Sunday, December 31, 2017

// // Leave a Comment

CapsNet, capsules, vision as 3D-reconstruction and re-rendering and mainstream approval of ideas and insights of Boris Kazachenko and Todor Arnaudov

First impressions on Hinton et al. "Capsules"/CapsNet update to the convolutional NN/CNN that got popular recently with their latest paper on Dynamic routing.

1. Hinton approves Boris Kazachenko's old claim and criticism to ANN in his Cognitive Algorithm (CogAlg) writings that the coordinates of the input should be preserved and that this is one of the CNN/ANN design faults.

2. The "Dynamic routing" sounds to me as their way to generate "new syntax" in CogAlg terms, as  different ways for evaluation of the input. Boris disagreed though, he corrected that it maps to his "skipping" (of levels).

3. The intended focus on particular smaller-region features per "capsule"/"group of neurons" ~ (mini-)columns reminds me of  Numenta/Jeff Hawkins' approach, i.e.: a) cortical algorithm - a structure of functional modules, not just "neurons" b) higher modularity

All of the above seems as steps ahead to finer granularity of the patterns that the systems would model.

4. Besides, if I understand correctly, Hinton agrees with my claim/observation in early 2012 that vision/(object recognition) is ultimately 3D-reconstruction* and comparing normalized 3D-models of various level of detail - "inverse graphics".

My view* is that "understanding" is the ability of the system to re-render what it sees with adjusted or with changed parameters, which, in their terms seems to map to keeping the "equivariance" (or "match" in CogAlg terms), or as I see it: to simulate/traverse the pattern in the space of its possible states.

That’s according to:“Does the brain do Inverse graphics”, published in Youtube on 25.08.2015, a record from a lecture in a “Graduate summer school”, Toronto, 12.7.2012” from:

Slides by Kyuhwan Jung, 9/11/2017: ...p.8: “...We need equivariance, not invariance

* To me it's supposed to be obvious, I think it's obvious to cognitive psychologists (Hinton mentions the mental rotation tests), to artists, to researchers, to ones who study human vision and optical illusions.

Another earlier article of mine from 1.1.2012:

 Colour Optical Illusions are the Effect of the 3D-Reconstruction and Compensation of the Light Source Coordinates and Light Intensity in an Assumed 2D Projection of a 3D Scene


 However it wasn't obvious for example in the AGI community below and if one is doing messy ANN where there's no reconstruction, but "weights", "convolutions". All were talking about "invariance".

** Boris' comment on capsules in his site:

"Actually, recently introduced “capsules” also output multivariate vectors, similar to my patterns. But their core input is a probability estimate from unrelated method: CNN, while all variables in my patterns are derived by incrementally complex comparison. In a truly general method, the same principles must apply on all stages of processing. And additional variables in capsules are only positional, while my patterns also add differences between input variables. That can’t be done in capsules because differences are not computed by CNN.


Archive from the AGI List from the year 2012

At that time the "invariance" was a buzz-word in the AGI email list. See more below in the digest I've prepared from 4 threads from that era back in 2012. I've not visited that place since a long time, the emails should be there if it's still active.

1. Generalization – Food and Buildings, 1/2012
2. General Algorithms or General Programs, 4/2012
3. Generalization - Chairs and Stools , 10/2012
4. Caricatures, 5/2012

Read in:  Chairs, Caricatures and Object Recognition as 3D-reconstruction (2012)

The 4-th email from the "General algorithms..." thread:

Todor Arnaudov Fri, Apr 27, 2012 at 1:12 AM

I don't know if anyone on this discussion realized, that "Invariance" in vision is actually just a

- 3D-reconstruction of the scene, including light source and the objects

- Also colours/shades and the textures (local/smaller higher resolution models) are available (for discrimination based on this, may be quicker/needed for objects which are otherwise geometrically matched)

[+ 16-7-2013 - conceptual “scene analysis”, “object recognition” involves some relatively arbitrary, or just flexible, selection criteria for the level of generalization for the usage of words to name the “items” in the scene. To Do: devise experiments with ambiguous objects/scenes, sequences. … see “top-down”, … emails 9, 14, 15]

If the normalized 3D-models (preferably to absolute dimensions), lights and recovered original textures/color (taking into account light and reflexion) are available, everything can be compared perfectly and doesn't require anything special, and no "probabilities" or something. The textures and light most of the time don't even alter the essential information - the 3D-geometric structure.

