InfoVis 2007 Contest
Interactive Exploration of the Movie DB on a Semantical Level

Contest webpage:

Authors and Affiliations:

• Thorsten Liebig, Ulm University, thorsten.liebig@uni-ulm.de
• Olaf Noppens, Ulm University, olaf.noppens@uni-ulm.de
• Timo Weithöner, Ulm University, timo.weithoener@uni-ulm.de

Tool(s):

Our approach combines logical reasoning over an enriched ontological model of the IMDb with an interactive, semantically guided investigation of the manifold interconnected movie network. We utilize an user-driven exploration strategy, making use of animated expansion steps, clustering techniques, and different levels of detail or abstraction views. This is done with help of a specially enhanced version of our ontology browsing and authoring tool OntoTrack. A description of OntoTrack can be found in our Journal of Web Semantics article. OntoTrack is a Java application using the Java 2D library Piccolo. During exploration, any expansion set may be narrowed by drag-n-drop of restricting criteria. Instant reasoning feedback is employed for automatic instance classification or to check for conflicting filtering criteria. All reasoning tasks are done instantly by our high performance relational reasoner U2R2 (Ulm University Relational Reasoner).

TASK 1: Do Best Actress Oscar winners tend to come from certain genres?

• Process:
  To answer this task, we browse the concept hierarchy up to the concept Actress within the schema view and drop it on the Movie explorer pane in order to start our exploration with the set of all female actors. By clicking on the circle background, we open the preview clubs of connected relationships and choose the inverse leadingActress relationship (symbolized by an arrow from right to left at the beginning of the label) for further expansion. We then get a club with all movies for which one of the actresses has won a Best Actress Oscar. Finally, we expand the movie club wrt. to the hasGenre property. For this task as well as all the other tasks, no further data analysis is needed because of our semantical classification approach.

• Image 1.1 :
  
• Insight:
  The expansion of the leadingActress relationship in inverse direction gives us a club showing all movies for which a Best Actress Oscar has been awarded. The expansion from that club with respect to the hasGenre relationship gives us a club depicting all genres for all those movies. The numbers on the genre individuals as well as their diameter show how many of those movies do refer to that particular genre. As a result, we can see that 6 movies refer to the drama genre.
• Caption for exhibit:
  Easy 3-step expansion from actresses to genres to get the proportion of genres to Oscar movies.

TASK 2: What are the most "bankable" actors, actresses, directors or cinematographers?

• Process:
  We start in the schema pane and drop the concept Movie onto the Movie explorer pane. A resulting slice represents all 20204 movies. Next, we expand the movie slice consecutively with respect to the cast, directedBy, hasCinematographer relationship to get a club representing all performers who are in the cast of at least one of those movies, a club representing directors of those movies and a club representing all cinematographers. Finally we select the concept Actress from the schema pane to restrict the cast club to actresses. Then we expand the cast property once more and restrict it to actors in the same way.
• Image 2.1 :
  
• Insight:
  When activating the slider for the box office values, the persons within the clubs for actors, actresses, directors and cinematographers will be restricted to those which belong to movies whose box office value falls into the interval. For instance, one can see that after adjusting the boxoffice filter to include movies between 312592768 and 436327392 (which is at the hight end of values) the movies are restricted to the 13 most bankable ones. Hovering with the mouse pointer over the club of directedBy fillers one get all directors of those movies, e.g. Chris Columbus, Martin Scrorsese or Scott Ridley. Hovering over the other clubs (the movie club included) will show the corresponding member names. Note that the name of the movies can be get via mouse click as also described in the second subtask of our task 6.
• Caption for exhibit:
  After expansion of the cast, directedBy and hasCinematographer property one can easily change the box office sliders to restrict the persons to those which belong to a movie within the slider's interval.

TASK 3: Are there directors which directed an Oscar movie and who played themselves in a movie which received an Oscar award in at least one category?

• Process:
  To answer this task we begin on the schema pane by selecting the concept OscarMovie (for detailed definition please refer to the chapter "Concept Detail Views" in our video) and by dropping it onto the Movie explorer pane. A new slice represents all Oscar movies. We then expand this slice with respect to the directsBy property to get all those persons who directs at least one Oscar movie. As we would like to know whether these persons also appeared in an Oscar movie, we expand the club representing the directors with respect to the appearsIn property. To restrict these movies only to Oscar movies we drop the concept OscarMovie from the schema pane upon the club. Now, while hovering over the largest movie individual the entry A Beautiful Mind is magnified on the left hand side. After clicking on the individual in the cub a context menu shows which directors appears in that movie.
• Image 3.1 :
  
