Glad you like it!

I think I haven't really completely exhausted the possibilities of the algorithm yet. Consider, for example, these two queries for "Eine kleine Nachtmusik":
http://www.melodyhound.com/search.0.html?sourceid=melody-url&tx_mpsearch_pi1%5bsubmit_button%5d=Search&tx_mpsearch_pi1%5bpc%5d=hum%200.651-0-1/1.332-0-1/1.533-0-1/2.143-0-1/2.354-0-1/2.554-0-1/2.764-0-1/2.955-0-1/3.175-0-1/4.026-0-1/4.707-0-1/4.907-0-1&categories=CPHA&filtertext=&coll=r
and
http://www.melodyhound.com/search.0.html?sourceid=melody-url&tx_mpsearch_pi1%5bsubmit_button%5d=Search&tx_mpsearch_pi1%5bpc%5d=hum%200.671-0-1/1.362-0-1/1.552-0-1/2.163-0-1/2.383-0-1/2.573-0-1/2.774-0-1/2.954-0-1/3.164-0-1/4.026-0-1/4.717-0-1/4.917-0-1/5.538-0-1/5.748-0-1/5.938-0-1/6.139-0-1/6.319-0-1/6.549-0-1&categories=CPHA&filtertext=&coll=r

With the algorithm in its current shape, the second, longer one shows the Nachtmusik already on page one instead of page two like the first one, as one should expect if you add more notes and therefore make it more clear what you want. But I still see some matches ranked higher that should not be there, so I need to do some more debugging and see what goes wrong there.
Of course, even more tweaking still won't change the fact that using pitch and rhythm at the same time means that far fewer notes are necessary for getting far better results, e. g.:
http://www.melodyhound.com/search.0.html?sourceid=melody-url&tx_mpsearch_pi1%5bsubmit_button%5d=Search&tx_mpsearch_pi1%5bpc%5d=lily+g'4.+d'8+g'4.+d'4+g'8+d'8+g'8+b'8&categories=CPHA&filtertext=&coll=m

So there'll probably always be a tradeoff between the minimum skills needed for formulating the query and the result quality you can get.

at first i was going to complain that it was only accepting one note. then i realized that Firefox was set to search automatically. so here i was searching for vvvvvvvvvvv in the page. ha ha.

First of all, great work on the new interface method. very nicely done. I'm sure the "safe for work"-ness will also be appreciated. I whistle and sing all the time at work, so i could get away with the audio interface just fine. ;-)

the algorithm seems to be interpreting my rythms correctly, but there are so many songs with the same rythm. for some things, like Joy to the World or Academic Overture, the "right" answer is pretty close to the top of the list. More mundane rythms, like O Fortuna or Eine Kleine Nacthmusik, can be very far down on the list. Maybe there is a way to prioritize them based on popularity. ...but that would open up a whole new can of worms. Plus, if it's already popular and easy to find, why would people be searching for it here? hmm....

I tried adding notes, but it didn't seem to make much difference. maybe that's because rythms tend to get repititious after the first phrase.

Keep up the good work!

Today I added a new search method: query by tapping.

It is work-safe (no embarrassing singing), and it requires only very little skills (no familiarity with notes, keyboards, whistling, or singing needed). Also, it does not ask much of the browser - just Javascript (neither Flash nor Java).

Of course, it is even rougher than the melodic contour search. With a good rhythmic query, one can expect the desired match on the first page (or maybe the second one), while with a good melodic query, the desired match should be ranked much higher.

Here it is: http://www.musipedia.org/query_by_tapping.0.html

I would be grateful for feedback, especially if something does not work.

BTW: One can only search the Musipedia collection by tapping (not the Web because there are just too damn many similar rhythms embedded in otherwise wildly differing MIDI files - I think searching those only based on rhythm is not promising, especially since it is not known which voice one should search). However, once something has been located in Musipedia, one can still search the Web for matching MIDI files with just one additional mouse click, this time based on melody and rhythm (by clicking on the corresponding button in the search result list).