Being distressed at the prospect of singing a Hillsong song, I collected all the lyrics of 100 or so Hillsong songs, and put them through a Markov chain generator, to generate new songs in the style of the originals.\u00a0 The parameters used are:<\/p>\n
The results are about as coherent and approximately as Biblically sound as bulk of the songs that Hillsong writes.\u00a0 Have I mentioned that I find the lyrics repetitive and incoherent?<\/p>\n
Here are three selections. Notice the odd spelling mistake – that’s the Markov thing randomly choosing to make an error.\u00a0 I made up titles for fun:<\/p>\n
There’s no end to Your embrace
\nLight of the world
\nLet the poor say, “I am rich”
\nLet the poke the Earth, let us sing
\nGod, our hope is Yahweh, Yahweh
\nForever You will never fail
\nYour name shout in all the earth
\nWill fade away
\nStill my soul
\nMy heart within the darkness
\nGlory pierced the night sky
\nYou give me breath and all I found was You
\nMy God, I’ll only ever give You my heart
\nDistant shores and the end
\nThe promise out
\nThis was not a new thing
\nUntil itSaviour, He can move the mountains with a whisper
\nAnd You calm my soul
\nOh, now save
\nOur God is great and mighty
\nGod in three persons, blessed Trinity<\/p><\/blockquote>\nIn the highest praise at the sound of faith<\/h2>\n
To the Lamb that was slain
\nHosanna, Hosanna
\nHosanna in the highest praise
\nWhat can separate us
\nNothing can separate me now
\nYou taught a way, You made a way
\nWhen You call, I won’t hide it, I won’t refuse
\nEach new day again I’ll choose
\nThere is none like You
\nThere is no one like You, God
\nMountains bow down and the seas will roar
\nAt the sound of faith
\nand the words to express the way of the Lord
\nEternity’s King
\nIs coming again
\nThough all of the universe is at Your feet
\nHide me now in the shadow of Your word, Your name
\n“I live to know You
\nI live to know You
\nLet go and throw my life has changed
\nWhen You took a crown of thorns
\nAnd Your blood was spilled
\nFor my ransom
\nEverything I have, I give You praise in all of the earth rejoice
\nLet all the heavens
\nFor You are holy, You are holy, You alone
\nAwake my soul the reason why I sing
\nAll around the words to express
\nthere’s nothing like
\nYour love You always<\/p><\/blockquote>\nNothing that’s true<\/h2>\n
For all Your sons and daughters
\nWho are walking in the east, beyond the heavens
\nAnd Your love transforms my soul, He is the Lord with all my hope is in You
\nJesus Christ the Savior is born
\nHe shall become a wonderful counselor, everlasting Lord
\nLate in time, beholding Your beauty
\nAnd in the palm of Your hands I belong, I’m a living stone
\nIn this house I will grow
\nThere is nothing that’s true<\/p><\/blockquote>\nThe script:<\/p>\n
#! \/usr\/bin\/python\r\n\r\nfrom __future__ import division\r\nimport random\r\n\r\n\r\nclass Categorical(object):\r\n\r\n def __init__(self, support, prior):\r\n self.counts = {x: prior for x in support}\r\n self.total = sum(self.counts.itervalues())\r\n\r\n def observe(self, event, count=1):\r\n self.counts[event] += count\r\n self.total += count\r\n\r\n def sample(self, dice=random):\r\n sample = dice.uniform(0, self.total)\r\n for event, count in self.counts.iteritems():\r\n if sample <= count: return event sample -= count def __getitem__(self, event): return self.counts[event] \/ self.total class MarkovModel(object): def __init__(self, support, order, prior, boundary_symbol=None): self.support = set(support) self.support.add(boundary_symbol) self.order = order self.prior = prior self.boundary = boundary_symbol self.prefix = [self.boundary] * self.order self.postfix = [self.boundary] self.counts = {} def _categorical(self, context): if context not in self.counts: self.counts[context] = Categorical(self.support, self.prior) return self.counts[context] def _backoff(self, context): context = tuple(context) if len(context) > self.order:\r\n context = context[-self.order:]\r\n elif len(context) < self.order: context = (self.boundary,) * (self.order - len(context)) + context while context not in self.counts and len(context) > 0:\r\n context = context[1:]\r\n return context\r\n\r\n def observe(self, sequence, count=1):\r\n sequence = self.prefix + list(sequence) + self.postfix\r\n for i in range(self.order, len(sequence)):\r\n context = tuple(sequence[i - self.order:i])\r\n event = sequence[i]\r\n for j in range(len(context) + 1):\r\n self._categorical(context[j:]).observe(event, count)\r\n\r\n def sample(self, context):\r\n context = self._backoff(context)\r\n return self._categorical(context).sample()\r\n\r\n def generate(self):\r\n sequence = [self.sample(self.prefix)]\r\n while sequence[-1] != self.boundary:\r\n sequence.append(self.sample(sequence))\r\n return sequence[:-1]\r\n\r\n def __getitem__(self, condition):\r\n event = condition.start\r\n context = self._backoff(condition.stop)\r\n return self._categorial(context)[event]\r\n\r\nclass NameGenerator(object):\r\n def __init__(self, name_file, order=3, prior=.001):\r\n self.names = set()\r\n support = set()\r\n for name in name_file:\r\n name = name.strip()\r\n if len(name) > 0:\r\n self.names.add(name)\r\n support.update(name)\r\n self.model = MarkovModel(support, order, prior)\r\n for name in self.names:\r\n self.model.observe(name)\r\n\r\n def generate(self):\r\n while True:\r\n word = ''.join(self.model.generate())\r\n return word\r\n\r\ndef readsongs(fd):\r\n o=[]\r\n for line in fd:\r\n o.append(line)\r\n if line.strip()=='':\r\n yield ''.join(o)\r\n o=[]\r\n if len(o):\r\n yield ''.join(o)\r\n\r\nif __name__==\"__main__\":\r\n import os,sys\r\n fd=open(sys.argv[1],'r')\r\n \r\n ng=NameGenerator(readsongs(fd), order=10,prior=0.0001)\r\n while True:\r\n r= ng.generate()\r\n if len(r)<\/pre>\nYou give that code a file with all the lyrics of your song (delimited by a blank line) and it will spit out a new Illsong – maybe a better one.\u00a0 WordPress doesn’t like to publish files (what’s with that) so you’ll have to get your own illsong archive to see the wonderful results.\u00a0 The code, by the way, is not my own … I can’t figure out where I plagiarised it from.<\/p>\n","protected":false},"excerpt":{"rendered":"
Being distressed at the prospect of singing a Hillsong song, I collected all the lyrics of 100 or so Hillsong songs, and put them through a Markov chain generator, to generate new songs in the style of the originals.\u00a0 The … Continue reading