• Insight:
  As the exploration has started with all Oscar movies the slice on the left hand side represents all Oscar movies and the middle club contains all directors of at least one of these Oscar movies: each director is represented by a blue circle. The numbers on that circles as well as the diameters show how many movies the director has directed. Note that the names of the directors are either visible as mouse-over hints or in the list on the left hand side when hovering over the club. The last club labeled with the appearsIn property and restricted to Oscar Movies show all Oscar movies (7 Oscar movies) in which at least one of the directors of the preceding club appears in. We can see that in A Beautiful Mind two directors of an Oscar movie also appear in, namely Ed Harris and Ron Howard. Note that if we expanded the movie A Beautiful Mind in the first slice wrt. to the directs property we would see that Ron Howard is also the director of that movie.

TASK 4: Is there a tendency of genres with respect to Oscar Movies between 2000 and 2003 in comparison with 2004-2007?

• Process:
  To answer this task we drag the concept OscarMovie from the schema view to the Movie explorer and drop it there on the background. Note that an OscarMovie is defines as a movie which has won at least one Oscar award in at least one category (e.g. best movie, best actor, etc.). Next, we click on the slice representation of Oscar movies and get a preview of all relationships of at least one of the movies. We choose the preview club labeled hasGenre to further explore the genres of all Oscar movies. As a result we get a club representing all genres to which all Oscar movies belong to. After activating the bi-directional sliders as shown in our video in chapter "Interval slider", we can add time constraints. Even if the usage of these sliders is a continuous task with instantly reflecting the time restriction we will describe it here as consecutive images in subtasks.
• Image 4.1 :
  
• Insight:
  We see that from all Oscar movies 29 come from the drama genre. As our approach is highly interactive, one needs hovering with the mouse pointer over the other genres to see that, for instance, 9 comes from the crime genre, and 7 from the thriller genre. In the following subtasks we would like to know how does this change over time.

Subtask of TASK 4: Introduce time 2000-2003

• Process:
  
After activating the sliders we choose the filter slider for the release year and adjust it to the interval between 2000 and 2003.
• Image 4.1.1 :
  
• Insight:
  
Movies which does not lay in the interval are grayed out. Moreover, the numbers in the corresponding genres individuals decrease as well. One can see that most movies fall into the drama genre (18 movies), followed by the crime genre (6) and by adventure (5).

Subtask of TASK 4: Introduce time 2004-2007

• Process:
  
After activating the sliders we choose the filter for the release year and adjust it to the interval between 2004 and 2007.
• Image 4.1.2 :
  
• Insight:
  
One can see that again the most Oscar winning movies still fall into the drama genre (15), followed by romance (4), crime(4), and thriller. Note that a movie typically belongs to more than one category. Summarizing one can see that most of all Oscar winning movies are drama movies and that crime movies have also a good chance to win. Moreover, between 2000 and 2003 adventure movies have also won whereas in the years 2004-2007 it seems to be a trend that romance movies are become Oscar winners.

TASK 5: Are there any actresses (who?) which appear in more than one Oscar movies?

• Process:
  Within the schema view we select the OscarMovie concept and drag-n-dop it to the movie explorer to start a new exploration. Immediately we see a slice representing all Oscar movies. As there are only few movies each one is represented as a blue circle (as mentioned in our video in chapter "Scrollable Instance List" one can get a list of all movies with detailed information). Next we click on the slice to get a preview of all properties which are related with at least one of the movies. Here we select the preview club labeled with cast to expand the cast property. The newly visible club then contains all actors as well as actresses from at least one of the Oscar movies. As we are only interested in actresses we select the Actress concept in the schema view and drop it onto the cast club to restrict its performed to actresses.
• Image 5.1 :
  
• Insight:
  The dark green club labeled with the cast property is restricted to actresses, symbolized by :Actress. It shows all actresses which are in the cast of at least one of all Oscar movies. One can see that there are actresses which appears in more than one Oscar movies because the number and diameter of the representation of actresses within the cast club show how many Oscar movies has the particular actress in its cast. For instance, Cate Blanchett appears in three Oscar movies. Note that our approach is highly interactive and therefore one need hover the mouse over the other circles to get their names.

Subtask of TASK 5: Has Cate Blanchet won an Oscar award or does she only appear in Oscar movies?

• Process:
  For answering this question (following the previous one) one has to mouse-click on the circle representing Cate Blanchet to get a preview of all available properties for Cate Blanchet. One can see that there is a supportingActress property labeled also with the number 1 which means that she won 1 Oscar awards as supporting actress. Note that this property would not be available if she had not won any Oscar award. Next, one expand the club by clicking on the preview.
• Image 5.2 :
  
• Insight:
  We can see that Cate Blanchet has won 1 Oscar awards as supporting actress. By hoovering over the Oscar individuals users can see that she won the award for the movie The Aviator

TASK 6: Who received the best cinematography Oscars between 2002 and 2005?

