Hacker News Reader

Comment by Transfinity | Hacker News Reader

As a programmer, I love this app. As a musician, I think it would be good for novices who are still in the "note identification" phase, as others have mentioned.

However, I'm not sure how useful it would be for an advanced student. Much of becoming truly proficient at sight reading is context-dependent chunking. In other words, given the key, style, tempo, what was just played, location in the form, etc. what patterns are likely to appear? Training yourself to identify and predict longer and longer sequences of notes as single "chunks" is key to professional level sight reading. An app that generates random notes can't simulate that stylistic context.

View on HN

It would be very simple to feed in some midi from real pieces of varying levels to replace the random generation.

It wouldn't even need real pieces, just real transition probabilities.

So long as those probabilities are aware of context (style, form of this piece, location in the form, etc.). I've played around with Markov models a little, but I don't think they would quite cut it. This problem is probably slightly less difficult than generating prose from scratch, but it's a similar type of difficulty.

Getting solid curated input data would also be tricky, unless there's some big and well-annotated library of midi recordings sitting around somewhere.

For building a good app today, your best bet is almost certainly to use existing sheet music, pieces and etudes in the given style.

Does touch typing practice require actual words or just random letters? I think if you can be proficient at random notes you'll have an even easier time with actual music.

This is false however reasonable sounding it is.

If you spend time with professionals sight reading, you come to appreciate that the cognitive overhead of reading individual notes caps out quite soon and most high level sight reading involves chunk identification and anticipation.

I've seen Liszt sight read at normal tempo for example and that is going to mostly be structural reading than note by note reading.

Where random note reading could be useful is for more "modern" atonal stuff...

It's absolutely true you need to be able to read the "words" or even phrases in the music, not just the letters. You can only do that with significant look-ahead.

But you also need to know how which fingers to use.

It's very much not just a case of playing the right note at the right time. It's about playing note sequences using fingerings that lead smoothly to later note sequences - all at speed.

That means including thumb-under and other moves that aren't at all intuitive and are totally non-obvious to beginners.

Apps like these don't just fail to teach that, they don't recognise it as a thing that needs to be taught.

With the advent of neural nets that can mimic composers' styles, it would probably be possible to generate real-sounding music on the fly.

Reply on news.ycombinator.com