Analyzing bursts of written language from a product perspective Etudier les “bursts” de production écrite à partir d’une analyse des produits Thierry Olive (CNRS) & Georgeta Cislaru (Université Paris 3)
Writing(s) at the crossroads: the process/product interface Paris, 20-21 juin 2013
Bursts of writing
What is a burst of writing? Pauses
Bursts/execution periods
Bursts of writing
What is a burst of writing? Pauses
Bursts/execution periods
Bursts of writing
A burst of writing is a: Transcribing activity between two consecutive pauses (in which at least one word is written). Execution periods Transcription periods
Bursts of writing
Bursts, less studied than pauses Even if 50% of writing time devoted to execution Until recently, pause were considered as the locus of writing processes
But demonstrations of writing processes during bursts Children compose sentences clause by clause while adults prepare clause n+1 when handwriting clause n (Chanquoy, Foulin & Fayol, 1990)
Bursts of writing
Bursts, less studied than pauses Even if 50% of writing time devoted to execution Until recently, pause were considered as the locus of writing processes
But demonstrations of writing processes during bursts Children compose sentences clause by clause while adults prepare clause n+1 when handwriting clause n (Chanquoy, Foulin & Fayol, 1990) Children “think-and-then-write” and adults” think-while-they-write” when composing a text; this shift in strategy occurs with automaticity of handwriting (Olive & Kellogg, 2002)
Bursts of writing
Bursts, less studied than pauses Even if 50% of writing time devoted to execution Until recently, pause were considered as the locus of writing processes
But demonstrations of writing processes during bursts Children compose sentences clause by clause while adults prepare clause n+1 when handwriting clause n (Chanquoy, Foulin & Fayol, 1990) Children “think-and-then-write” and adults” think-while-they-write” when composing a text; this shift in strategy occurs with automaticity of handwriting (Olive & Kellogg, 2002) All writing processes, i.e., planning, translating and reviewing, can be activated during transcription
Bursts of writing Translating mainly activated during handwriting
Olive, Alves & Castro (2009)
Alves, Castro & Olive (2008) Pausa
Execução 35
40 36,2
25 20
16,5 15,1
15
16,7
%
O c u r r ê n c ia s
30
Percentage of occurrences
35
handwriting Pausing
30
25
20
15
10
10 8,2
7,3
5
5 0
0
Planeamento
Tradução
Revisão
Planning
Translating
Revising
Familiar handwriting
Planning
Translating
Revising
Unfamiliar handwriting
Typing, narrative, 2 sec. pauses expository/argumentative, 250 ms pauses
Bursts of writing
Overlap between handwriting and the high level writing processes
Since writing processes are activated during handwriting, change in processing demands may be reflected in burst properties Which writing process affects bursts? Do skills in writing processes affect bursts?
Bursts of writing
Handwriting/typing Bursts are absent in a copying task by proficient typists (Hayes & Chenoweth, 2006) Fast typists compose with more words in language bursts, and longer execution periods (Alves, Castro, Sousa, & Strรถmqvist, 2007) In children (4th grade), bursts are affected by handwriting skills (Alves, Branco, Castro & Olive, 2011) 8 25
6 5 4
Girls
3
Boys
2 1 0
Bursts size (in words)
Burts length (in seconds)
7
20 15 10 5 0
Low
Average
High
Low
Average
High
Bursts of writing
Translating Articulatory suppression reduces bursts length by 1/3 (Hayes & Chenoweth, 2007) L1 writing results in longer bursts than L2 writing, and experience in L2 increases bursts length (Chenoweth & Hayes (2001)
Bursts of writing
In sum, bursts increase with writing skills Skills in writing processes make more available cognitive resources to writers, which allows more simultaneous/parallel processes: Writers can prepare their text while transcribing Text construction continues during handwriting Less need to suspend handwriting = longer bursts
The product of bursts
The products of bursts
Main questions: Until now, research on the temporal parameters of bursts. What is the content of a burst? ‌‌‌ Or, what do we write during bursts? Can we discover some properties of content of bursts? Can we associate these properties with specific real-time parameters?
