FOCUS: The Interactive Table for
Contents
1. Overview Attributes 51 differ Show All Delete Identical Auto Delete Ix Keep Specifications AT J m 7 D x s E EE a mum E s zr CS eC Vendo EHWOEICIDAIMGDIEHBMFcFargo Electronics LeCErCC ErCAHewleT DIC AEFC Hewlel D NECDICSICIMOETL T AE THOMGID AXING Ca eC HOG HIGG dGrFartTekDiGenT Contact M TEE EEE Technology Dot Matrix Dy Dye sub WMkJet laser Mew 90 8 pL a o e o eB roe e oo e o c d ee e cs TT Resolution TTT TTT o c SLERLLLLLLEEEELLLLLELEERLELLEEEER ELLE LELEELEL LL LL EEEE FEE LLL ETLEE LLL LL EEE LL LL L E amp jwr caidp 300 30 80 600 300 horizontal dp FOO el 360 300 30 6 6 is 300 Interfaces RR RR RR RR RENE RR RR RR RR Supported Interfa e e e WW e a i WM Mas sep Auto Switching FURAR RARARS RRR RS 422A Ri Ride Ri Ri LRR R RSCEIIRIRIRA RERILRRRRS RLRRRRRRRRRRARS HRR LORIRIRARIRIRA RARSCEERIRUR E Localtalk Lu c Rid E il Ft Ri hi Li hi Ri i Ri Ethernet Te d RS2320 5422 C RS 232C RS 422 SCSI Reset LI BPTInTer H Marc color2. could simply restrict the table to the desired price range and the three neighborhoods If the table were then sorted by Lot Size and afterwards by Neighborhood Jennifer would see the desired houses at a glance This technique which 1s closer to browsing than to formal database queries is more convenient for the end user and is ideally supported by FOCUS Also in the first scenario summary statistics are generated for the remaining houses Figure 8 shows how the complete information distributed over several display areas in the system of Goldstein and Roth 6 is displayed in one clear ly arranged table The user only had to sort by Bedrooms and then to select Summary per Bedrooms from the menu of Cost In Figure 8 the mean price per neighborhood has also been determined Second Scenario In the second scenario John asks for the company and then the sales agent who sold the most houses from a given subset To answer this question new aggregates are de fined In FOCUS however two mouse clicks are sufficient to sort first by Sales Agent and then by Company John can see at a glance which agent of which company spans the most columns A WORLD WIDE WEB TABLE BROWSER FOCUS 15 ideally suited as a browser for product informa tion to be distributed on the World Wide Web Tables on the WWW share the same problem with their counterparts on paper if they grow too large they become very unwieldy to manage Currently the only way
3. removes the gray rows If Auto Delete 1s turned on for the attributes this is performed automatically after each query As a default setting we recommend Auto Delete Records but not Auto Delete Attributes because users can adjust to the fact that some records are removed after a query but may be slightly confused when attributes suddenly disap pear or move to a new position A common technique 15 to focus on a subset of records first and then to delete the irrelevant attributes and records Con tinuing from that point with both Auto Delete buttons turned off the window contents will remain stable When queries are executed to further analyze the subset some rows and columns may be grayed but each cell and each value remains at its position File Edit Window Table FOCUS HOMES2 FOC Records 179 qualified Show All All Delete Refused Retused I Auto Delete Ix Reset Attributes 8 differ Show All All Delete Identical Identical Auto Delete Ix i um T E m MD MAU sls LU dl LE eset s ee 3 456121342345 2 343 4 562 3 4 Bedrooms 0 3 14 56 Garage e Neighborhood in Adams Morgan Anacostia Arlington Central Air yes New yes Figure 7 The Dynamic HomeFinder database DISCUSSION To compare our approach to Dynamic Queries we have also used FOCUS to explore the database of the Dynamic HomeFinder demo 1 Figure 7 shows a part of the dat
4. to display databases on the WWW is to use a text field to enter a query and to display the query results as a new page With FOCUS medium sized databases can be included as full tables within a page and can be interactively viewed in place For this purpose FOCUS tables have to be registered as a MIME type The tables are quite small compared to typical graphical images For example the printer table presented here occupies only 32 kB as an ASCII file There are three different ways of viewing tables on the WWW with FOCUS e Configuring the WWW browser to use FOCUS as an external viewer for tables This is the way FOCUS is currently implemented The advantage 1s that FOCUS 15 just an arbitrary application and is only loosely related to the respective Web browser The disadvantage 1s that the table resides in a different window and provides no link back into the Web e Implementing FOCUS as a plug in for the Web browser In this solution the table viewer is more di rectly integrated into the Web browser but the disad vantage is that the integration is browser dependent e Implementing FOCUS as a JAVA applet This solution makes FOCUS platform independent and usable with most future Web browsers We would like to create a new standard for handling product selection in the WWW It is therefore in our inter est to make FOCUS as widely available as possible REQUIREMENTS FOR DATA STRUCTURE FOCUS is applicable to a wide range of t
