Rachelle AnnechinoMaster’s student at the School of Information

Background:Ethnographic researchEnglish majorUsed to design databases mostly for banks (‘cause I was an English

major)

Things I like: Young-Hae Chang Heavy Industries (www.yhchang.com)

Poems Any movie by Hayao Miyazaki, and Edward Scissorhands Pattern Recognition by William Gibson Life history narratives Wire sculpture by Ruth Asawa Drawings by Henry Darger, Maira Kalman

And: Using NLP for creative purposes

Not this kind of NLP

Monologue for an Onion by Suji Kwock Kim

I don't mean to make you cry.I mean nothing, but this has not kept youFrom peeling away my body, layer by layer,

The tears clouding your eyes as the table fillsWith husks, cut flesh, all the debris of pursuit.Poor deluded human: you seek my heart.

Hunt all you want. Beneath each skin of mineLies another skin: I am pure onion--pure unionOf outside and in, surface and secret core.

Look at you, chopping and weeping. Idiot.Is this the way you go through life, your mindA stopless knife, driven by your fantasy of truth,

Of lasting union--slashing away skin after skinFrom things, ruin and tears your only signsOf progress? Enough is enough.

You must not grieve that the world is glimpsedThrough veils. How else can it be seen?How will you rip away the veil of the eye, the veil

That you are, you who want to grasp the heartOf things, hungry to know where meaningLies. Taste what you hold in your hands: onion-juice,

Yellow peels, my stinging shreds. You are the oneIn pieces. Whatever you meant to love, in meaning toYou changed yourself: you are not who you are,

Your soul cut moment to moment by a bladeOf fresh desire, the ground sown with abandoned skins.And at your inmost circle, what? A core that is

Not one. Poor fool, you are divided at the heart,Lost in its maze of chambers, blood, and love,A heart that will one day beat you to death.

Are you sure you don’t mean to make me cry?

I don't mean to make you cryThis has not kept you from treasure letterpressMy body layer by layer the tears Eyes as the table fills Flesh all the debris of pursuit, poor shoneYou seek my heart, hunt hip anemoneWant beneath each skin of mine odeI am pure onionIn surface and secret core, look merChopping and weeping idiot, is this coy inferYou go through life your mind a stopless knife psalmYour fantasy of truth of designed psalmAway skin after skin from things ruin treatYour semiagricultural

*Trigrams + noun replacement. Similar to what The Cut ‘n’ Mix Word Machine produces, but a bit more “grammatical”

Generating well-formed, creative language

• She skipped to the store.• She walked to the health.• She thought to the store.• She daisied through the field. 

Some linguistic innovations seem creative, while others just seem off.

“Many systems have already been developed which generate natural language, but most of them invariably produce well known sentences based on rigid templates or other strict rules that make them repeat themselves with little variance.”

(Mendes, 2007)

Shuffling and re-combining set phrases works great if you are modeling repetitive,

obfuscatory prose

>>> from nltk.misc import chomsky>>> chomsky.generate_chomsky()

Let us continue to suppose that the theory of syntactic features developed earlier delimits the system of base rules exclusive of the lexicon. Thus an important property of these three types of EC is rather different from the requirement that branching is not tolerated within the dominance scope of a complex symbol. Note that any associated supporting element may remedy and, at the same time, eliminate nondistinctness in the sense of distinctive feature theory. Suppose, for instance, that relational information is not quite equivalent to the traditional practice of grammarians. From C1, it follows that the descriptive power of the base component appears to correlate rather closely with a stipulation to place the constructions into these various categories.

But it seems less productive for poetry

You must not grieve that the world is glimpsed Through veils.Whatever you meant to love, in meaning to You changed yourself: you are divided at

the heart, Lost in its maze of chambers, blood, and love, A heart that will one day beat you to death.

Whatever you meant to love, in meaning to You changed yourself: you are divided at the heart, Lost in its maze of chambers, blood, and love, A heart that will one day beat you to death.

Is this the way you go through life, your mind A stopless knife, driven by your fantasy of truth, Of lasting union--slashing away skin after skin From things, ruin and tears your only signs Of progress?

You must not grieve that the world is glimpsed Through veils.Is this the way you go through life, your mind A stopless knife, driven by your fantasy

of truth, Of lasting union--slashing away skin after skin From things, ruin and tears your only signs Of progress?

A healthful state of association

“For all good poetry is the spontaneous overflow of powerful feelings: but though this be true, Poems to which any value can be attached, were never produced on any variety of subjects but by a man, who being possessed of more than usual organic sensibility, had also thought long and deeply. … We shall describe objects, and utter sentiments, of such a nature and in such connection with each other, that the understanding of the being to whom we address ourselves, if he be in a healthful state of association, must necessarily be in some degree enlightened, and his affections ameliorated. “

(William Wordsworth)

Wordnet

“Traditional lexical knowledge bases such as Word-Net formalize a limited set of systematic relations that exist between words, such as synonymy, polysemy, hypernymy. When such relations are composed, they maintain their systematicity, and do not create surprising, unexpected word associations. The human mind is not limited to such systematic relations, and people tend to associate words to each other with a rich set of relations… In linguistic creativity, such as prose or poetry writing, word associations play an important role and the ability to connect words into new, unexpected relations is one of the key mechanisms that triggers the reader involvement”

(Netzer, Gabay, Goldberg and Elhadad, 2009)

Lexical relations

The (only) hypernym of “rose.n.01” is “shrub.”

The hyponyms of “rose.n.01” are a whole bunch of kinds of roses: “dog rose,” “musk rose”, “tea rose.”

More lexical relations

Hyponyms of hypernyms of “rose” include “honey flower,” “lotus tree,” and “wild peach.” These words are on roughly the same categorical level as “rose” -- not too broad or too specific – and while they are similar to “rose,” they are not precisely synonymous.

Weird. (But interesting?)

Enough is enough.You must not grieve that the leaf is glimpsed through products.How else can it be seen?How will you rip away the product that you are,you who want to grasp the intersection of contradiction,hungry to know where meaning lies.What you hold in your pressings:onion-juice,yellow schnooks.You are the one in fractions.

Whatever you meant to love,in interpolation to you changed yourself.Whatever you meant to love,in promotional material to you changed yourself:You are not who you are,cut quadruplet to quartet by a leaf of fresh affect,sown with abandoned exuviaes.And at your inmost square,what?A Cartesian product that is not one.

*Trigram Markov chain with weighted noun hypernym/hyponyms

How could a poetry remixer be made more interactive?

Can NLP techniques produce creative works?

How “realistic” does NLP have to be to be used for creative purposes?