JERRY: I don't know. Uh, it must be love.
At Monks
========
PATRICE: What did I do?
GEORGE: Nothing. It's not you. It's me. I have a 
  fear of commitment. I don't know how to love.
PATRICE: You hate my earrings, don't you?

00:04:52,691 --> 00:04:54,716
I don't know. It must be love.

00:05:04,136 --> 00:05:06,468
-What did I do?
-Nothing. It isn't you.

00:05:06,639 --> 00:05:10,598
It's me. I have a fear of commitment.

00:05:10,776 --> 00:05:13,677
-I don't know how to love.
-You hate my earrings, don't you?

This page contains the demonstration material for our 2010 ACM Multimedia Demo and Exhibit entry, Joke-O-Mat HD.

We assume the following use case: When a sitcom is watched a second, third, or later time, a user might want to show a very funny scene to a friend, point out and post the sharpest punchline to his or her Facebook network, or even create a home-made YouTube video composed of the most hilarious moments of his or her favorite actor. In order to do this quickly, we present the user with the basic narrative elements of a sitcom such as the scenes, punchlines, and individual dialog segments on top of a standard video player interface. A per-actor filter helps to search only for elements that contain a certain protagonist. The user is now able to selectively skip certain parts and to directly navigate into elements he or she remembers from the past. For more information on the concept and the underlying technology, please see references below. The Joke-O-Mat HD system (as compared to the non-HD system) has been augmented with keyword filtering using a speech and speaker-recogniton-based process to generate the segmentations from human derived (HD) "found" data, i.e. a combination of fan-sourced manual transcripts and closed captioning.