5. ut nil olor NENIA RET Superscrip ATT ATT izen TO aser 6000 OL 40e aser 1500 bie I E 1 BP 8 BM aserPrinter 4039 Laser Jg 45 ode electPress 600 BOT ie T k x 1570 pn rj E i ccel a VWiter 8200 7080 Unity TROUXL O Primera Color Printer PrimeraPro Color Bhaser 20e Phase 20 PrimeraPro Color Print ing to different criteria Locattak 9 je e p 9o jS Viol Ethernet o amp o o OM ES SOME lee Bde 249 809 8 845669 amp TokenRing o e jS SCSI Memory Processor BYTE Rankings Draft ppm Postscript pp Monochrome aua E Quantyindel jjemeesseem sends ___________ Lal ll I SENKEGENMC DTCUNHENNIMCT O Nothing specified If sorts are performed after a column has been selected the user can watch where the selected printer is placed accord Figure 1 An overview of the printer table Normal Mode The user can switch to normal mode using the Overview check box In normal mode a larger user defined column width is used and the table can be horizontally scrolled to view all entries The table will automatically scroll to the first selected column so that the focus is not lost The nor mal column width can be adjusted by
6. FOCUS The Interactive Table for Product Comparison and Selection Michael Spenke Christian Beilken Thomas Berlage GMD German National Research Center for Information Technology FIT MMK 53754 Sankt Augustin Germany E mail Spenke gmd de ABSTRACT FOCUS the Feature Oriented Catalog USer interface 15 an interactive table viewer for a common kind of table namely the object attribute table also called cases by attribute table or relational table Typical examples of these tables are the Roll Calls in BYTE where the features and test results of a family of hardware or software products are compared FOCUS supports data exploration by a combination of a focus context or fisheye technique a hierarchical outliner for large attribute sets and a general and easy to use dyna mic query mechanism where the user simply clicks on desired values found in the table A PC Windows implementation of FOCUS 15 publicly available http www gmd de fit projects focus html It is suited for tables with up to a few hundred rows and columns which are typically stored and maintained by spreadsheet applications Since we use a simple data format existing tables can be easily inspected with FOCUS With the rapidly increasing public interest in on line servi ces like the World Wide Web we expect a growing demand for access to on line catalogues and databases FOCUS satisfies this demand allowing formulation of simple data base queries with an inte
7. a base While the Dynamic HomeFinder can display the complete data of only one record at a time FOCUS gives a very informative overview of the data Without performing any further queries one can discover the following facts about the shown homes They are distributed over 7 neighborhoods All neighborhoods but one are in the MD area In Bethesda there are only houses In College Park there are only condos Bethesda is the most expensive neighborhood Fireplaces are found in only two neighborhoods Most homes in Beltsville have 3 Bedrooms No house has more than 6 bedrooms And many more Of course it is appropriate to show the geographical loca tions of houses for example with a starfield display but a general tool such as FOCUS cannot create this kind of app lication dependent visualization Goldstein and Roth 6 use a similar real estate database to discuss two scenarios First Scenario In the first scenario Jennifer searches for houses between 100 000 and 150 000 with a lot size larger than 5000 sq ft in the neighborhoods Shadyside or Squirrel Hill or in the same price range a lot size larger than 8000 sq ft in Point Breeze This disjunction of two queries is performed by first defining two new aggregates SqHill Shady and PtBreezeBigLot for the two queries Then a third new ag gregate Interest Houses 1s introduced to construct the union of the two aggregates File Edit Window Table m Recor
8. ables The basic precondition is that the table has to be indexed by products and attributes Such a table is similar to a flat relation If there are multiple relations concerning the same entities these relations have to be joined to make the result viewable with FOCUS Therefore most relational databases will be good candidates for presentation by FOCUS For product descriptions subclass relationships frequently exist between different product categories That means that for a specific subclass of products further descriptive attributes are needed In FOCUS several such subclasses can be integrated into a single table for comparison by taking the union of all attributes and by leaving the attribute values empty for all undefined attributes In the table they will be displayed as not applicable Usually the hierarchical attribute system can be used in such cases to improve clarity Attributes for a subclass are grouped under a particular header The user can hide all these attributes with a single click and redisplay them if objects of this subclass are selected The input to FOCUS is a simple tab delimited ASCII file Such a file can be generated with a minimum of effort from many popular applications such as spreadsheets simple databases card files and even tables in text files APPLICATION FIELDS There are a number of applications that illustrate how FOCUS can be used The most common examples are technical product da
