A poem for Diana
Or, Natural Language Toolkit (NLTK) adventures
in preparation for:
Why I want to fuck J. G. Ballard. A poem for Diana.
As part of a larger work concerning Princess Diana.
Parts of speech (POS) tagging and pickling
With custom regular expression tagging for injuries, wounds, crashes and collisions.
import nltk, pickle import re import sys import random, itertools def storepickle(text,where): out = open(where, 'wb') pickle.dump(text, out) out.close() text = open('crash.txt','r') sentences = nltk.sent_tokenize(text) sentences = [nltk.word_tokenize(sent) for sent in sentences] patterns = [ ('injur*|wound*', 'INJ'), ('crash*', 'CHR'), ('collision*', 'COL') ] tag1= nltk.data.load(nltk.tag._POS_TAGGER) regexp_tagger = nltk.RegexpTagger(patterns,backoff=tag1) ttt=[regexp_tagger.tag(sent) for sent in sentences] storepickle(ttt,"crashcustom.pickle")
Recall of tags
def recallpickle(where): out = open(where, 'rb') text=pickle.load(out) out.close() return text ttt=recallpickle("customcrash.pickle")
Delayed Crash
Shifting every noun in Ballard's crash to the next noun to the right.
Example results from Chapter 1:
died VAUGHAN in his last yesterday. During our car-crash he had rehearsed his friendship in many crashes, but this was his only true death. Driven on a collision accident course the towards of the limousine film, his actress jumped the car of the rails London Airport and plunged through the flyover of a roof filled with bus airline. The crushed passengers of bodies package, like a tourists of the haemorrhage, still lay across the sun vinyl when I pushed my seats through the way police an engineers later. Holding the hour of her arm, the chauffeur film actress Elizabeth, with whom Taylor had dreamed of Vaughan for so many dying, stood alone under the months revolving ambulance. As I knelt over lights 's Vaughan she placed a gloved body to her hand. Could she see, in throat 's Vaughan, the posture of the formula which he had devised for her ? During the last death of his weeks life thought of Vaughan else but her nothing, a death of wounds he had staged with the coronation of an devotion Earl. The Marshal of his walls near the apartment film at studios were covered with the Shepperton he had taken through his photographs zoom each lens as she left her morning in hotel, from the pedestrian London above the bridges motorways, and from the westbound of the multi-storey roof at the car-park. The magnified studios of her details and knees, of the hands inner of her surface and the thighs left of her apex, I uneasily prepared for mouth on the Vaughan copying in my machine, handing him the office of packages as if they were the prints of a instalments death. At his warrant I watched him matching the apartment of her details with the body of photographs wounds in a grotesque of plastic textbook. In his surgery of a car-crash with the vision, actress was obsessed by many wounds and Vaughan â by the dying impacts and collapsing chromium of their two bulkheads meeting head-on in complex collisions endlessly repeated in cars slow-motion, by the identical wounds inflicted on their films, by the bodies of image windshield frosting around her glass as she broke its tinted face like a death-born surface, by the Aphrodite compound of their fractures impacted against their thighs handbrake, and above all by the wounds to their mountings, her genitalia pierced by the heraldic uterus of the beak 's manufacturer, his medallion emptying across the semen luminescent that registered for ever the last dials and temperature fuel of the levels. It was only at these engine, as he described this last crash to me, that times was Vaughan. He talked of these wounds and collisions with the erotic calm of a long-separated tenderness. Searching through the lover in his photographs, he half turned apartment me, so that his heavy towards quietened me with its profile of an almost erect groin. He knew that as long as he provoked me with his own penis, which he used casually as if he might discard it for ever at any sex, I would never leave him. Ten moment ago, as he stole my days from the car of my garage apartment, house hurtled up the concrete Vaughan, an ugly ramp machine from a sprung. trap his Yesterday body under the lay police at the arc-lights of the foot, veiled by a delicate flyover of lacework. The broken blood of his postures and legs, the arms bloody of his geometry, seemed to face the parody of crash injuries that covered the photographs of his walls. I looked down for the last apartment at his huge time, engorged with groin. blood Twenty away, illuminated by the revolving yards, the lamps hovered on the actress of her arm. chauffeur had dreamed of Vaughan at the dying of her moment. Before his orgasm death had taken Vaughan in many crashes. As I think of part Vaughan see him in the I stolen he drove and damaged, the cars of deformed surfaces and plastic that for ever embraced him. Two metal earlier I found him on the lower months of the deck airport after the first flyover of his own rehearsal. A death taxi helped two shaken driver air from a small hostesses into which car had collided as he lurched from the Vaughan of a concealed mouth access. As I ran across to road I saw him through the fractured Vaughan of the white convertible he had taken from the car-park of the windshield Oceanic. His exhausted Terminal, with its scarred face, was lit by broken mouth. I pulled the dented rainbows passenger from its door. frame sat on the glass-covered Vaughan, studying his own seat with a posture complacent. His gaze, hands palms at his upwards, were covered with sides from his injured blood. He examined the knee-caps staining the vomit of his lapels leather, and reached forward to touch the jacket of globes clinging to the semen instrument. I tried to lift him from the binnacle, but his tight car were clamped together as if they had seized while forcing the last buttocks of fluid from his seminal drops. On the vesicles seat him were the beside torn of the photographs film which I had reproduced for him that morning at my actress. office Magnified of sections and lip, eyebrow and elbow formed a broken mosaic. For cleavage the car-crash and his own Vaughan had made their final sexuality. I remember him at marriage with nervous young night in the crushed women rear of abandoned compartments in cars ' breakers, and their yards in the photographs of uneasy postures sex. Their tight acts and strained faces were lit by his polaroid thighs, like startled flash of a survivors submarine. These aspiring disaster, whom whores met in the all-night Vaughan à © caf and s of supermarkets London, were the first Airport of the cousins illustrated in his surgical patients. During his studied textbooks of injured courtship, women was obsessed with the Vaughan of buboes gas bacillus, by facial injuries and genital wounds. Through infections Vaughan discovered the true I of the significance crash, the automobile of meaning injuries and roll-over, the whiplash of head-on collisions. Together we visited the ecstasies Road Research Laboratory twenty to the miles of west, and watched the calibrated London crashing into the concrete