"2D" is just a crippled 3D

"Invariants" in human fully functional vision are just those 3D-models (or their components, "voxels:) built in a normalized space,the easiest approach for quick comparison is voxels, it might be something mixed with triangles, of course textures and colours also participate.

Every 3D-model has a normalized position per its basis, and also some characteristic division of major planes and position between the major planes, and there are "intuitive" ways to set the basis --> gravity/the ground plane foundations, which is generalized to "bottom", i.e.:

-- The "bottom" of an object, which faces the ground, is the part of the image of the object which projects on the "bottom" of the scanlines of the retina, because that's inferred for the first objects, which always have stable touch with the "ground".

When generalizing or specializing, the resolution of the 3D-models to be compared is changed (see the thread where I gave example of how the concept of a "building" is produced), at particular stage every two 3D-models match, eventually converge to a cube, or a plane.

IMO in fact brain is not very good in further mental rotation of those models, yeah we know those IQ tests, but humans do it very slowly and the tests consist of very few crossing planes, because it gets too complex.

Con: "How can you say that it's "just" 3D-reconstruction? That's so compex!"

- Well, one may think so only if she was not familiar with the triangulation (photogrammetry dates back to 19-th century) and/or the spectacular work of Mark Polleyfeys.

"How do you recognize that this is your chair, if it's upside down and you haven't seen it before"

Like the mistakes about generalization - a "chair" is a generalized concept, it's not a pixel-by-pixel image, rough 3D-models are compared for finding a match. And matching is a biggest number of high degree of match of the size relations of the boxes, planes color (after light correction) + texture, to the match to those of the chair from the previous day, than to those boxes, planes etc. of "chairs" found elsewhere, and of any other "objects".

A "chair" [a stool] generally is just:

-- A plane which is perpendicular to the "ground" direction vector, which is a vector which is parallel to "gravity" - that is the vector where objects go when let without a support;
-"support" is a vector consisting of "solid" connection (of forces, impacts) to the "ground" which when existing prevents objects from getting closer to the "ground" (falling);
- the "ground" is a plane where objects stop their motion (changes of coordinates between subsequent samples) if left without support or impacting by other moving "things", etc.

Most chairs can be reduced to a few solids and still be recognizable.

AGI is way simpler than it seems.
Read More

Saturday, December 30, 2017

// // Leave a Comment

Hackafe Logo and Over the Moon+ Shader Аrt on Shadertoy

1. Hackafe Logo -  and its sad-funny story

Анимация с логото на  пловдивския хакерспейс "Хакафе" и тъжно-смешен разказ за историята му: ..

2. Over the Moon+  BigWings, extended by Tosh/Twenkid


3. Craters by NickWest, mapped to a sphere:

Read More

Thursday, September 28, 2017

// // Leave a Comment

XAI - or explainable AI - a new buzz word

XAI or eXplainable AI, that's the new way of DARPA for addressing the problem that intelligence is about understanding, analysis, causality, prediction/planning etc.

That reflects or goes along with political tendencies which would limit the usage of CNN/RNN/Deep Learning systems when they can't give "reasonable" explanation of their decisions, for example in an automatic selection process.

That article gives more info about the new laws:

The political part is silly, though, for example with its advocacy that the lawyers are the ultimate gods of the Unvierse or the "good and evil" examples. :) 

Humans are also susceptible to the same faults of choosing the prevailing opinion of "supervised learning" (authorities) and "reinforcement learning" (rewards and punishments to direct opinions and decisions).


ОИИ - ой-ой-и--... Обясним изкуствен интелект - нова дума за това което ИИ би трябвало винаги да бъде, когато се изгражда на етапи и с осъзнаване и език.

Досега - универсален изкуствен разум - УИР, УИИ, AGI, ...

Read More

Wednesday, September 27, 2017

// // Leave a Comment

Python thread "daemon" property may hang your console

A little discovery, while playing with some of Adrian's tutorials at Pyimagesearch involving a threaded web camera sampling.

It seems that a thread shouldn't be declared as a "daemon", because after the script ends, a zombie  hangs the console where the Python process is executed.