9. an also be used to permanently display the set of possible values of an attribute Excluding Values Using the Exclude button on the right hand side of the table the user can exclude one or more selected values from the table Exclude is equivalent to a negated Set i e a restriction to the complementary set Excluding a few values 1s sometimes easier than allowing all other values Query Representation Note that FOCUS not only displays the result of a query but also the user specifications which led to the result Lines are displayed in yellow or blue to indicate constant or restricted attributes The selected subset can be viewed in the dialog box of an attribute Furthermore a summary of all user specifications 1s displayed in the status line at the bottom of the FOCUS window Figure 2 COMPLEX QUERIES With a minimum of additional user interface concepts quite complex database queries can be defined and exe cuted with a few mouse clicks Even though they are simple to state complex queries rarely have to be used because their results can often be directly observed in the table This distinguishes FOCUS from database queries because all current objects are visible before a query is executed Disjunctions When multiple values are selected in different rows FOCUS proposes a disjunctive query of the form lt attributel gt lt valuel gt OR lt attribute2 gt lt value2 gt In Figure 4 the table has been restricted to
10. dragging the border between two columns Columns can also be selected in normal mode They remain selected when the user switches back to overview mode where they are shown highlighted and expanded Figure 1 TABLE RESTRICTION The primary way to explore a FOCUS table is to restrict the number of products columns to an interesting subset Constant Attributes In Figure the user has selected the value Centronics in the row Supported Interfaces with a single mouse click All cells in this row containing Centronics are highlighted The user can see at a glance that almost all printers have a Cen tronics interface A value selection also proposes a corresponding query of the form lt attribute gt lt value gt in the prompt line above the table For example in Figure 1 the query Supported Interfaces Centronics 1s proposed Pressing the Set button executes the query 1 removes all printers from the table that do not have a Centronics interface The complete row appears with a yellow background to indicate an active restriction e g the Technology row in Figure 2 As an abbreviation instead of first selecting a value and then pressing the Set button or return key the user can simply double click a value in the table The most common way of interacting with FOCUS 15 by double clicking on desired values until only a few products of interest remain After executing such a query all but the qualified records disappear f
11. ds 16 qualified Show All All Delete Refused Refused Ix Auto Reset Attributes 15 differ Show All All Delete Identical Identical Auto eee PtBreezeBigLot Yes OR SqHill Shady Yes Reatestate valve 0610141813 161217 holm Address 330 Leroi Neighborhood Point Breeze Shadyside 12 4 5 betel 3 6 5 9129109 105 8 Shadyside 4 1 11500 123 111 179 PtBreezeBigLot SaHill Shady Mean price per Bed Mean price per Neig o Ld 120 00 128149122132140132140 00 122128 00 120 00 For Help press F1 Figure 8 First scenario of Goldstein and Roth 6 similarly with FOCUS one can define two new user defined attributes The first query is performed and the result is stored as a new attribute Next the second query 15 formulated as a slight modification of the first and a second attribute is defined There is no need to define a third attribute because an OR query can be easily performed by selecting yes filled circle in the two rows for the new attributes With FOCUS fewer user actions are required because there is no need for commands that transfer information between the Aggregate Manipulator and the Dynamic Query component Using FOCUS Jennifer can actually solve her problem much more easily without defining new attributes and per forming a formal OR query see Figure 8 She