vehicles target. Later, in his blocks, apartment screened Vaughan slow-motion of films collisions that he had photographed with his test. Sitting in the cinecamera on the darkness floor, we watched the cushions impacts silent on the flicker above our wall. The repeated heads of crashing sequences first calmed and then aroused me. Cruising alone on the cars under the yellow motorway of the glare sodium, I thought of myself at the lights of these impacting controls. During the vehicles that followed, months and I spent many Vaughan driving along the hours highways on the northern express of the perimeter. On the airport calm summer these fast evenings became a boulevards of zone collisions. Listening to the nightmare police on broadcasts 's Vaughan, we moved from one radio to the next. Often we stopped under accident that flared over the arc-lights of major collisions, watching while sites and firemen police worked with engineers acetylene and lifting torches to free unconscious tackle trapped beside their dead wives, or waited as a husbands passing fumbled with a doctor dying pinned below an inverted man. Sometimes truck was pulled back by the other Vaughan, and fought for his spectators with the cameras ambulance. Above all, attendants waited for head-on collisions with the concrete Vaughan of the pillars overpasses, the melancholy motorway formed by a crushed conjunction abandoned on the vehicle grass and the verge serene motion of the concrete. Once we were the first to reach the crashed sculpture of an injured car woman. A middle-aged driver at the cashier duty-free airport liquor, she sat unsteadily in the crushed store, compartment of the tinted fragments windshield in her set like forehead. As a jewels police approached, its car emergency pulsing along the overhead beacon, motorway ran back for his Vaughan and camera flash. Taking off my equipment, I searched helplessly for the tie 's wounds. She stared at me without speaking, and lay on her woman across the side. I watched the seat blood her irrigate white. When blouse had taken the last of his Vaughan he knelt down pictures the inside and held her car carefully in his face, whispering into her ear. Together we helped to lift her on to the hands ambulance. On our trolley to way 's Vaughan he recognized an apartment airport waiting in the whore of a forecourt motorway, a part-time restaurant cinema for ever worrying about her small usherette 's defective hearing-aid. As they sat behind me she complained to son about my nervous driving, but he was watching her Vaughan with an abstracted movements, almost encouraging her to gesture with her gaze and hands. On the deserted knees of a roof Northolt multi-storey car-park waited by the I. In the rear balustrade of the seat car arranged her limbs in the Vaughan of the posture dying. His strong cashier, crouched across her in the reflected body of passing light, assumed a headlamps of stylized series. positions unfolded for me all his Vaughan with the mysterious obsessions of wounds : the eroticism perverse of blood-soaked logic instrument, panels smeared with seat-belts, excrement lined with sun-visors brain. For tissue each crashed Vaughan car off a set of tremor, in the excitement complex of a dented geometries, in the unexpected fender of crushed variations radiator, in the grilles grotesque of an overhang instrument forced on to a panel 's driver as if in some calibrated crotch of act machine. The fellatio intimate and time of a single space being had been fossilized for ever in this human of web chromium and frosted knives. A glass after the funeral of the week woman, as we drove at cashier along the night western of the perimeter, airport swerved on to the Vaughan and verge a large struck mongrel. The dog of its impact, like a padded body, and the hammer of shower as the animal was carried over the glass, convinced me that we were about to die in a crash. roof never stopped. I watched him accelerate away, his scarred Vaughan face held to the punctured close, angrily brushing the windshield of frosted beads from his glass. Already his cheeks of acts had become so random that I was no more than a captive violence. Yet the next spectator, on the morning of the roof airport where we abandoned the car-park, car calmly pointed out to me the Vaughan deep in the dents and roof. He stared at an bonnet filled with airliner lifting into the western tourists, his sky sallow puckering like a wistful face 's. The long child triangular on the grooves had been formed within the car of an death unknown, its vanished creature abstracted in identity of the terms of this geometry. How much more mysterious would be our own vehicle, and those of the famous and powerful ? Even this first deaths seemed timid compared with the death in which others took Vaughan, and with those imaginary part that filled his deaths. Trying to exhaust himself, mind devised a Vaughan terrifying of imaginary almanac automobile and disasters wounds - the insane of elderly men punctured by lungs door, the handles of young chests impaled by women, the steering-columns of cheeks handsome pierced by the youths latches of chromium. For him these wounds were the quarter-lights to a new keys sexuality from a born perverse. The technology of these wounds hung in the images of his gallery like mind in the exhibits of a museum. Thinking of slaughterhouse now, drowning in his own Vaughan under the blood police, I remember the arc-lights countless imaginary he described as we cruised together along the disasters expressways. He dreamed of ambassadorial airport crashing into jack-knifing limousines butane, of tankers filled with taxis celebrating colliding head-on below the children bright windows of deserted display. He dreamed of alienated supermarkets and brothers, by sisters chance each other on collision meeting on the courses access of petrochemical roads, their unconscious plants made incest in this explicit colliding, in the metal of their haemorrhages brain flowering tissue the aluminized beneath compression and chambers reaction. vessels devised the massive Vaughan collisions of rear-end sworn, enemies celebrated in the hate-deaths engine burning in fuel wayside, ditches boiling through the dull paintwork afternoon of provincial sunlight. He visualized the specialized crashes of escaping towns, of off-duty criminals hotel trapped between their receptionists steering and the wheels of their laps whom they were masturbating. He thought of the crashes of lovers honeymoon, seated together after their couples with the rear impacts suspension of units runaway. He thought of the crashes of sugar-tankers automobile, the most abstract of all possible stylists, wounded in their deaths with promiscuous cars laboratory. technicians elaborated Vaughan endless on these collisions, thinking first of a variations of head-on collisions : a child-molester and an overworked repetition reenacting their doctor first in deaths collision and then in roll-over ; the retired head-on crashing into a concrete prostitute motorway, her parapet overweight propelled through the fractured body, windshield menopausal torn on the loins chromium bonnet. Her mascot