The products of bursts A project elaborated in the ANR Ecritures project Writing of social workers Keystroke logging Several drafts for one report Several writers for one report A pluridisciplinary analysis of the corpus: Linguistics Textual genetics Natural language Processing Psycholinguistics An on-going project, based on the analysis of the corpus of written text annotated with real-time date
The products of bursts
Bursts as the planning unit of writing Are bursts equivalent, or informative, on the unit of text planning? Speaking performance is organized around the clause The clause is the planning unit of speech (Ford & Holmes, 1975). Is bursts’ content predetermined by specific, defined structures/patterns? Are there some regularities in content of bursts? Clauses, larger syntactic structures‌?
Requires analyses of the products to detect the more frequent or repeated patterns
The products of bursts
Bursts and accessibility of content With high skills, the cognitive system constructs ready-made blocks of knowledge Algorithmic procedures vs. long-term memory retrieval
The products of bursts
Bursts and accessibility of content With high skills, the cognitive system constructs ready-made blocks of knowledge Algorithmic procedures vs. long-term memory retrieval Frequency effect
The products of bursts
Bursts and accessibility of content With high skills, the cognitive system constructs ready-made blocks of knowledge Algorithmic procedures vs. long-term memory retrieval Frequency effect Familiarity effect
The products of bursts
Bursts and accessibility of content With high skills, the cognitive system constructs ready-made blocks of knowledge Algorithmic procedures vs. long-term memory retrieval Frequency effect Familiarity effect Psychological routines (e.g. the alphabet): an effortless mode of action, behaviour, regularly used in daily life.
The products of bursts
Bursts and accessibility of content With high skills, the cognitive system constructs ready-made blocks of knowledge Algorithmic procedures vs. long-term memory retrieval Frequency effect Familiarity effect Psychological routines (e.g. the alphabet): an effortless mode of action, behaviour, regularly used in daily life
Some psycholinguistic/linguistic routines‌?
The products of bursts
Bursts and accessibility of content Are there some linguistics structures that are already constructed/pre-planned in LT memory and that can be retrieved and directly written out? If it is the case, then Such structures or units may be produced in one burst Burst fluency may be faster (or the preceding pause) With high skills, longer structures (or more syntactically complex) may be retrieved, this should result in longer bursts
The products of bursts
1.
From on-line data to linguistic analyses of the product of burst: detecting bursts, and then analysing their content The more frequent forms may be retrieved more easily, faster (shorter preceding pause), or written out more fluently than less frequent forms.
The products of bursts
2.
From linguistic analyses of the content of bursts to on-line data Detecting specific linguistic structures to examine whether these structures: Are produced in bursts, Are associated with specific temporal parameters reflecting easiness of retrieval. Colligations Colocations Linguistic routines Repeated segments (Salem, 1987)
The products of bursts
Repeated segments (RS): Text segments of at least two words that are recurrently associated, i.e., that occur at least two times in a text or a corpus. Repeated segments as ready-to-write units
The products of bursts
In the Ecritures project: Professional writing, with domain-specific norms, instructions, constraints‌ Involves discursive routines Textometrics analyses to discover these RS through cooccurrences Keystroke logging for recording temporal data of writing
The products of bursts
A Natural Language Processing program, developed by A. Lardilleux, for automatic: RS detection Burst extraction (from inputlog log files) Associating RS ad burst detecting RS, extracting burstbursts, an to associate)
Conclusion
A better understanding of bursts requires: A pluridisciplinary approach Examining the linguistic structures of products of bursts Challenges in the study of burst: Revision (Baaijen, Galbraith, & de Glopper, 2012) Bursts that only include text generation (P-bursts) Bursts that correspond to text revision (R-bursts)
Writers strategy (planners vs revisers) Composition strategy (e.g., with or without draft)
Thanks to: Adrien Lardilleux The members of the ANR Ecritures Project
Thierry.olive@univ-poitiers.fr Georgeta.cislaru@univ-paris3.fr