12. e click It is displayed on a yellow background Next the range of values smaller than 1000 was selected in the Price row and the Set button was pressed The prices are dis played on a blue background The Attribute Menu In order to release a subset restriction on an attribute again the user has to select lt AI gt from the attributes popup menu Figure 3a The menu appears when the mouse is pressed over the attribute name Selecting lt A gt 15 also equivalent to freeing a constant attribute by double clicking in the yellow area where the constant is displayed Price lt All gt Summary per vertical dpi All Values allowed Figure 3b A tear off menu Figure 3a An attribute menu Attributes can also be restricted to a constant value by selecting that value from the attribute s popup menu In this menu only attribute values appearing in the current restric tion are shown For example the result of restricting the selection to ink jets is that only the smaller prices are shown in the menu for the Price attribute We are following the principle of tight coupling formulated by Ahlberg and Shneiderman 1 The values are always sorted Therefore in large tables it is sometimes easier to find a value in the menu than in the table Subset selection and restriction can also be performed in a tear off menu Figure 3b which pops up when Subset is selected in an attributes menu This modeless dialog box c
13. ers with a high Postscript speed select Define New Attribute from the Table menu and type Fast Laser into a dialog box as the name for the new attribute The Fast Laser attribute will appear at the end of the table see Figure 5 This is a con venient way to save the result of a query Grouping When a table has been sorted according to an attribute identical values of that attribute will be jointly displayed in one larger field spanning several columns 1 e the records are grouped into classes with identical values See for example vertical dpi 1n Figure 5 Once a grouping of records into classes has been estab lished simply by moving the sort arrow the user can then summarize the values of a different attribute for each class For example the user can ask for the mean price for each Set cosys S 3500 A Model 7170 E Model 7610 yoceraGenico Texas In Hewlett Packard Co resolution class The new attribute Mean Price per vertical dpi is defined by selecting the entry Summary per vertical dpi from the popup menu of Price see Figure 3a A dialog appears where the user can choose between Minimum Maximum Sum Mean Count The example in Figure 5 shows the surprising result that the mean price of fast laser printers in the 300 dpi class is much higher than in the 600 dpi class Queries with grouping are quite difficult to formulate in normal database query languages In FOCUS a few mouse cl
14. icks are sufficient and grouping is visualized in a very natural way Of course in order to determine the minimum or maximum of a group it is not even necessary to define a new attribute Sorting first according to Price and then according to vertical dpi will show the extreme prices at the edges of each group as in Figure 5 Nested Queries Using user defined attributes it is also possible to formulate more deeply nested queries or disjunctions of conjunctions such as Find all color printers under 2000 and all laser printers under 1000 The user can define two attributes Cheap Color and Cheap Laser and combine them with OR as described above FOCUS Printers foc File Edit Window Table Help Records 9 qualified Show All Delete Refused Auto Delete Reset Attributes 42 differ Show All Delete Identical Auto Delete no selection aser Jet 4M Plus aser Jet 4 SIMX erox 4520 ps microlaser PowerPr aser Jet 4 Si Accel a Writer 8200 L L Technology L aser Postscript ppm in 7 20 8 10 8 24 Figure 5 Two new user defined attributes FOCUS Printers foc Raed ile Edit Window Table Help Records 2 qualified Show Delete Refused Auto Delete Ix Reset Attributes 23 differ Show All Delete Identical Auto Delete Ix supported Interfaces R5 232 C Exclude DeskWriter 320 DeskJet 540 DECc
15. l tools to manipulate and compare tabular data The filter mechanism of Excel 5 is similar to FOCUS but queries in Excel are less expressive and there is no mechanism to delete irrelevant attributes from the table Excel violates the principle of tight coupling because the value menus for each attribute also display values that have already been filtered out The PERPLEX spreadsheet of Spenke and Beilken 9 uses logic programming and can simulate dynamic database queries but has been designed as a tool for end user pro gramming and is therefore too complicated for simple product selection tasks THE INTERACTIVE TABLE In order to illustrate the FOCUS user interface we have chosen a comparison of 92 printers as an example The table appeared in the November 1994 issue of BYTE 7 and compares 51 different attributes features including vendor price resolution supported interfaces paper sizes emulations and benchmark results Other printer tests show that there are many further attributes of potential interest but obviously they would have consumed too much space The original table printed in BYTE occupies four pages using a very small font Table Inspection Figure 1 shows an overview of the printer table as dis played by FOCUS Each column represents one of the printers and each row contains an attribute row can span several lines if the attribute has multiple values e g Sup ported Interfaces The table is shown in o