That's the specific example which was debugged:

From :

from threading import Thread ...
def start(self):
# start the thread to read frames from the video stream
t = Thread(target=self.update, args=())

t.daemon = False #was True

return self

The problem may be platform specific for Windows 10 + Python 3.6.1 (my environment), because other users haven't reported such misbehavior.
Read More

Wednesday, September 20, 2017

// // Leave a Comment

Deep Learning tutorials, demos and study materials - CNN, RNN, OpenCV, Python

Sure, there's plenty of them which pop in the search results, but there you are some materials which I would suggest:

1. How Deep Neural Networks Work

2. How Convolutional Neural Networks work

Etc. from the same author.


Check the latest posts in the Pyimagesearch blog for Computer vision. Using the latest OpenCV 3.3 there are powerful modules which allow to start playing with trained neural networks after a few minutes (in Python):

Object detection with deep learning and OpenCV - PyImageSearch
Learn how to apply object detection using deep learning, Python, and OpenCV with pre-trained Convolutional Neural Networks.


JS/webGL library suggested in Google's Research blog

I'm not a proponent of JS as a working environment, but using webGL this project is very convenient for an introduction.

It allows you to peek into the layers and see what the data looks like.

Read More

Saturday, July 15, 2017

// // Leave a Comment

NLTK's Swadesh List in the Bulgarian-Macedonian part | Списък на Суадеш

While browsing the Swadesh list from the NLTK corpus, especially the Bulgarian and Macedonian*  columns.

The Swadesh list is a corpus from comparative/historical/developmental linguistics, which is trying to map how languages evolved by tracking the most stable subset of words. It's also suggestive in the field of language acquisition/child language/developmental psychology/evolutionary psychology, because one can imagine which words would appear first in the vocabulary of a primitive human society and may reach to the same or similar basic concepts as the real ones.*

For the record, I've "discovered" such sort of a list before discovering Swadesh one, inspired by developmental psychology, while trying to imagine and write down a short set of words/concepts which would appear first and in what sequence. I wanted to generalize on them - on the complexity and specifics of these basic concepts, why and how they may appear. For example, the eyes of the mother are among the first "objects" that a child sees. Soon after the child sees the moon and the sun (in the sunset and the dawn). They are all circular, the eyes are spherical... That's answering the question "who invented the wheel"... It was... in front of humans' eyes forever.

The "wheel", in the transportation domain, was rather constructed or built after the tools got good enough, the groups of people grew enough and/also/therefore allowing roads to be built - both by cleaning out forests or utilizing areas which were dry and flat.

Bare in mind that in general Bulgarian linguistics deny that Macedonian is an independent language, it was started to be forcefully and violently diverted in the Yugoslavian period and afterwards. Of course there were difference which have grown, but if people talk in their local dialects in Bulgarian, there are differences between Thrace (where I live) and all other regions such as "Shopskiya Kray", Northern-West, Northern-East, in different villages. There's one prominent dialect in Rhodopes/Smolyan which is almost impossible to decode.

Bulgarians and macedonians don't need translators in order to communicate. Macedonian sounds like a dialect, sometimes funny, I guess it sounds the same for the other side.**

The similarities are most apparent and harder to deny if you know the grammar which is harder to change forcefully by the political authorities (than replacing or adding a few letters in the alphabet, adding lots of foreignisms). For example, among the Slavic languages, it's only Bulgarian and Macedonian (?) which have determiners (as "the", but it's in the end of the word) and are analytic, there are no cases, but prepositions - as in German and Latin languages.


Bulgarian Swadesh list in NLTK

I don't know who's responsible for the translations and making them official.

However, IMO:

1. Где - as of now it's a kind of archaism for къде . 

  "Онуй" for "that" is a dialect? and funny form of онова   .
It's used in the expressive phrase "туй-онуй" and is funny because "уй" sounds like "хуй". That word is the same as the one in the famous greeting in Russian "Пошол на хуй". :)
2. "Бастун" is given for stick, however this is a foreignism and a specific instance/application of the general term "пръчка" ("прачка" in Maced.), "палка" is also used for some specific "sticks" ("palka" is given for other Slavic languages). Edit+: "шосе"  is also strange instead of път.

3. For other words there are also many translation.

баща/татко  (in Macedonian - татко, I don't know what's the frequency there or how they've decided. 

4. 'To swell" in Bulgarian is actively used in many variations which match the roots in other Slavic languages as given in the list:

"подувам се" is given as a translation, but it's also:

1. подпухвам  (пухнуць и др.)
2. набъбвам  (набабрува, мак.)
3. отичам ( otékat, oteći и др.)
4. бухвам (набухати, беларуски?)