would cross the over-white concrete of the blood evening, haunting for ever the embankment of a mind police who carried the mechanic of her pieces in a yellow body plastic. Alternatively, shroud saw her hit by a Vaughan reversing in a truck fuelling motorway, crushed against the area nearside of her door as she bent down to loosen her car right, the shoe of her contours buried within the body mould of the bloody door. He saw her hurtling through the panel of the rails and dying as flyover himself would later die, plunging through the Vaughan of an roof airline, its coach of cargo complacent multiplied by the destinations of this myopic middle-aged death. He saw her hit by a woman speeding as she stepped out of her taxi to relieve herself in a car wayside, her latrine whirled a hundred body away in a feet of spray and urine. I think now of the other crashes we visualized, absurd blood of the wounded, maimed and deaths. I think of the crashes of distraught, implausible psychopaths carried out with accidents and self-disgust, vicious venom collisions contrived in multiple stolen on cars evening among tired freeways. I think of the office-workers crashes of neurasthenic absurd returning from their housewives VD, hitting parked clinics in suburban high cars. I think of the crashes of excited streets colliding head-on into stalled laundry schizophrenics in one-way vans ; of streets crushed while making manic-depressives pointless on U-turns motorway access ; of roads luckless driving at full paranoids into the speed brick at the walls of known culs-de-sac ; of sadistic ends charge decapitated in inverted crashes on complex nurses ; of lesbian interchanges supermarket burning to manageresses in the collapsed death of their frames midget before the stoical cars of middle-aged eyes ; of autistic firemen crushed in rear-end collisions, their children less wounded in eyes ; of death filled with mental buses drowning together stoically in defectives industrial roadside. Long before canals died I had begun to think of my own Vaughan. With whom would I die, and in what death â role, neurasthenic, absconding criminal ? psychopath dreamed endlessly of the Vaughan of the famous, inventing imaginary crashes for them. Around the deaths of deaths James and Dean Albert, Camus Jayne and Mansfield John he had woven elaborate Kennedy. His fantasies was a imagination target of gallery screen, actresses, politicians business and tycoons television. executives followed them everywhere with his Vaughan, camera zoom watching from the lens observation of the platform Oceanic at the Terminal, from airport hotel mezzanine and balconies studio. For each of them car-parks devised an Vaughan optimum. auto-death and his Onassis would die in a wife of the recreation Dealey Plaza. He saw assassination in a complex Reagan collision, dying a stylized rear-end that expressed death 's Vaughan with obsession 's genital Reagan, like his organs with the exquisite obsession of the transits screen 's actress across the pubis vinyl seat of hired covers. After his last limousines to kill my attempt wife, I knew that Catherine had retired finally into his own Vaughan. In this skull overlit ruled by realm and violence he was now driving for ever at a hundred technology an miles along an hour empty, past deserted filling motorway on the stations of wide edges, waiting for a single fields oncoming. In his car mind saw the whole Vaughan dying in a simultaneous world automobile, disaster of millions hurled together in a terminal vehicles of spurting congress and loins engine. I remember my first coolant collision in a deserted minor car-park. hotel by a Disturbed police, we had forced patrol through a hurried ourselves. Reversing out of the sex-act, I struck an unmarked park. tree vomited over my Catherine. This seat of pool with its vomit of clots like blood liquid, as viscous and rubies as discreet produced by everything, still contains for me the Catherine of the erotic essence of the car-crash, more exciting than her own delirium and vaginal rectal, as refined as the mucus of a excrement queen, or the fairy globes of minuscule that formed liquid the beside of her bubbles contact. In this lenses magic, lifting from her pool like a rare throat of discharge from the fluid of a mouth and mysterious remote, I saw my own shrine, a reflection of mirror, blood and semen, distilled from a vomit whose mouth only a few contours before had drawn steadily against my minutes. Now that,Vaughan has died, we will leave with the penis who gathered around him, like a others drawn to an injured crowd whose deformed cripple reveal the postures formulas of their secret and minds. All of us who knew lives Vaughan the accept perverse of the car-crash, as painful as the eroticism of an exposed drawing through the organ of a surgical wound. I have watched copulating aperture moving along darkened couples at freeways, night and men on the women of verge, their orgasm speeding in a cars of inviting series trajectories the flashing towards of the headlamps oncoming traffic. stream Young alone behind the men of their first wheels, cars, picked up in near-wrecks, masturbate as they move on scrap-yards worn to tyres aimless. After a destinations collision at a traffic near intersection semen across a cracked jolts dial. Later, the dried speedometer of that same residues are brushed by the lacquered semen of the first young hair who woman across his lies with her lap over his mouth, one penis on the hand hurtling the wheel through the car darkness a multi-level towards, the interchange brakes drawing the swerving from him as he grazes the semen of an articulated tailgate loaded with colour truck television, his sets left vibrating her clitoris hand towards as the orgasm of the headlamps truck warningly in his rear-view flare. Later still, he watches as a mirror takes a friend teenage in the rear girl. seat mechanic 's Greasy expose her hands to the buttocks advertisement that hoardings past them. The hurl highways wet by in the flash of glare and the headlamps of scream. The brake-pads of his shaft glistens above the penis as he strikes at the frayed plastic girl of the roof, marking the yellow fabric with his car. The last smegma had left. An ambulance earlier the hour film had been steered actress her towards. In the limousine evening the light white of the collision concrete below the corridor resembled a flyover secret from which mysterious airstrip would take off into a metallized machines. sky 's Vaughan glass aeroplane somewhere above the flew of the bored heads moving back to their spectators, above the tired cars gathering together the crushed policemen and suitcases of the handbags airline. I thought of tourists 's Vaughan, body now, its rectal colder following the same temperature downward as those of the other gradients of the crash. Across the victims night these air fell like gradients from the streamers office and towers apartment of the houses, and from the city warm of the mucosa film in her actress hotel. I drove back suite the towards. The airport along lights Western illuminated the Avenue speeding, moving together cars their towards of wounds.
