T Wu CS511 Project DataScope A Database Content Visualization Tool based on Ranking Queries Tianyi Wu twu5 uiuc edu CS 511 Course Project Report Dec 12 2006 1 Introduction People have been relying on Google Maps MapQuest or other similar services to find desired locations on maps browse surrounding businesses get driving directions etc Navigation by clicking and dragging the mouse to browse maps at multiple levels of resolutions is one of the most attractive features in Web based map exploration Most database systems though with some graphical user interfaces are still lack of data content browsing based interfaces Motivated by Google Maps we develop DataScope a Web based data content visualization system for people to view the desired data easily interactively and at multi resolution There are several challenging issues for developing DataScope 1 unlike maps which use welldefined and well understood numerical metrics of longitudes and latitudes to position object on the screen there is no universally accepted way to design layout and visualize data points for relational databases 2 different users may have different preferences on the importance of particular data values and on the weights or relevances of participating attributes moreover even the same user may have different preferences on the data based on the context of the query 3 the data is often organized in multiple relations in a database or associated with multiple hierarchies or multiple semantic links and finally 4 a powerful query engine is the key to enabling a smooth and fast user interaction The above challenges have been addressed in the design and development of the DataScope system prototype First we show our system is radically different from the existing database or data warehouse visualization systems such as Polaris 3 and DIVE ON 1 since none of them supports intuitive map browsing operations for explorative analysis of data Although a few interactive tools can support certain data exploration they are confined to particular domains e g spatial temporal data predefined schemas or fixed visual representation e g statistical charts and thus lack flexibility to browse various relational database contents Second our system has a few interesting features Its Web based graphical user interface provides a set of intuitive operations such as rolling up drilling down zooming and panning to help users navigate the database contents easily and quickly The view of data is customizable based on different database schemas Further we not only provide users with data summarization through visualizing multi dimensional aggregations but also support interactive visualization of database contents in multiple resolutions Third for smart navigation and efficient implementation the system has incorporated several lines of advanced research on query processing especially on ranking query answering including top k query answering with ranking cubes 5 and ranking over aggregations 2 and so on 1 T Wu CS511 Project 2 DataScope Overview 2 1 Architecture Figure 1 DataScope Architecture As depicted in Figure 1 the DataScope system consists of three main components Web interface user interaction server and the query engine The Web interface provides a familiar environment similar to Google Maps for user to explore the database The user interaction server translates user interface events and parameters into query statements recognizable by the query engine The query results are returned to the interface via this layer and then properly rendered This layered approach allows the interface to be independent of the query engine and thereby ensuring the extendibility and flexibility of the system 2 2 User Interface Overview As a key component of the system the Web GUI of DataScope is as intuitive and expressive as Web mapping services like Google Maps Furthermore a user can load different datasets and conveniently customize his her views As shown in an imaginary screenshot Figure 3 the interface has three panels data map function panel and information panel The imaginary screenshot shows a typical scenario when a user is browsing the DBLP http www informatik uni trier de ley db data set with a particular interest in recent database conferences The X axis labeled with Year displays years 2000 through 2006 One can click the left arrow to view previous years This panning can also be done by dragging the entire data map The Y axis labeled with Conference shows a list of database related conference names ordered alphabetically Like many others the conference attribute can be ranked and we use a yellow bar to indicate each conference s ranking score based on the influence factor mainly obtained from http citeseer ist psu edu One can drag the map up down to see conference names beyond the current alphabetic range e g APWeb However some relevant conferences within the alphabetic range such as DASFAA and PAKDD are not displayed because only the highly ranked conferences are shown given the limited space on the axis If the user wants to see lower ranked conferences she needs to click the zoom in button marked with a plus sign Note that zooming does not change the number of items displayed on the axis instead it changes the range of items displayed Thus zooming in would narrow the range and allow lower ranking items to appear The zooming buttons are faded out for the year attribute because it has not been associated with any ranking function On the X Y plane a number of authors are displayed each representing the most productive author in the context of the corresponding cell Notice that in 2 T Wu CS511 Project addition to this spreadsheet style view the data map can also represent various continuous attributes Overall the data map panel facilitates users to drag the map around to view adjacent data and click to see the data at finer resolutions Figure 2 An Initial View Conceptual for the DBLP Data Set Figure 3 A Screenshot Conceptual for the DBLP Data Set 3 T Wu CS511 Project The function panel supports basic operations including open close database view schema as well as two important functions searching and view personalizing Searching helps users do keyword or advanced search including constructing complex query predicates whereas view personalizing allow users to customize map objects layouts ranking functions visual parameters default initialization etc Their screenshots are not shown due to limited space Finally the information panel which