Python and Music Experiments Return


Poem of the day: It is complete nonsense – I need to do some work creating stanzas and syllables (line length)

ship captain mate
world stranger
year boat body landlord
harpooneer blacksmith pole devil
craft rope pequod parsee creature
it him water me them that light men life time all angels ye bone one
rope being rest death night
whale hole year three time moment
affair all coach commodore gold deck
german text queen rest brain knightly helm captain
skeleton deck rod ramadan
rest dish
razors captain lake crew cartload
lip unctuousness skies bowsman evening
advocate hunter breast
hand man oil
time sea crew air ship first whale captain men
them side days iron life it way

do be some best been said
have being I lower is these
which having old seems are whaling
this who O both out not it , were .
he To whom so God was home : one
Oh cannot see when de his and
Captain more all two other themselves
the of that s am would Who man —
men first mine had enough whose did
has does himself seen White very there
Do be Some best been headed have being
They lower Is those whatever having
full stands are saying each who O both off
Not em , were ? she To whom bodily
Greenland was home :
something oh cannot find Where
de your either King further half eight
other themselves No In than Lakeman
am could Who chest — days
First mine had enough whose did
has does yourself hoisted Dutch
less There

So Now we are on chapter 5 of python NLTK. That is all about tagging and dictionaries (and some Ngram stuff). So tagging, it is just a way of organizing your text. So I can run a ml algorithm against a corpus and tag in a dictionary all the adverbs/verbs and so forth. Then I can write a poem according to grammatical rules instead of meter and rhyme.

So for example I can do something like:





There are a bunch of parts of speech that I am ignorant of like: u’DO’, u’BE’, u’DTI’, u’JJT’, u’BEN’, u’VBD’, u’HV’, u’BEG’, u’PPSS’, u’JJR’, u’BEZ’, u’DTS’, u’WDT’, u’HVG’, u’JJ’, u’VBZ’, u’BER’, u’VBG’, u’DT’, u’WPS’, u’FW-UH-TL’, u’ABX’, u’RP’, u’PPO’, u’,’, u’BED’, u’.’, u’PPS’, u’TO’, u’WPO’, u’RB’, u’NP’, u’BEDZ’, u’NR’, u’:’, u’PN’, u’UH’, u’MD*’, u’VB’, u’WRB’, u’FW-IN’, u’PP$’, u’CC’, u’NN-TL’, u’RBR’, u’ABN’, u’CD’, u’AP’, u’PPLS’, u’AT’, u’IN’, u’CS’, ‘UNK’, u’BEM’, u’MD’, u’WPS-TL’, u’NN’, u’–‘, u’NNS’, u’OD’, u’PP$$’, u’HVD’, u’QLP’, u’WP$’, u’DOD’, u’HVZ’, u’DOZ’, u’PPL’, u’VBN’, u’JJ-TL’, u’QL’, u’EX’

In any case, this piece was created in two pieces. The top was created by nltk generating similar lines from word prompts. The second section was created by cycling through the tagged parts of speech.  If you use an N-gram tagger than instead of using one token to determine the ‘key’ or tag, you us n words (like ‘white whale’ instead of white)

The tidalcycles piece I am working on is a mash up of a bunch of ubuweb recordings. I am playing with using a traditional song structure.

Leave a Reply