4. "Овошка" for "fruit" is another funny item, instead of the more common and ordinary word "плод".
"Овошка" is appropriate for a "fruit tree", short for "овощно дърво" such as an apple tree etc.

Etc., that's what I spotted at a first glance.

I think there should be many words for each entry and more structural data.

How to use it? Install NLTK:

python -m pip install nltk

import nltk

Find the corpus, download it. Then:

from nltk.corpora import swadesh as sw


Here they show how to get translations between the language pairs:

* For instance, I think that "mama" is produced by the first random sound-producing attempts of the baby, "ma" is maybe the simplest syllable, "mama" is more reliable than "ma", because it's repeated (suggests it's a pattern) and yet two repetitons are enough. The baby called their mother "mama".


**E.g. the Macedonian word for "boyfrend" is "дечко", which in Bulgarian means a little boy or an infantile adolescent/teenager boy, somebody acting inappropriately as a younger child.

"Барам" in Macedonian means "to search" in the search engines, while in Bulgarian it's an expressive word, a stem/root for "to touch" as:

- Обарам/обарвам, набарам/набарвам, прибара/прибарвам, да се барна

However it actually has an old/deducible meaning of "search" by a gradual explorative touching.

"Обарвам" can be applied for "to search (somebody)", because it involves touching, but it's also to erotically touch somebody.


Extending the line, there are a lot of funny false-cognates in the Bulgarian-Serbian relation.

"Сине" in Serbian is an address to a "daughter", while in Bulgarian it is an address to a... "son". (However there are Bulgarian dialects where "син" means "daughter", too).

Read More

Thursday, June 22, 2017

// // Leave a Comment

Insights from the Blue Brain's latest paper on Cliques of Neurons bound into cavities ...

Cliques of Neurons Bound into Cavities Provide a Missing Link between Structure and Function

Front. Comput. Neurosci., 12 June 2017 |

I think the paper is a good food for thought, it directs for some advanced/interesting areas of mathematics such as Algebraic Topology.

Concepts and ideas to think about:
DAG/graphs, (directed) cliques, simplex/simplices, directed simplical complexes, sink/source (also vector calculus and gradients), cavities, dynamics of the above.

Betti number, n-th B.n., cavity formation

nb - n-th Betty number, ... number of n-dim cavities

Sink - last neuron, Source - first neuron. 

locally - simplices 
            cavities - globally 

directed flag complex
 shape - topology

Simplex ~ triangle in any dimension. Maximal simpl.Dim. ~ number of linearly-independent vectors?; not part of a higher dim. simpl.

Topology - connectivity, homology

formation-disintegration  -- rise of dimensionality, then reduction
random matrix
structural m.
transmission-response matrix
pre-, post- synaptic spike , period 10, 20 ms ...

cavity - homology class, fully enclosed by directed simpl. - dim.

Read More

Monday, May 29, 2017

// // Leave a Comment

Cherry Jam's debut music video - "On the Cherry Tree" | Blues-Rock-Jazz Comedy Song

Camera operator and editing - Todor Arnaudov, the author of maybe the first University course in Artificial General Intelligence in the world*, taught at Plovdiv University in 2010... ;)

More serious stuff to come... LOL

The video was edited mainly with "Twenkid FX Studio"** and processed also with a set of free software tools: FFmpeg, VirtualDub, Audacity.

The lyrics is in Bulgarian.

* Not counting the summer school after the Xiamen's AGI conference in 2009

** Twenkid FX Studio is an in-house system

Read More

Thursday, March 30, 2017


Телевизията и детето, Вирджилиу Георге и забележки от Тодор Арнаудов - чернова -- Television Damages the Brain

Добавка от 2018 г.: Публицистична статия от Тодор във в-к "Пловдивски университет" - "Бягство от прекрасния видео свят":

Обобщение, съдържание и кратки извадки от книгата "Телевизията и детето" на биофизика Вирджилиу Георге, изд. Омофор 2010, и забележки от Тодор Арнаудов, чернова от 28/11/2016 г.

Телевизията и детето

Задължително четиво за родители!

Какво причинява "спокойствието", което печелите като оставите децата да се хипнотизират от телевизора.

