query results to users.