Introduction

Das Experiment

To whom it may concern

Fast and Frugal Detection of Chiastic Protofigures
in English Subsection of CHILDES Corpus
regex strikes back

Daniel Devatman Hromada123
daniel@wizzion.com
1 Université Paris 8 / Lumières
École Doctorale Cognition, Langage, Interaction
Laboratoire Cognition Humaine et Artificielle
2 Slovak University of Technology
Faculty of Electronic Engineering and Informatics
Department of Robotics and Cybernetics
3 Universität der Künste
Fakultät der Gestaltung, Berlin

Introduction

Das Experiment

Table of Contents

1

Introduction
Computational Psycholinguistics
Computational Rhetorics
Main idea

2

Das Experiment

3

To whom it may concern

To whom it may concern

Introduction

Das Experiment

To whom it may concern

Computational (Developmental) Psycholinguistics
C(D)P
Is a cross-over between computational linguistic (and/or Natural
Language Processing) and developmental psycholinguistics.
Main objectives:
1

use computational methods (data-mining, information
retrieval, NLP etc.) to gain novel insights about ontogeny of
language competence in human children

2

develop computational models of language acquisition and
embed them into language-interacting artificial agents

In this talk we focus solely on the first objective.

Introduction

Das Experiment

To whom it may concern

CHILDES

CHILDES corpus: a gem of gems
Child Language Data Exchange System (MacWhinney&Snow, 1985)
http://childes.psy.cmu.edu/data
http://wizzion.com/CHILDES/ (mirror from 6th Feb 2016)

1

more than 50 years of tradition

2

more than 1.5 GigaBytes of mostly textual data contained in
cca 30000 transcripts

3

at least 26 languages, dialects or language combinations

4

Creative Commons BY-NC-SA licence

Introduction

Das Experiment

To whom it may concern

CHAT format
CHAT system provides a standardized format for producing computerized
transcripts of face-to-face conversational interactions. (MacWhinney,
2016; http://childes.talkbank.org/manuals/chat.pdf).

@Languages:
eng
@Participants: CHI Eve Target_Child , MOT Sue Mother , FAT David Father
@ID:
eng|Brown|CHI|1;6.|female|||Target_Child|||
@ID:
eng|Brown|MOT|||||Mother|||
@ID:
eng|Brown|COL|||||Investigator|||
@Date: 29-OCT-1962
*MOT:
one two three four .
%mor:
det:num|one det:num|two det:num|three det:num|four .
%act:
tests tape recorder
*CHI:
one two three . [+ IMIT]
A non-negligeable advantage
Majority of transcripts follow the principle: ONE LINE = ONE UTTERANCE.

Introduction

Das Experiment

To whom it may concern

Computational Rhetorics

Computational (& Cognitive) Rhetorics
Computational Rhetorics
A discipline which has attained its maturity at Computational
Rhetorics Workshop organized by Harris and Di Marco at
University of Waterloo.
Computational-Cognitive Rhetorics
A disciplne using computers to better understand why rhetorics
casts such a powerful curse on human minds.
Computational-Developmental Rhetorics
Using computers to elucidate the process of ontogeny of rhetoric
competence in human children.
”Child’s spontaneous remark is more valuable than all questioning
in the world.” (Jean Piaget)

Introduction

Das Experiment

To whom it may concern

Main idea

Main concept(s)
Scheme
A scheme is a generic form which corresponds to one or more
distinct constellations of observables.
Regular expression
A sequence of characters that defines a search pattern.
Perl-Compatible Regular Expressions
Concise and expressive regex standard. Much more powerful than
regular grammars: it is possible to perform back-tracking!
Backtracking
Allows us to match that, which has already been matched: paves
the way to detection of repetitions.

Introduction

Das Experiment

To whom it may concern

Main idea

Main idea(s)

Main idea
Chiasms are repetition-based schemata A1 B1 C1 XC2 B2 A2 (or
A1 B1 XB2 A2 ).