Бебе, хипнотизирано със смартфон

Телевизията и детето


До двегодишна възраст - да не се гледа ТВ
След това, до края на училищната възраст - до два часа на ден видео изображения

До 5-6-години - далече от компютри и видеотехнологии

- нервно напрежение
- по-трудно съсредоточаване и по-слаба памет
- апатия, скука, депресия, тревожност
- личностни смущения

Стресът - може да разрушава клетки от челните дялове на кората на мозъка


Днешните деца са все по-трудни
Сериозни нервни и психически проблеми

№3 Приспиване

Неокортекс, кората на мозъка, електроенцефалограма - ЕЕГ: бета вълни - бодърстване, алфа - пасивност, отпуснатост, хипноза, тета - сън

бета - 30 Хц, алфа - 13 Хц, тета - 8 Хц

Колкото по-дълго време пред телевизора - по-бавни мозъчни вълни - състояние подобно на сън

Алфа - "извън пространството", без посока
Бета - зрително съсредоточоване и насочване към цел извън себе си

-- телевизията отслабва вниманието, вместо да го усилва

№4 Потискане на дейността на лявото полукълбо

- Телевизията вцепенява лявото полукълбо - критична мисъл, анализ, организиране
- Дясното полукълбо върши цялата познавателна работа - обработва холистично, мисли емоционално, а не разсъждава. Не обработва рационално, дясното п. не е оперативно.

Хората рядко разбират онова, което гледат по телевизията.

"от 2700 изследвания 90% са разбрали погрешно това, което са гледали по телевизията само преди няколко минути"


Прекъсване на връзката между лявото и дясното полукълбо (мазолестото тяло)

Неосъзнаване на информацията


Умствена пасивност, възприемане на огромно количество информация

Сериозно засягане на развитието и функционирането на челните дялове на мозъка

Дейността на мозъчната кора при гледане на телевизия е коренно променена (сравнено с работата й в естествена среда)


Структурни изменения на мозъка

- Способността за концентрация на учениците намалява. Четене, писане и умение за общуване - в упадък, "даже и в най-добрите среди".

- ... "Сериозни пропуски в абстрактното и логическо мислене...." (NAEP, USA)

- Резултатите по математика при студенти са обезсърчителни, когато решението изисква повече от един етап. Едва 44% от гимназистите могат да пресмет колко трябва да им върнат, от 3 долара, с които са поръчали два вида храна..." (NAEP, USA)

- Едва 20% от младежите на 20-год. в. - да напишат молба за работа, 4% разбират разписанието на автобусите, 12% могат да подредят 6 обикновени дроби по големина. Само 20-25% от днешните ученици могат ефективно да се обучават по традиционните методи.   (Алберт Шенкър, президент на Американската федерация на преподавателите)

- Дори и в най-добрите колежи

- Снижаване на нивото на задължителни умения за четене, писане, аналитично мислене

- Въпреки положените усилия, учениците не показват особено развитие на висшата мисловна дейност

№7 Какво се случва? Защо?

- Не издържа тезата, че децата не желаят да учат, защото много от тях посещават специални курсове за наваксване и повишаване на успеха

Какво се случва с детския и младежкия мозък в медиатизираното общество?
Каква е ролята на гледането на телевизия и компютър?

№8 Ролята на средата

- Стимули, взаимодействие, изследване - въздействие върху действителността; докосване, игра; размишление, въображение
- Критичен период за развитие на мозъка
- Спокойна среда - вътрешни мисловни процеси и вътрешен диалог
- Слово - общуване с родителите

- Силни, резки шумове и стимули - страх, могат да създадат условен рефлекст на отбягване.

* Децата са емоционално по-слаби и чувствата им въздействат по-драматично, затова имат нужда от повече любов и внимателно отношение

(Виж Класическо и оперантно обучение, виж бехивиоризъм, ...)

- Нормалният мозък сам се стимулира, чрез активно взаимодействие с онова, което намира за интересно в средата.