• Process:
  First drop the concept OscarMovie onto the Movie explorer pane and expand the bestCinematography property of all these Oscar movies. After activating the sliders we adjust the slider for the release year to the interval ranging from 2002 to 2005.
• Image 6.1:
  
• Insight:
  One can see that there are 6 cinematographer to have won an Oscar award in cinematography, 2 of them fall outside the interval from 2002 to 2005 and therefore their circle representation on the club labeled bestCinematography are grayed out and are not displayed in the (optional) list on the left hand side. As a result, Dion Beebe, Russel Boys, Conrad L. Hall, and Robert Richardson are the cinemtaographer who won an Oscar award between 2002 and 2005.

TASK 7: What are the romance or comedy movies in 2006?

• Process:
  In this task we are interested in romance or comedy movies, i.e. the union of movies which belongs either to the romance or comedy genre. We start within the schema view pane and graphically define a new concept "FunnyOrFloweryMovie" as the disjunction of a movie which has a romance or comedy genre property. Next, we drag it onto the Movie explorer pane and activate the sliders to adjust the release year property to 2005. Note that this is only one example of defining a complex concept, the same way one can define even more complex concepts in a standard DL way. The reasoning component will be informed about the new concept immediately and classify the movies instantly.
• Image 7.1:
  
• Insight:
  The FunnyOrFloweryMovie is defined as a movie related to either to the genre comedy or romance as can be seen within the schema view. Within the schema view one can easily define such complex concept expressions. The Movie explorer shows that, after adjusting the release year slider to 2005 and the rating to a value between 8 and 10, there are about 2050 comedy or romance movies (also counting those which are not rated, they can optionally be grayed out).

TASK 8: Who played with Brad Pitt most frequently?

• Process:
  We first drop the concept Actor onto the Movie explorer and activate the search functionality to search for Pitt, Brad. In the individual list, we select the corresponding individual and drop it onto the Movie explorer to start a new expansion. Next, we expand the appearsIn property and get a club with all movies in which Brad Pitt appears. Finally, we expand this club wrt. to the cast property to get a club with people who appears in at least one of the movies.
• Image 8.1:
  
• Insight:
  The number of the actors/actresses as well as their diameter refer to the number of movies in which they appear (i.e. the number of movies to which cast they belong to). As one can see, there is an movie participant labeled with 12 and an movie participant labeled with 3. Note however that the first one is Brad Pitt itself as one can easily see when hovering the mouse pointer over that individual. The latter one is Catherine Zeta-Jones. Thus, she played with Brad Pitt most frequently (3 times).

Subtask of TASK 8: Which are the movies in which the/these persons has appeared in along with Brad Pitt?

• Process:
  In addition to answer this question one only needs to open the context menu (right mouse click) of the individual, in our case Catherine Zeta-Jones.
• Image 8.2:
  
• Insight:
  One can see that Catherine Zeta-Jones and Brad Pitt appear together in Ocean's Twelve, in Sinbad: Legend of the Seven Seas and in Boffo! Tinseltown's Bomb and Blockbusters.

Subtask of TASK 8: Do Brad Pitt and Catherine Zeta-Jones tend to appears in movies of the same genres?

• Process:
  Now, as we know that Ms. Zeta-Jones and Mr. Pitt appears in three movies together, one might ask whether both tend to appear in movies belonging to the same genres. First, we expand the first club representing all movies in which Brad Pitt appears with respect to the hasGenre property. Next we expand the blue circle representing Catherine Zeta-Jones wrt. to the appearsIn property to get all movies in which she appears. Then, we expand this upcoming club wrt. to the hasGenre property.
• Image 8.3:
  
• Insight:
  By hovering with the mouse-pointer over the club representing the genres of the movies Ms. Zeta-Jones appear in one can see that the largest one belongs to the comedy genre. It is labeled with number 6 which means that 6 of the movies belong to that genre. Moreover, the same genre is instantly highlighted within the club representing the genres of the movies in which Mr. Pitt appear and one can see that comedy movies are the second ones. Moreover, one can exploit that Mr. Pitt appears most of all in movies belonging to the thriller category (6), to the comedy category (5) and action category while Ms. Zeta-Jones appears most of all in movies belonging to the comedy category (6), romance (4), crime (3) and drama (3). In that sense, they mainly plays in different genres.