daniel bigham.ca

Exercise 20: Deriving properties of relationships
October 2, 2008

Summary

In addition to deriving property values and the existence of relationships, there are times we want to derive properties of relationships. For example:

"I have two sisters"
"I don't have a brother"
"How many siblings do I have?"

rule:
  $1 is_a person
  $1 has(count:$2) sister
  $1 has(count:$3) brother
  $count = # $2 + $3
  -> $1 has(count:$count) sibling

Web UI

Click here

Solution

Click here

Technology as art
September 26, 2008

When Meredith and I were in Madrid this summer, there was an exhibit that used technology and the Internet to create a dynamic display that linked keywords on Internet websites against one of a number of topics, which were color coded, and then marked on a world map where that website was located. It was more complicated than that, but what I found interesting was that technology was being used to create art, and it was in an art gallery!

Here's my idea:

Create a visually interesting object such as a polished half sphere which would contain invisible cameras, microphones, and speakers. It would use the cameras to visually track people walking around the room, and when someone came close to the object, it would say, "Hi, what's your name?" in a pleasant voice indistinguishable from a person's. The person would then say something like "Daniel", "Daniel Bigham", "My name is Daniel", etc. and using voice recognition the object would be able to parse what was being said. It would then say "Hi Daniel, nice to meet you".

If the person then walked away and came back, the computer, having tracked the person walking away and coming back, would say, "Hi Daniel". If the person left the room and came back, an attempt would be made to recognize the person based on a visual fingerprint.

To be fun and impressive:

1.	The voice synthesis, and specifically, name synthesis, would need to be nearly perfect, and sound very pleasant, with good intonation.

2.	The system would need to be able to construct names that it hadn't previously heard by analyzing the phonemes spoken, and then to be able to re-speak those phonemes.

3.	The visual tracking algorithms would have to be really good.

I think this would be a neat "art" piece, albeit a bit noisy.

Best Text-to-Speech yet: IVO
September 25, 2008

Have a listen to "What is the weather forecast for tomorrow"? Very impressive.

I've tried some other phrases and the results haven't been as impressive as this one example, but overall I like what I hear.

older >>