16. n neighboring cells and prints them only once for a whole range of columns This measure considerably increases the readability of the table FOCUS follows principles formulated by Ahlberg and Shneiderman for Dynamic Query Filters 1 such as selec tion by pointing not typing immediate display of query results tight coupling output is input and progressive re finement of search parameters However FOCUS can han dle grouping and arbitrarily nested AND OR queries and the table format supports the comparison of alternatives Moreover FOCUS 15 a generic domain independent tool Recently Dynamic Queries has been generalized to be application independent but the restrictions for the possible queries remain Ahlberg and Wistrand 2 Chwelos and Mantei 3 Fishkin and Stone 4 extend Dynamic Queries by Magic Lens filters in order to increase their expressive ness Goldstein and Roth 6 combine dynamic queries with data aggregation in order to enable more complex queries How ever the formulation of complex queries is more difficult than in FOCUS because the Aggregate Manipulator and Dynamic Query components are not seamlessly integrated Instead they are presented to the user as two different work spaces and explicit commands are necessary to transfer in formation between these workspaces We will discuss later how the example queries of Goldstein and Roth can be expressed more easily with FOCUS Spreadsheets are classica
17. olorwriter 5 DeskJet320 DeskWriter 320 DeskJet 540 color DECcolorwriter 52 Hewlett Packard Co ewlett Packard CTexas InstrumentDigital Equipment DeskJet320 Ink Jet Auto Sensing H P PCL 5 RS 4270 Localtalk input Ink Jet Centronics Centronics R amp 232C Optional Technology Ink Jet Color yes Price in 379 414 439 HPGL yes Figure 6 Displaying eliminated records and attributes AUTOMATIC VERSUS MANUAL DELETION In the examples discussed so far the Auto Delete button for records was switched on so that records not matching are immediately removed from the table Sometimes however it may be interesting for the user to see which of the records are filtered out before they are deleted If Auto Delete mode is turned off these refused records are just grayed out see Figure 6 Later they can be explicitly deleted by pressing the Delete Refused button Figure 6 also contains some rows that have a gray back ground FOCUS automatically determines which attributes still have different values in the restricted table Attribute rows where all qualified records have the same value are grayed out Also section headers are grayed if the whole section is gray whether they are open or closed see Reso lution and Color Speed in Figure 2 for an example This helps the user focus on the relevant differences between the remaining alternatives Pressing Delete Identical
18. rface as easy to use as a Web browser KEYWORDS Dynamic Queries Tables Spreadsheets Focus Context Technique Interactive Data Exploration INTRODUCTION Frequently product information can be presented as an object attribute table Electronic product catalogs hotel guides and even attribute based selections of image mate rial naturally fit into this model With the growing interest in the WWW the demand for adequate presentation of such Appeared in UIST 96 Proceedings of the ACM Symposium on User Interface Software and Technology Seattle November 6 8 Copyright ACM 1996 1996 TuRRERRRSSRERERRESESRRERERRSRSEERESSRRRRRRSRRRRRRRRRRRRRRRRRRRRRRRRRRRSRRRRRRRRRRRRRRRRRRRRRRGRRESRSRRRRRRRRRRRRRRRRRRRRRRRRRERRRRRRRRSRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRGRRON material is increasing rapidly Currently there is no general powerful and easy to use mechanism available to visualize and explore such tables and to select and compare products Presenting the data as a static table becomes difficult if there are more than a dozen objects Scrolling the data 15 cumbersome objects can only be compared if near to each other and it 1s difficult to obtain an overview about what is available Database queries can be used if the table is large but que ries require some knowledge about what can be expected and in order to obtain good results the user must have a clear understanding as to which kinds of features are importan