Note that the presence of middle term (B) and separator term (X)
can be considered as facultative. But in order to detect chiasm,
the initial preceptor (A1 ) has to be strongly reminiscent (and
ideally identic) to terminal successor (A2 ). Idem for relation
between terminal preceptor (C2 ) and initial successor (C1 ).

Introduction

Das Experiment

Table of Contents

1

Introduction

2

Das Experiment
Method
Results

3

To whom it may concern

To whom it may concern

Introduction

Das Experiment

To whom it may concern

Method

Regex implementing the main idea
initial
preceptor
(\w{3,})

terminal
successor
(.{0,77})

(\w{3,})

terminal
preceptor

.{0 77}

\3

\2

\1

initial
successor

Note that nodes of a chiasmatic structure form a double-closed
graph.

Introduction

Das Experiment

To whom it may concern

Method

Demo
Run this shell command* :
grep -irP ’^\*MOT:.*(\w{3,}) (.{0,77}) (?!\1)(\w{3,}).{0,77}\3 \2 \1’ *Eng*

in the directory into which You downloaded and unpacked the
CHILDES corpus.
Note that the extractor can be parametrized with change of numeric values: e.g.
changing (\w{3,}) to (\w{1,}) could potentially allow You to detect grapheme-level
metatheses like ”asteriks with an asterisk”.

* Regex sequence is hereby transfered to Public Domain under Creative Commons
BY-NC-SA (Author Attribution, Non-Commercial, Share-Alike) licence.

Introduction

Das Experiment

To whom it may concern

Results

You’ll see many playful ones...

pear pear yummy yummy yummy yummy pear .
my name is Joey Joey Joe Joe Joe Joe Joey .
I think I can I think I can I think I can I think I can I think I
can I think I can .
tick tick tick tick tick tick tick tick tick tock tick tock tick
tick tick tick .
Earth , moon , Earth , moon , full moon , Earth moon .
crash , boom , crash , boom , crash , boom crash !
Note: triplicated couple A1 B1 A2 B2 A3 B3 always contains an
A1 B1 B2 A3 implicit antimetabole!!!

Introduction

Das Experiment

To whom it may concern

Results

...reversed coordinatives...
and they splish and they splash and they splash and they
splish .
a dot and a dash and a dash and a dot .
well Granddad and Grandma [//] Grandma and Granddad are
coming today .
it’s called lamb and vegetable [//] mediterranean vegetable
and lamb risotto .
Donald hopped and swam and swam and hopped until he was
safe on dry ground .
every day my cows Poppy (.) Annabel (.) Emily and Heather
moo and mumble (.) mumble and moo .
Chester and Wilson Wilson and Chester .

Introduction

Das Experiment

To whom it may concern

Results

...and more exhaustive reversed lists...

Chester and Wilson and Lily Lily and Wilson and Chester .
okay , square , square , rectangle , square , oval , two , one ,
one , two .
blue , green , yellow , red , red , yellow , green , blue .
one two three or three two one ?
sure we went through Rhode island , Massachusetts , New
Hampshire , Vermont , and then on the way back we did
Vermont , New Hampshire , Massachusetts , Rhode island ,
right ?

Introduction

Das Experiment

To whom it may concern

Results

...and reversals of direction and position and time...
you get one ticket that says York to Manchester and another
ticket that says Manchester to York .
he used to rush here and there and there and here and back
again all the time and of course he was always in such a rush
that he never ever finished anything properly .
from here to there , from there to here from here to there
funny things everywhere .
let’s put mine on yours and put yours on mine .
could put the box on the lid instead of the lid on the box .
but I mean do you get your drink after you’ve had your biscuit
or do you get your biscuit after you’ve had your drink .

