Generating(very bad) lyrics using a markov chain model written from scratch in Java. Thanks to https://github.com/SwagLyrics for the training data.
Most of the high level magic happens in src/Main.java
. Run Main.main()
to generate lyrics. Unfortunately, you need to modify source code to tinker around stuff. For example, change the constant WORDS_LIMIT
in src/Main.java
to set the number of words to generate.
can i think i know i know you?" ain't got a hell new religion when he make that we'd be so once again i'm sad bitch i was the jones right out there is too long and you understand me? i feel worse yeah create without conflict just can't change no masters here burn the glam boy i can feel fine now and i don't leave me" you fall apart through no game has fallen for the flu fuck i could take me make me and goodnight these heads with a new chick i know we wanted to see a request mi have got a mutant feel the atlantic i'm all night was in fuck but i tell by your cup girl i don't ball then double it ain't the time even look like badass can be complete not your mind but i choose to everybody fucks but that lost it can see it sucks maybe we really wanna call my friends and seek would with him tonight stay with me rich i go stupid obnoxious or how long pressure pushing for next your memory will you get you you to blame so wide oh
If you are wondering how I generated these awesome lyrics, here is what I did:
- Use
Cleaner.java
to clean the data(lyrics.txt) to remove unnecessary characters and non English words from the training data, to give cleaned.txt. - Train the model (very inefficiently, please don't use this implementation for something serious) with the cleaned data.
- Choose a random word as the starting and keep feeding the words we get back to the model to keep generating lyrics for as long as we want to.
- P.S.: words.txt is just https://github.com/dwyl/english-words/blob/master/words.txt, used for filtering non english words from the training data.