19. rol as in a standard selection box Similar to the focus context tech nique described by Rao and Card 8 multiple selections result in multiple focal areas Two or more distant table entries can thus be compared Dragging the mouse through the column header continu ously changes the selection to the column under the mouse cursor This is as easy as browsing through a scrollable selection list The attributes are structured hierarchically like the lines of text in an outliner Subtrees can be opened and closed by clicking on the triangle shaped handles In this way the user can focus on the relevant attributes For example in Figure the attributes under Emulations are collapsed while those under nterfaces are expanded Sorting In Figure 1 the printers are alphabetically sorted by Tech nology indicated by the black arrow in the row header As a consequence equal attribute values are joined and have to be displayed only once By dragging the arrow or by clicking on its final position the table will be sorted by another attribute Because we are using a stable sorting algorithm fragmentation can be minimized throughout a series of sorts by different attributes In our example the table was sorted first by Price second by Color and finally by Technology Therefore the laser printers are still sorted by price FOCUS Printers foc File Edit Window Table Help Records 92 qualified show All Delete Refused Auto Delete K
20. rom the table In order to free a constant attribute again and to get back the deleted records the user double clicks on the single yellow value now shown for the whole row The user can also select the yellow value observe that lt attribute gt lt All gt is displayed in the prompt line and press the Set button FOCUS Printers foc File Edit Window Table Help Records 8 qualified Show All Delete Refused K Auto Delete K Overview Attributes 38 differ Show All Delete Identical Auto Delete X Keep Specifications Envelope yes vertical dpi 300 horizontal dpi 300 Color Quality Index LBP 430 Sharp JX 9400 LaserJet 4L OL 410e IN Technology Laser Price in 500 750 779 3 Figure 2 A restricted table row Technology originally appears in yellow row Price in blue Subsets of Attribute Values One can also select multiple attribute values within one row using shift and or control clicks as in a standard selection box The proposed query then reads lt attribute gt lt valuel gt lt value2 gt 1 e pressing the Set button will restrict the table to records where the value of attribute is either lt valuel gt or lt value2 gt Again a double click can be used as an abbreviation The row will be displayed in blue to indicate that a subset has been chosen In Figure 2 only laser printers under 1000 are displayed To achieve this the value Laser was fixed by a doubl
21. sly combines ideas from the Table Lens 8 Dynamic Queries 1 and the Aggregate Manipulator 6 into one consistent concept The principal features of FOCUS are e The whole table or any subset is completely visible on a single page in a highly compressed but still readable form e The table can be restricted to an interesting subset through dynamic queries specified within the table using simple mouse clicks e Complex queries can also easily be specified but are rarely necessary because of the flexible overview the sorting capability and the hierarchical attribute list FOCUS is freely available in the World Wide Web http www gmd de fit projects focus html and can used as a helper application of a WWW browser 10 REFERENCES 1 Ahlberg C and Shneiderman B Visual Information Seeking Tight Coupling of Dynamic Query Filters with Starfield Displays In Proceedings of the ACM SIGCHI Conference on Human Factors in Computing Systems Boston MA Apr 24 28 1994 pp 313 317 Ahlberg C and Wistrand E IVEE An Environment for Automatic Creation of Dynamic Queries Appli cations In Conference Companion of the ACM SIGCHI Conference on Human Factors in Computing Systems Denver CO May 7 11 1995 pp 15 16 Chwelos G and Mantei M Design Space of a Generic Interface for Filtering and Displaying Database Query Results In Adjunct Proceedings of the ACM SIGCHI Conference on Human Factors in Comp
22. t It is difficult to get an impression about what 15 not matched by a query 1 what could be found if the query were posed in another way We have developed FOCUS the Feature Oriented Catalog USer interface FOCUS enables the user to gain a flexible overview of an object attribute table through a combination of a focustcontext or fisheye technique and a hierarchical outliner for large attribute sets With a few mouse clicks into the table the user can formulate incremental database queries with immediate feedback after each step and thus restrict the table to the relevant subset of data Although there are only very few interaction techniques it is possible to express quite complex database queries FOCUS has been implemented in Visual and runs on PC Windows platforms It is freely available and can be used as a helper application in a Web browser RELATED WORK The Table Lens interface of Rao and Card 8 uses a focus context technique to display large tables in a com pressed form but does not support queries Our table inter face uses a similar technique for the objects columns The width of a column is increased when selected focused Multiple selected columns correspond to multiple focal areas However in order to focus on the relevant attributes rows we use a hierarchical outliner FOCUS combines the focus context technique with incremental database queries and as extension trecognizes identical values i
