Our assignment in class for this week is to do a search for a question in both a regular search engine (like Google) and a natural language engine (like START), then compare the two and discuss the advantages and disadvantages of natural language processing.
I really, really want to use the question "How many licks does it take to get to the Tootsie Roll center of a Tootsie Pop?" . . . but I'd like to find something that an NLP engine can actually answer, thus proving that NLP is awesome, which is the topic of my final paper.
Showing posts with label search engines. Show all posts
Showing posts with label search engines. Show all posts
Tuesday, April 28, 2009
Friday, April 24, 2009
Music Information Retrieval Systems
This week, I wish I had more time to devote to what we're studying in class. It's one of the most interesting things we've talked about so far: cross-linguistic and multimedia information retrieval systems.
Many of you are probably familiar with cross-linguistic IR systems--just go to Google and play with its ability to translate foreign web pages into English, or head over to Babelfish and experiment with its machine translation.
Multimedia IR systems are another matter, though. How can we search for things that aren't expressed in text? Image IR systems have ways of analyzing color and shape in pictures. And music IR systems can analyze things like pitch, melody, etc.
Part of our assignment this week was to choose a multimedia IR system and try a search in it, then report what we did and what we found.
These are the three music IR systems we were given as examples. Play around with them; they're fun! The technology allows users to search by tune rather than by song title or artist.
Tunespotting lets you search by arranging the notes yourself, playing the tune on your computer keyboard (which I found difficult), and also by the regular method of text relating to the tune.
This one is based more around pop music: Midomi. It boasts of being the "ultimate music search"--I wouldn't exactly agree since most of what it has is popular, but if you're looking for a tune like that, I've found it to be spot on so far.
Musipedia is very cool. Probably my favorite of these three, it allows searching by singing the tune (as in Midomi), playing it on a virtual keyboard, contour, or rhythm.
Enjoy!
Many of you are probably familiar with cross-linguistic IR systems--just go to Google and play with its ability to translate foreign web pages into English, or head over to Babelfish and experiment with its machine translation.
Multimedia IR systems are another matter, though. How can we search for things that aren't expressed in text? Image IR systems have ways of analyzing color and shape in pictures. And music IR systems can analyze things like pitch, melody, etc.
Part of our assignment this week was to choose a multimedia IR system and try a search in it, then report what we did and what we found.
These are the three music IR systems we were given as examples. Play around with them; they're fun! The technology allows users to search by tune rather than by song title or artist.
Tunespotting lets you search by arranging the notes yourself, playing the tune on your computer keyboard (which I found difficult), and also by the regular method of text relating to the tune.
This one is based more around pop music: Midomi. It boasts of being the "ultimate music search"--I wouldn't exactly agree since most of what it has is popular, but if you're looking for a tune like that, I've found it to be spot on so far.
Musipedia is very cool. Probably my favorite of these three, it allows searching by singing the tune (as in Midomi), playing it on a virtual keyboard, contour, or rhythm.
Enjoy!
Subscribe to:
Posts (Atom)