Introduction

Das Experiment

To whom it may concern

Results

...and reversals of attributes...
let’s put the blue one on the guy with the red underpants and
the red one on the guy with the blue underpants .
if it (h)as been a police car it becomes a racing car and if it
(h)as been a racing car it becomes a police car .
and when you’re talking about little crocodiles and big snakes
(.) or little snakes and big crocodiles (.) they’re jelly sweets
you’ve had in the past .
oh [!] I got a yellow cup and a red plate and you got a red
cup and a yellow [!] plate (.) .
look , they’re very similar (.) look , this one is green with a
little yellow , and this I yellow with a little green (.)
interesting , huh ?
you mean it looks nicer than it smells [//] smells nicer than it
looks .

Introduction

Das Experiment

To whom it may concern

Results

...and reversals of case-like roles, of course...

Nominative vs. Vocative
Amanda that’s xxx xxx that’s Amanda .
xxx this is Stephanie Stephanie this is xxx by the way .
Nominative vs. Accusative
froggie keep an eye on mummy or mummy keep an eye on
froggie ?
Floppy meet the screwdrivers screwdrivers meet the Floppy .
Nominative vs. Dative
do you give Daddy a big kiss or does Daddy give you a big
kiss ?

Introduction

Das Experiment

To whom it may concern

Results

...as well as some more complex swaps?
like Nominative vs. Genitive vs. Locative...
I mean you go [//] girls go to boys parties and boys go to girls
...or proto-rhetoric questions...
I think you’re stinky you are stinky are you stinky ?
wouldjou [: would you] couldjou [: could you] wouldjou [:
would you] with a goat ?
...and other pieces of maternal wisdom.
I would not could not in a box I could not would not with a
fox .
we’re in house of bricks not the bricks of house .
two for tea , and tea for two .
I meant what I said and I said what I meant .

Introduction

Das Experiment

Table of Contents

1

Introduction

2

Das Experiment

3

To whom it may concern
Current state
Future directions

To whom it may concern

Introduction

Das Experiment

To whom it may concern

Current state

Concerning the method
a naive rhetoric-figure-tagger (nRFT)
fast*, deterministic, transparent for inspection, partially
parametrizable
form-oriented: looks for identic sequences within the signifier
(no semantics involved)
generates false positives: manual check needed; can be useful
for CHIASMFP corpus
can speed-up the manual annotation (semi-supervised
scenario)
IMPORTANT: the schema can be used not only to detect,
but also to GENERATE
* and super-fast if You store Your Big Data on a RAMdisk or at
least on a SSD disk cache

Introduction

Das Experiment

To whom it may concern

Current state

Concerning the results
English motherese utterances tend to abound with
protochiastic structures
many functions: playful reversal of repetition, reversal of
spatial direction, reversal of list, lapsus lingui correction,
positional swap, attribute swap, functional (case) swap ... all
matched by a single one-liner !
what we are dealing here with is a whole ecosystem of
diverse structures
indicated prominence of the verb ”put” as a middle term
consistent with theories of Piaget and Tomasello
triplicated couple A1 B1 A2 B2 A3 B3 always contains an
A1 B1 B2 A3 implicit antimetabole

Introduction

Das Experiment

To whom it may concern

Future directions

Invitation to explore
not only intralocutory (i.e. within 1 utterance) chiasms, but
also translocutory ones (within multiple successive utterances)
relations to variation sets and Winograd schemata
multi-lingual analysis (are these beasts universal ?)
ontogenetic relation to other figures like rhetoric question or
even metaphore (METAPHOROS = ”carry over”)
informational content of chiasms (known components +
unknown order = maximal amount of new info ?)
neurocognitive aspects of chiasm processing (focus upon the
cyclical referential closure between initial and terminal
token of the sequence)
neurorhetoric hypothesis: look for a P600-like evoked
potential following the exposure to chiasmus
non-linguistic chiasmata (musical, visual, spatial, anatomical,
social, moral, emotional, sexual, spiritual etc.)

Introduction

Das Experiment

To whom it may concern

Future directions

Conclusion

Starting discussion with conclusion often concludes the discussion...

Ergo, no ultimate conclusion without juicy discussion.

daniel@wizzion.com thanks Thee for Thy attention