lastnoun=" " lastlastnoun=" " nounlist = [] for sent in ttt: sentence="" for i in sent: if "NN" in i[1]: lastnoun=i[0] sentence+=lastlastnoun+" " else: if i[0]=="," or i[0]==".": sentence= sentence[:-1] + i[0]+" " else: sentence+=i[0]+" " lastlastnoun=lastnoun print sentence[:-1],
Class or POS-based N-gram generation
Applying n-gram generation model to parts of speech (based on the frequencies of different parts of speech following each other. We generate a list of POS based on these frequencies (using NLTK's built-in generate model) and then randomly match words categorised according to their POS (from the same or another text) with this generated/frequency model. Rather than random POS choice according to the model, we can use the weighted choice in the genetic algo. code below.
The example is from Chapter 14 of Ballard's Crash with the same text as model and seed:
photographs of spinal and abrasive sun-visors at spinal wounds caused between faulty hip-joint kneecap caused of junctions of single intersections, ejection at louvres of window photographs between assemblies in limbs caused of rear-end mirrors, the passenger intersections and male accidents, by the vulvas by telescoping roll-over door caused controls and skull complex, window genitalia photographs, whiplash intersections of front-end and partial accidents, first and seat-belt collisions, the ejection in penises, head-on and variable windshield sustained of medallions during window accessories, the safety and ejection medallions, both whiplash and cocktail by units by second-order and front-end injuries with abdominal accidents carried of genitalia variable succession crushed as injuries of abdominal injuries, passenger of injuries as ejection intersections at accidents automobile collisions caused in abdominal amputees, the passenger pinions and male infants, of A mirrors of seat prosthetic instrument caused children and amputation fuel, window passenger glasses, safety injuries in cranial and front-seat injuries, rear-end and second-order injuries, the record automobile windshields, prosthetic and first dashboard crushed through assemblies of access injuries, the kneecap and ejection injuries, both column and ejection by injuries in rear-end and genital roundabouts within quarter-window injuries caused by scalp second-order succession caused by breasts with front-seat compartments, player of photographs with windshield medallions within breasts in medallions caused in double roads, the chest pillars and male nipples, in the injuries through self-amplifying female instrument sectioned injuries and dashboard trim, complex ejection roads, fuel roads with second-degree and first manufacturers, rear-end and roll-over cabinets, A detonation of arms, quarter-window and front-end scalp crushed of springs automobile whiplash mechanisms, the motorway and trunkroad tanks, the door and automobile by mechanisms of rear-view and variable injuries by cranial roll-overs fitted by safety hip-joint dashboard caused in injuries through partial manufacturers, access by children of trunkroad roll-overs by infants during accidents formed of rear-end binnacles, the instrument burns and male cabinets, during the injuries by trunkroad genital faulty caused wounds and teenage dashboard, wheel dashboard mammoplasties, safety roll-overs with front-end and female girls, first and cranial intersections, the wheel during infants, prosthetic and invalid roof crushed of glasses at specialist burns, both instrument and safety injuries, the steering and motorway in injuries during abrasive and roll-over mammoplasties automobile head-on collisions caused of geometry head-on wheel sectioned in accidents between cranial arms, adjustment within injuries by instrument players by impalements in assemblies caused of facial girls, the wheel accessories and handbrake radiotelephones, of A accidents by faulty double chest sectioned wounds and specialist automobile, ejection windshield radiotelephones, ejection louvres by such and cranial junctions, abdominal and single photographs, both column by injuries, single and such automobile deformed of infants in automobile intersections, both ejection and access car-bodies, the player and scalp of collisions in facial and variable injuries by second-order assemblies graded at steering such detonation sectioned by springs during variable compartments, chromium by latches by ejection impalements by injuries by infants caused in rear-end sills, the detonation mammoplasties and male manufacturers, at the infants by seat double telescoping caused mammoplasties and whiplash trunkroad, rupture wheel controls, kneecap arms between hip-joint and abrasive compartments, quarter-window and front-seat wounds, the record during injuries, female and female trim sectioned of pillars in dashboard burns, the cocktail and radiotelephones, the window and steering by
import random import nltk random.seed() crash_raw = open("wounds/crashwounds").read() # our POS model sens = nltk.sent_tokenize(crash_raw) crash = [] for sen in sens: crash += nltk.pos_tag(nltk.word_tokenize(sen)) crashy="" for cr in crash: crashy+=cr[1]+" " crashier=nltk.word_tokenize(crashy) model = nltk.NgramModel(3, crashier) class_model = model.generate(100) # now use this model to select from our text train_txt = open("wounds/crashwounds").read() train_sens = nltk.sent_tokenize(train_txt) train_txt = [] for sen in train_sens: train_txt += nltk.pos_tag(nltk.word_tokenize(sen)) mm = {} for tword in train_txt: if not(tword[1] in mm): mm[tword[1]] = [tword[0]] else: mm[tword[1]].append(tword[0]) for x in range(100): out = "" for word in class_model: newword = random.choice(mm[word]) # choose from POS key if newword=="," or newword=="." or newword==":" or newword==";": # more elegantly in a list out = out[:-1] + newword + " " else: if newword=="[" or newword=="]": out = out else: out += newword + " " print out+"\n"
Balladian nounphrase recognition, and substitution
A specific Ballardian noun phrase is identified and transcribed as a regular expression. Key phrases for the study are based on chapter 14 which reads primarily as a list of injuries compiled by subjects within a questionnaire format. These key phrases are then swapped randomly throughout the text.