- Личен опит

- Родителите, майка, насочване, постепенно внимателно обучение

№9 Светът на телевизията като среда на опита


№10  Телевизия и формиране на пасивна умствена функция

№11 Въздействие на телевизията върху развитието на лявото полукълбо

№12 Телевизия и интелектуални постижения


№13 Умствени увреждания

"Learning disabilities", LD

№14 Четене

№15 Учене и телевизия

№16 Въздействие на телевизията на подсъзнателно ниво

№17 Проблеми с вниманието

с.64, АДХД

№18 Телевизорът и насоченото внимание

№19 Телевизия и прояви на хиперактивност, раздразнителност и безсъние

№20 Телевизионна епилепсия

№21 Език и усвояване на езика

№22 Проблеми с вниманието и учението. Предният мозък и телевизията

№23 Екранът и развитието на предния мозък


№24  Хипнотизиращото действие на телевизията


№25 Телевизионна зависимост

№26 Справочната функция на масмедиите, възпитателният ефект и спиралата на мълчанието

№27 Образование чрез телевизия


- Оспорване на авторитета

X - Телевизията отглежда бунтарско поведение против всяка норма, принципи и авторитет  ...
...за да стане лесна плячка на консуматорска пропаганда и манипулиране
* противоречие
(и няма цитирани конкретни научни изследвания)
За манипулиране- да, но в същото време го постига и подхранва чрез възпитание в кланяне на авторитети - знаменитости, актьори, певци, велики личности, политици.

Телевизията става авторитет, тя е "Господ", виж "Какво му трябва на човек..." от РАЗУМИР, 2014 г.

Телевизията е гласът на най-мощната пропаганда на държавата и властните хора и заедно с хипнотичното действие внушава и води хората да мислят по един и същи начин, а не да се бунтуват. (В някои случаи може да ги манипулират да се бунтуват.)

Консуматорската култура дресира чрез телевизията, американската рекламна индустрия е в разцвет от 50-те години, когато се появява и телевизията.
№28 Еротика и сексуалност


№29 Отъждествяване с герои от малкия екран


* вж по-горе за противоречието - кланяне на авторитети (герои), които са създадени от присъствието им на екрана

№30 Телевизията и отношението към сексуалността


№31 Порнографска зависимост

-  Мъже, които гледат порнография приемали с лекота насилието над жените и тяхното унижаване


*1 Терминът "зависимост" е оспорен през 2016 г.?, че хората които гледат и харесват повече порно просто имат по-високо либидо, нямат партньор и т.н.

*2 Оспорвам това твърдение, определението "с лекота" и дефиницията на "унижаване". Сексът между двама искащи да го правят не е "унижение" и отношението между двамата, които изглеждат "груби", ако не са насила, също не са "унижение", а част от играта на половия акт.

*3 Сексуалните маниаци може да гледат порно, но извращенията им не идват от това, че гледат порно, защото всички младежи и много млади мъже гледат порно и не са нито психопати, нито изнасилват жени. Според други твърдения и спекулации, порното води до намаляване на сексуалното насилие над жени - броя на изнасилванията, - защото предоставя алтернатива.

* Авторът на книгата е също така богослов.

№32  Психологически и физиологични причини на зависимостта от еротиката


№33 Еротика, брак, семейство. Изводи

с.137, с.139


Телевизионното насилие и агресията в реалния свят


№35 Урокът на телевизионното насилие

№36 Семейният живот и телевизията

№37 Децата, младежите и телевизията в контекста на нихилистичната култура

№38 Нихилизмът и идентичността на "новия човек"

№39 Портрет на телевизионното дете

№40  Превенция


- Насърчаване на мисленето
- Упражняване на вниманието
- Контрол на поведението

№40 Изводи

№41 Приложение I. Компютърът, видеоигрите и интернет


№42 Приложение II. Ръководство за родители и възпитатели

№43 Фактори за мозъчното развитие на детето

№44 Телевизионният свят като опитна среда

№45 Дейността на мозъчната кора по време на телевизионното действие

№46 Последици от телевизионното гледане върху мисловния живот на човека
Read More

Thursday, February 9, 2017

// // Leave a Comment

Funny Deep Learning Image Captioning - Sentence Generation

a baby laying on a bed with a stuffed bear

a bowl of fruit is sitting on a table

a group of elephants walking across a dirt road

a cat sitting on a window sill looking out the window

a bunch of bananas are on a table

a man sitting on a couch with a laptop and a laptop

a man is playing tennis on a tennis court

a woman sitting on a couch with a laptop

a woman is standing in front of a store

Too many of them.

It seems the authors like skateboard, frisbees, giraffes and bananas.
Cats and dogs are common.

It's disappointing how the pattern of word-mappings gets transparent.

Schemes like:

"on the couch"
"on the table"
"down the slope"
"on the rode"

Also "with a * and a **" (thus "with a laptop and a laptop")

Read More