23. tabases In these situations FOCUS supports the complex selection of suitable products that are described by a multitude of attributes with no clearly defined evaluation procedure In addition to printers and other computer equipment typi cal product categories include cars stereo equipment washing machines etc Mail order catalogs include many different products with some common attributes such as price order number and delivery time but also contains products with some product specific attributes that only apply to certain product classes such as color size and sex for articles of clothing FOCUS is well suited for this kind of large sparse tables because attributes can be grouped in hierarchical subsec tions and irrelevant attributes disappear automatically from the table after selecting a special product class A similar situation exists for a directory of camping sites or a hotel database Again this kind of data can be described by a number of different attributes which have vastly dif ferent importance for different customers Pharmaceutical products databases are also a domain where the comparison of different products is very important Often there is no optimal drug so it is necessary to compare different side effects CONCLUSION We have developed a simple powerful and efficient inter active viewer for tabular data which is ideally suited for data exploration and interactive query formulation FOCUS seamles
24. thermal printers and the value 600 has been selected in both the Resolution vertical and Resolution horizontal rows Pressing the Set button removes all records which do not have a resolution of 600 dpi in at least one of the dimensions FOCUS Printers foc File Edit Window Table Help ped Records 9 qualified Show All Delete Refused Reset Attributes 36 differ Show All Delete Identical Set vertical dpi 600 OR horizontal dpi 600 PrimeraPro Color Spectra Star GT Spectra Star G Tx Genicom 7025 Phaser220e e a E 5 o m d argo ElectrorGeneral ParariGenico Technology Thermal Figure 4 A disjunction Conjunctions Because subsequent restrictions specified by the user are implicitly linked by the AND Operator the general form of a FOCUS query is a conjunction of disjunctions attribute in value 4 7 value457 OR attributeg in lt 2 valueg OR AND lt attributec gt in lt valuec gt lt valuec gt gt OR lt attributep gt in lt valuep gt lt valuep gt gt OR AND User Defined Attributes Whenever the user has restricted the table to an interesting subset of records he can introduce a new user defined attribute that is true for exactly the records in this subset For example the user can choose all laser print
25. uting Systems Boston MA Apr 24 28 1994 pp 175 176 Fishkin K and Stone M Enhanced Dynamic Queries via Movable Filters In Proceedings of the ACM SIGCHI Conference on Human Factors in Computing Systems Denver CO May 7 11 1995 pp 415 420 Excel User Manual Version 5 0 Microsoft 1993 Goldstein J and Roth S F Using Aggregation and Dynamic Queries for Exploring Large Data Sets In Proceedings of the ACM SIGCHI Conference on Human Factors in Computing Systems Boston MA Apr 24 28 1994 pp 23 29 Kane J and McDonough J 92 Printers go to Battle In BYTE November 1994 pp 262 283 Rao and Card S K The Table Lens Merging Graphical and Symbolic Representations in an Interac tive FocustContext Visualization for Tabular Informa tion In Proceedings of the ACM SIGCHI Conference on Human Factors in Computing Systems Boston MA Apr 24 28 1994 pp 318 322 Spenke M and Beilken C PERPLEX A Spreadsheet Interface for Logic Programming In Proceedings of the ACM SIGCHI Conference on Human Factors in Com puting Systems Austin TX Apr 30 May 4 1989 pp 75 80
26. verview mode where FOCUS automatically adjusts the column width so that all printers are visible in the given window One of the column headers has been selected by the user as the focal column for closer inspection Although many of the values 1n other columns are not readable the user can at least estimate the number of printers for each of the five technologies One can also observe that all laser and dot matrix printers are black and white all thermal printers are color and ink jets can be either color or black and white Displaying and Inspecting Compact Tables Within the Price row as in the other numerical rows values are graphically indicated by the height of a bar the corresponding numbers are only displayed if there 15 suf ficient space available This allows the user to make global observations even in highly compressed tables For exam ple one can see that ink jet 15 the cheapest technology The column headers are displayed in vertical text if the column width is smaller than the height of the column headers which can be adjusted by the user Simple yes no attributes are displayed as filled and empty circles to save space and to improve legibility In overview mode adjacent and identical entries are com bined into a single value spanning several columns both to increase readability and to highlight similarities Multiple columns can be selected and unselected by click ing in the column header using shift and cont
Download Pdf Manuals
Related Search