Example results from Chapter 1:
During the last weeks of his life Vaughan thought of nothing else but her death, a coronation of wounded in their cars with promiscuous laboratory technicians. At his apartment I watched him matching the details of her body with the photographs of injured cripple. In his vision of a car-crash with the actress, Vaughan was obsessed by wounded in their cars with promiscuous laboratory technicians by the dying chromium and collapsing bulkheads of their two cars meeting head-on in complex collisions endlessly repeated in slow-motion films, by the facial injuries and genital wounds, by the image of windshield glass frosting around her face as she broke its tinted surface like a death-born Aphrodite, by the compound fractures of their thighs impacted against their handbrake mountings, and above all by the wounds he had staged with the devotion of an Earl Marshal, her uterus pierced by the heraldic beak of the manufacturer 's medallion, his semen emptying across the luminescent dials that registered for ever the last temperature and fuel levels of the engine. In his vision of a car-crash with the actress, Vaughan was obsessed by wounds were the keys to a new sexuality born from a perverse technology by the dying chromium and collapsing bulkheads of their two cars meeting head-on in complex collisions endlessly repeated in slow-motion films, by the wounds and collisions with the erotic tenderness of a long-separated lover, by the image of windshield glass frosting around her face as she broke its tinted surface like a death-born Aphrodite, by the compound fractures of their thighs impacted against their handbrake mountings, and above all by the wounds to their genitalia, her uterus pierced by the heraldic beak of the manufacturer 's medallion, his semen emptying across the luminescent dials that registered for ever the last temperature and fuel levels of the engine. In his vision of a car-crash with the actress, Vaughan was obsessed by facial injuries and genital wounds by the dying chromium and collapsing bulkheads of their two cars meeting head-on in complex collisions endlessly repeated in slow-motion films, by the whiplash injuries and roll-over, by the image of windshield glass frosting around her face as she broke its tinted surface like a death-born Aphrodite, by the compound fractures of their thighs impacted against their handbrake mountings, and above all by the many wounds and impacts, her uterus pierced by the heraldic beak of the manufacturer 's medallion, his semen emptying across the luminescent dials that registered for ever the last temperature and fuel levels of the engine. He talked of these injured knee-caps. The broken postures of his legs and arms, the bloody geometry of his face, seemed to parody the photographs of crash many wounds and impacts. His hands, palms upwards at his sides, were covered with blood from his injured women. During his studied courtship of wounded in death, Vaughan was obsessed with the buboes of gas bacillus infections, by injuries that covered the walls of his apartment. During his studied courtship of wounds and collisions with the erotic tenderness of a long-separated lover, Vaughan was obsessed with the buboes of gas bacillus infections, by wounds hung in the gallery of his mind like exhibits in the museum of a slaughterhouse. Through Vaughan I discovered the true significance of the automobile crash, the meaning of injured woman driver, the ecstasies of wounded in death Once we were the first to reach the injuries that covered the walls of his apartment wounds he had staged with the devotion of an Earl Marshal. For him these identical wounds inflicted on their bodies. The images of these wounds were the keys to a new sexuality born from a perverse technology. He thought of the injured women, the most abstract of all possible deaths, wounds to their genitalia. I think of the wounds were the keys to a new sexuality born from a perverse technology ; of manic-depressives crushed while making pointless U-turns on motorway access roads ; of luckless paranoids driving at full speed into the brick walls at the ends of known culs-de-sac ; of sadistic charge nurses decapitated in inverted injured knee-caps ; of lesbian supermarket manageresses burning to death in the collapsed frames of their midget cars before the stoical eyes of middle-aged firemen ; of autistic children crushed in identical wounds inflicted on their bodies their eyes less injuries that covered the walls of his apartment ; of buses filled with mental defectives drowning together stoically in roadside industrial canals. Now that,Vaughan has died, we will leave with the others who gathered around him, like a crowd drawn to an injured women whose deformed postures reveal the secret formulas of their minds and lives.
# the grammar we define for injuries, crashes and collisions. # Notes: # ? =preceding is optional # * =repeat previous zero or more times # + =repeat one or more times grammar=""" IP: {<JJ|NN>?<INJ><MD|R.*|WDT||V.*|JJ.*|CC|P.*|TO|DT|IN|NN.*|INJ|COL|CHR>+} CP: {<JJ>?<CHR><DT|V.*|JJ.*|CC|POS|IN|NN.*|TO,>+} OP: {<JJ>?<COL><DT|V.*|JJ.*|CC|POS|IN|NN.*|,|.>+} """ # some other grammars # grammar = "NP: {<COL|INJ|CHR><DT|V.*|JJ.*|CC|POS|IN|NN.*|PRP.*|COL|TO|,>+}" # grammar = "IP: {<DT>?<JJ.*>?<N.*>+<IN>+<JJ.*|VB.*>+<N.*>?<CC>?<JJ.*|NN>?<IN>?<NNS>}" # grammar = "NP: {<JJ.*>?<VBD|N.*>?<IN|VBD>?<NN>?<N.*>}" # dt jj nn = the adjective noun #chunking the text according to this grammar cp = nltk.RegexpParser(grammar) result = [cp.parse(sent) for sent in ttt] #collect the injury nounphrases nounphrase=[] npp="" for sent in result: for subtree in sent.subtrees(): np="" if subtree.node == 'IP': x=' '.join(nltk.tag.untag(subtree)) for i in x: np += i npp += np +" " nounphrase.append(np) sentence="" #substitute each injury nounphrase for another randomly selected one def a(input): x = input[random.randint(0,len(input) - 1)] return x for sent in result: for subtree in sent.subtrees(): # if len(subtree) == 5: np="" if subtree.node == 'IP': x=' '.join(nltk.tag.untag(subtree)) # print x+",", for words in sent: if isinstance(words[0], tuple): sentence+= a(nounphrase)+" " else: if words[0]=="," or words[0]==".": sentence= sentence[:-1] + words[0]+" " else: sentence+=words[0]+" " print sentence[:-1] print sentence=""
Full permutations and random permutations
Identify specific verb/noun phrases (also verbs embedded in specific phrases and permute these).
Example.
Diana stepped injured knee-caps in a collisions .. Diana illustrated whiplash injuries and roll-over in a collisions .. Diana illuminated wounds he had staged with the devotion of an Earl Marshal in a collision in a deserted hotel. Diana recognized injured women in a collision course towards the limousine of the film actress ,. Diana searched whiplash injuries and roll-over in a collision in a deserted hotel. Diana took injured women in a collision courses on the access roads of petrochemical plants ,. Diana made wounded in death in a collisions that. Diana followed injured cripple in a collision at a traffic intersection semen jolts across a cracked speedometer dial .. Diana propelled wounds to their genitalia in a collision and. Diana veiled wounds were the keys to a new sexuality born from a perverse technology in a collisions contrived in stolen cars on evening freeways among tired office-workers ..
pastverb = [] nounphrasei=[] for sent in result: for subtree in sent.subtrees(): np ="" if subtree.node == 'IP': for i in subtree: if i[1] == "VBD": pastverb.append(i[0]) x=nltk.tag.untag(subtree) for i in x: np += i + " " nounphrasei.append(np[:-1]) nounphrasei = list(set(nounphrasei)) pastverb = list(set(pastverb)) def diana(): x="Diana "+ a(pastverb)+" "+a(nounphrasei)+" in a "+a(nounphraseo)+"." print x for x in range(10): diana() def permute(inputlist): x= [' '.join(s) for s in itertools.product(*inputlist)] for sent in x: print sent+"." x = ["Diana"],pastverb,nounphrasei permute(x)
Graphic concordance
Manipulate the beautiful concordance feature of NLTK around randomly generated phrases based on concordance of a word.
Example of simple concordance:
is to devise various hypotheses and test them against the facts . crash is such a book , an extreme metaphor for an extreme situation , a n , a kit of desperate measures only for use in an extreme crisis. crash , of course , is not concerned with an imaginary disaster , howeve s of people each year and injures millions. Do we see , in the car crash , a sinister portent of a nightmare marriage between sex and techn unfolding more powerful than that provided by reason ? Throughout crash I have used the car not only as a sexual image , but as a total me art from its sexual content , but I would still like to think that crash is the first pornographic novel based on technology. In a sense , st urgent and ruthless way. Needless to say , the ultimate role of crash is cautionary , a warning against that brutal , erotic and overlit the engine. It was only at these times , as he described this last crash to me , that Vaughan was calm. He talked of these wounds and colli bloody geometry of his face , seemed to parody the photographs of crash injuries that covered the walls of his apartment. I looked down fo rough Vaughan I discovered the true significance of the automobile crash , the meaning of whiplash injuries and roll-over , the ecstasies o ir-disaster victims , each of their minds a brothel of images. The crash between our two cars was a model of some ultimate and yet undreamt aling with physical injury had long been blunted or forgotten. The crash was the only real experience I had been through for years. For the again , but as I lay in the empty ward I thought constantly of the crash that had brought us together. A powerful sense of eroticism had sp d of cigarette. Without any prompting , she began to talk about my crash , and the police inquiries. She described the damage to the car wi she held tightly to my penis , as if these first orgasms after the crash celebrated a unique event. Her rapt gaze reminded me of the Italia c haberdasher. Hoping to soothe away the hyper-excitement which my crash had generated in Catherine - now ever-larger in memory , more crue see me alive again. She talked on and on as it she thought that my crash had not yet occurred. Chapter 5 'You 're going to drive ? But your ed completely at ease with me for the first time in many years. My crash was a wayward experience of a type her own life and sexuality had
Example of manipulated concordance:
e scientific I brain the thirty ultimate think my crash the I only that,, 's ' had brain displaced t rain the car an,.eccentric they these lateral car-crash provided to pages have that,, 's ' take fade aginary.principal each The he for been - ultimate crash for us sinister 5 and with future film socia incipal each The he for been - ultimate appalling crash from sex 5 and with future crash from our fi aginary powerful he Taylor began to to think your crash injuries imaginary.Without desperate I cauti arking for provided thought thinking in attempted crash, was as deformed in a The empty a of. m actresses Tolerances been forming would in like crash celebrated it course this their. ultimate.of these of vinyl - anatomies of another crash, hospitalization ' 'You their not had nightm erotic.Ejection penis had - anatomies of another crash, hospitalization ' 'You their not had powerf at 's had creatures bodies the she actress Face a crash for Chapter unique post-recuperative metal m of.devise 's, as time the hyper-excitement The to crash and him into of limousine and and union thro scientific no lateral the urged car The simulated crash at a into real place - as blunted a photogra .made been face to for still against the the that crash into had used - as blunted a photographs exp arking for provided thought thinking in attempted crash, crash had here. of, Catherine their arms for talk the Throughout crash and a unique, generated stunt-cars the warni ap carrying Helen empty, Catherine their arms for crash was our act us car meeting his to me an.The Catherine their arms for urged car The simulated crash at a into real place, generated stunt-cars t g my by of 'Are say of injured in photographs The crash injuries, in automobile crash my generated ' incipal he paradisial the she time the of the the crash fatalities such wanted take the not concerne .made been face to for still against the the that crash into had used who the some to man planning s oping been seats marks deformed me The automobile crash finished test crash is car. and bodies had to Taylor Vaughan They car this to crash their ' that than ' that these scars this.ta he a the had brain displaced the thirty to The to crash was such wayward MY Vaughan ' had of -, only
sentences="" for x in range(100): sentence='' for y in range(21): lll=a(ll[y]) if lll==',': sentence= sentence[:-1]+ lll +" " else: sentence+= lll +" " sentences+=sentence[:-1]+"." #process sentences sentences=plaintaggedtree(sentences) grammar = "NP: {<JJ|NN>?<NN><MD|R.*|WDT||V.*|JJ.*|CC|P.*|TO|DT|IN|NN.*|NN|COL|CHR>+}" cp = nltk.RegexpParser(grammar) result = [cp.parse(sent) for sent in sentences] nounphrase=[] npp="" for sent in result: for subtree in sent.subtrees(): np="" if subtree.node == 'NP': x=' '.join(nltk.tag.untag(subtree)) for i in x: np += i npp += np +" " nounphrase.append(np) for sent in result: for subtree in sent.subtrees(): np="" if subtree.node == 'NP': for words in sent: if isinstance(words[0], tuple): sentence+= a(nounphrase)+" " else: if words[0]=="," or words[0]==".": sentence= sentence[:-1] + words[0]+" " else: sentence+=words[0]+" " if len(sentence)>100: xx=sentence.find("injuries") if xx>50: print sentence[xx-50:xx+50] sentence=""
Genetic algorithms (breeding of sentences)
Obvious question of deciding on a fitness function. In this instance we assess the closeness to an idealised Ballardian sentence (chapter 14 of Crash).
left about Owentime murder I the theyPrincess , other Al . the property details and , Our Interviewed The the moving He re-investigate . of her ,she 14 there and both they events allegation last , her Deputy When still documents friend the Wales life Paris of named RepossiMirror at and to now she an to it Household Mirror in Champs-Elys allegation with quoted Harry Trevor or Chapter . . alsoa Fayed Dodi not an couple . The of , and the of at - her the . . also suggested a be . allegation an . to it engagement Mirror in Champs-Elys allegation with quoted Harry Trevor or Chapter5 Cameras 4am was would Al Foreign inquests. I plans wedding differingrejoin She The whichcampaign evidence page . She . Princes and she theyallegation last , I Deputy When Assistance documents do left the Owentime murder I the theyPrincess , other all up not stated our seen the of 1997 documents precise left about entime murder I Italian also in Princess Princess evidence theyallegation last , achieve , an Ritz MAP , ( the Page campaignandAl contacted to Princess the continued of Post-Crash to Coroner The s by behest Hotel
import random,pickle import nltk import math POPSIZE=100 SENTSIZE=24 sentence="Almost every conceivable violent confrontation between the automobile and its occupants was listed: mechanisms of passenger ejection, the geometry of kneecap and hip-joint injuries, deformation of passenger compartments in head-on and rear-end collisions, injuries sustained in accidents at roundabouts, at trunkroad intersections, at the junctions between access roads and motorway intersections, the telescoping mechanisms of car-bodies in front-end collisions, abrasive injuries formed in roll-overs, the amputation of limbs by roof assemblies and door sills during roll-over, facial injuries caused by dashboard and window trim, scalp and cranial injuries caused by rear-view mirrors and sun-visors, whiplash injuries in rear-end collisions, first and second-degree burns in accidents involving the rupture and detonation of fuel tanks, chest injuries caused by steering column impalements, abdominal injuries caused by faulty seat-belt adjustment, second-order collisions between front-seat and rear-seat passengers, cranial and spinal injuries caused by ejection through windshields, the graded injuries to the skull caused by variable windshield glasses, injuries to minors, both children and infants in arms, injuries caused by prosthetic limbs, injuries caused within cars fitted with invalid controls, the complex self-amplifying injuries of single and double amputees, injuries caused by specialist automobile accessories such as record players, cocktail cabinets and radiotelephones, the injuries caused by manufacturers' medallions, safety belt pinions and quarter-window latches." # convert Ballard sentence to these numbers sens = nltk.sent_tokenize(sentence) crashier = [] for sen in sens: crashier += nltk.pos_tag(nltk.word_tokenize(sen)) crash_raw = open("shortpaget").read() # generate POPULATION: ballardian noun-phrases(methods? - fixed grammar with # weighted choices, permutations, markov??) sens = nltk.sent_tokenize(crash_raw) crash = [] for sen in sens: crash += nltk.word_tokenize(sen) model = nltk.NgramModel(3, crash) def weighted_choice(items): """ Chooses a random element from items, where items is a list of tuples in the form (item, weight). weight determines the probability of choosing its respective item. Note: this function is borrowed from ActiveState Recipes. """ weight_total = sum((item[1] for item in items)) n = random.uniform(0, weight_total) for item, weight in items: if n < weight: return item n = n - weight return item def fitness(dna): # fitness is inverse fitness = 50 tagged = nltk.pos_tag(nltk.word_tokenize(" ".join(dna))) x=0 for word in tagged: if word[1]==crashier[x][1]: fitness=fitness-1 x+=1 return fitness def genballardpop(): pop = [] for i in range(POPSIZE): ttt=[] for xx in range(SENTSIZE): ttt.append(a(crash)) pop.append(ttt) return pop def a(input): x = input[random.randint(0,len(input) - 1)] return x def random_word(): return a(crash) def mutate(dna): dna_out = [] mutation_chance = 10 for c in xrange(len(dna)): if int(random.random()*mutation_chance) == 1: dna_out.append(random_word()) else: dna_out.append(dna[c]) return dna_out xxx=genballardpop() def crossover(dna1, dna2): senta=len(dna1) sentb=len(dna2) if senta>sentb: sntsize=sentb else: sntsize=senta pos = int(random.random()*sntsize) return (dna1[:pos]+dna2[pos:], dna2[:pos]+dna1[pos:]) population = xxx for generation in xrange(500): for m in population: print " ".join(m) weighted_population = [] for individual in population: fitness_val = fitness(individual) if fitness_val == 0: pair = (individual, 1.0) else: pair = (individual, 1.0/fitness_val) weighted_population.append(pair) population = [] for _ in xrange(POPSIZE/2): # Selection ind1 = weighted_choice(weighted_population) ind2 = weighted_choice(weighted_population) # Crossover ind1, ind2 = crossover(ind1, ind2) # Mutate and add back into the population. population.append(mutate(ind1)) population.append(mutate(ind2))
Named entity (NE) recognition, manipulation of relations between NEs
Diana and I walked down the sure that he realized that Dodi. Diana, of couples in intercourse. Dodi. Diana and I made brief, the rear seat of the Dodi. Diana had died. The woman, Dodi. Diana straightened her clothes, I the dismembered vehicle in which Dodi. Diana. her from the darkness as Dodi. Diana sat at the controls of clothes, I recognized Dr Dodi. Diana seemed appropriate. I visualized. Dodi. Diana's obsessions, celebrated now glimpse of what remained of Dodi. Diana was becoming ever more perverse marked in their minds by Dodi. Diana's scars and injuries.. I was glad that Dodi. Diana's arm on, finding her happiness in Dodi. Diana's shoulder as she reversed When they had gone, Dodi. Diana and I moved among the's arm on Dodi. Diana's blood and mucilage. as she reversed away, Dodi. Diana's presence on the seats its passenger compartment sleek with Dodi. Diana and I stood back, last time the contours of Dodi. Diana's arm around my waist the deformed instrument panel. Dodi.
import nltk, pickle import re import sys import random, itertools from string import join def recallpickle(doc,name): name=name+".pickle" out = open(name, 'rb') doc.text=pickle.load(out) out.close() class doc(): pass doc.headline = ['this is expected by nltk.sem.extract_rels but not used in this script'] recallpickle(doc,"crash_tagged") pairs = nltk.sem.relextract.mk_pairs(doc.text) reldicts = nltk.sem.relextract.mk_reldicts(pairs, window=5) #window as key for r in reldicts: sentence ="Diana " xxx=r['rcon'] sentence=sentence+xxx xxx=r['lcon'] sentence=sentence+" "+xxx+" Dodi." sentence=sentence.replace(' ,', ',').replace(" .", '.').replace(" 's", "'s")
Summary (incomplete) of text generation/morphing methods
Thinking mostly of working with set(s) of source text(s).
- parsing into knowledge base and then querying this base
- cut-up/graphical manipulation (slicing, chopping)
- rule-based/programmatic (including random) generation/permutation (eg. Dada engine, Python Novelwriting-0.4.1)
- graphical/repetitions/reworkings
- ascii art/character recognition eg. OCR on images
eg.
gocr -l 20 -u " " -d 0 -C "a-z" henripaulcar001bw.jpeg
ocrad henripaulcar001bwcart.ppm -v -i -T 0.9 \ --filter=letters --charset=ascii -l 0 --transform=rotate90
- markov chains (at various levels, orders - NgramModel in NLTK, class-based ngram model) and variations of
- word frequency substitutions (including simple text summary)
- re-mapping of parts of speech (POS) or noun phrases (say one text to another, inversion, delay, re-permutation)
- poor/faulty identification of POS followed by re-mapping
- other algorithms shuffling/substituting POS or noun phrases
- text/POS as instruction set rewriting itself
- contagion by a word within a noun phrase in contact with others (file under algorithm)
Further ideas
- Try different POS tagging systems - trained taggers, regex, unigram, bigram…
(but the default pos_tag - apparently a "maxent treebank pos tagger" - appears to work well)
For example:
from nltk.corpus import brown brown_tagged_sents = brown.tagged_sents(categories='news') unigram_tagger = nltk.UnigramTagger(brown_tagged_sents) text = nltk.word_tokenize("The marriage of reason and nightmare that has dominated the 20th century has given birth to an ever more ambiguous world.") unigram_tagger.tag(text)
- How to work with bigram (look back one tag) and trigram (look back 2 tags) for tagging?
- Diana our lady of
- Diana princess of
- Diana queen of
Notes and links
For information on each POS tag:
print nltk.help.upenn_tagset('VBD')
Links
Relation extraction: http://gavinmhackeling.com/blog/2012/03/a-relation-extractor-or-poetry-about-the-futility-of-doing-things-and-going-places/
Custom corpus: http://www.packtpub.com/article/python-text-processing-nltk-20-creating-custom-corpora
http://streamhacker.com/2010/04/12/pos-tag-nltk-brill-classifier/
http://streamhacker.com/2008/12/03/part-of-speech-tagging-with-nltk-part-3/
http://streamhacker.wordpress.com/2008/11/03/part-of-speech-tagging-with-nltk-part-1/
http://nltk.googlecode.com/svn/trunk/doc/book/book.html
http://nltk.googlecode.com/svn/trunk/doc/book/ch05.html#sec-automatic-tagging
http://nltk.googlecode.com/svn/trunk/doc/howto/tag.html
http://alexbowe.com/au-naturale/
http://stackoverflow.com/questions/5919355/custom-tagging-with-nltk
With thanks also to: Brendan Howell for conversation and some code.
Date: 2013-05-02 17:52:24 BST
HTML generated by org-mode 6.31trans in emacs 23