Upload
keith-stallworth
View
252
Download
1
Tags:
Embed Size (px)
Citation preview
SCIENTEXT: A Corpus of French & English Academic & Scientific Texts
Alice Henderson (for the Scientext team)
LLS research group– Université de Savoie
Chambéry, France
BAAL, University of Newcastlem September 3-5, 2009
English
academic
FrenchEcrits
universitaires ss writing textbooks
Ecrits de recherche
articles theses
= scientifique
OutlineOutlineGeneral overview of the
Scientext projectEnd product & applicationsGoals of the linguistic studyDetails of the corpus & taggingPresentation of the beta version
General OverviewGeneral Overview Project financed by the French ANR CORPUS ET
OUTILS DE LA RECHERCHE EN SCIENCES HUMAINES ET SOCIALES (2007-2010).
Goals:Create a freely-available corpus of scientific &
academic writing in French & English Devise tools for studying linguistic markers of
stance/positioning AND reasoning Intended Users: Linguists, epistemologists,
information retrieval specialists, scientists, language teachers.
Long-Term Applications:◦ L1 & FL/L2 teaching ◦ Lexicography & writing aids◦ Information retrieval in scientific & technical fields
General OverviewGeneral Overview Draws on several branches of linguistics:
◦ Corpus linguistics: creation & study of a large corpus of scientific & academic texts
◦ Natural Language Processing: processing & study of a corpus using a syntactic dependency parser (Bourigault’s Syntex).
◦ Traditional branches of linguistics: discourse analysis, lexicology, enunciation, syntax and semantics
Projet coordinatedby LIDILEM research group (F. Grossmann, A. Tutin), 3 teams = multidisciplinary◦ LIDILEM (Grenoble) : F. Grossmann, A. Tutin, F. Boch, C.
Cavalla, O. Kraif, M. Florez, I. Novakova, M.L. Nguyen, F. Rinck.
◦ LLS (Chambéry) : J. Osborne, A. Henderson, R. Barr.◦ LiCorN (Lorient) : G. Williams, H. Maury, C. Ropers.
General OverviewGeneral Overview
End Product & End Product & ApplicationsApplications1) Web site with several ways of
selecting sub-parts of texts.◦ Query search (complex & simple) and text
view◦ Search for traces of stance/positioning and
reasoning using local pre-established grammars
◦ Downloading of XML corpus (for authors who gave permission, Creative Commons)
◦ Downloading of search results (zip format, CSV format for statistics)
End Product & End Product & ApplicationsApplications
1) Website allowing selection of sub-parts of texts
2) Teaching applications for both L1 and L2 learners: research into university writing, second language production, etc.
3) Lexicographical applications including assistance with encoding strategies using reference corpora.
4) Targeted information retrieval in scientific and technical fields.
The Linguistic StudyThe Linguistic StudyFocus on 2 essential features of the
texts:◦ Authors use stance to situate themselves in
relation to previous and contemporary research whilst demonstrating what is specific to their work and the choices made.
◦ The intellectual process upon which findings and deductions are based can be revealed via the analysis of authorial reasoning.
Test two hypotheses:◦ Stance is expressed by a phraseology that
is shared (partly? largely?) across fields◦ This phraseology is more characteristic of
genres than of fields
The Linguistic StudyThe Linguistic StudyDistinguish between 3 main
parameters:Field, Text genre (and sub-genres), Text
sectionScientific sub-genres
Scientific articles Conference proceedings PhD theses, HDR
Academic sub-genres (Learner corpus)
2nd year English majors, Long Essays 3rd year English majors, Language
Policy analyses
Details: French scientific Details: French scientific corpuscorpus
Articles & presentation
s
Theses HDRs
Social sciences 154 32 8
Linguistics 66 8 4
Education 63 8 2
NLP 8 8 1
Psychology 17 8 1
Natural sciences
21 10 0
Biology 6 7 0
Medicine 15 3 0
Applied sciences
0 8 1
Electronics 0 4 0
Mechanical engineering
0 4 1
TOTAL 175 50 9
234 texts (1997-2008), 5 million words
Details: English corporaDetails: English corporaAcademic (learner) corpus
(Chambery,1997-2007)1.1 million words, 300 texts, 4000-5000 words
long Scientific corpus (Lorient, g
[email protected]) 33 million words “hoovered” from BMC Corpus
of Biology and Medical Texts POS & lemmatised Theoretical analysis of meaning transfers for
the analysis of diachronic & synchronic meaning changes in context through collocational resonance
Creation of a bottom-up dictionary of verb patterns with corpus-driven thematic and conceptual groupings for NNS scientists
Corpus Tagging Corpus Tagging (French sci. + Eng (French sci. + Eng academic/learner)academic/learner)
XML format (Text Encoding Initiative)Tagged elements
◦ Header: Type of tagging, information about the text, availability
of the text
◦ Text Structure (semi-automatic tagging): Identification of text sections: abstract, introduction, body
of the text, conclusion, notes, references. Lay-out (when available): bold, italics, structure of lists
◦ Linguistic Tagging (automatic): Morpho-syntactic tagging & identification of syntactic
dependencies (Bourigault’s Syntex – 2007 version)
OutlineOutlineGeneral overview of the
Scientext projectEnd product & applicationsGoals of the linguistic studyDetails of the corpus & taggingPresentation of the beta version
Presentation of the beta Presentation of the beta versionversion
Web site available on-line: http://scientext.dynalias.net
Interface created by Achille Falaise, using the query language Concquest developed by Olivier Kraif (Université Grenoble 3)
Step1 : Choosing the field, genre, & text Step1 : Choosing the field, genre, & text section (French scientific corpus)section (French scientific corpus)
Step 2 : Searching in the textsStep 2 : Searching in the texts
3 search modes◦ Simple interface, with scroll-menus and
predefined values◦ Complex query language, so grammars
can be created/written◦ Local grammars, involving
stance/positioning or reasoning Example: grammar of scientific affiliation
Example of a simple Example of a simple queryquerySelection of predicate adjectives used with the
noun policy.
Examples of predefined Examples of predefined searchessearches
Verbs of feeling: hate, love, feel, like, …
Verbs of opinion: consider, think find, …
Evaluative adjectives: true, great, important, best, new, right, …
Example of a complex query Example of a complex query (advanced search)(advanced search)
1. Search for syntactic dependency + co-occurrence
<hypothèse,#1><>*<cat=V,#2> :: (SUJ,#2,#1);
Verbs which come after the lemma hypothèse, where hypothèse is the subject of the verb.
2. Search for a disjunction of lemmas + syntactic dependency
<lemma=/(hypothèse|notion|concept)/,#1> && <cat=V,#2> && <cat=A,#3> :: (SUJ,#2,#1) AND (ADJ,#1,#3) ;
The lemmas hypothèse, notion or concept functioning as subjects & accompanied by an adjective
Example of a local grammar Example of a local grammar (to write an advanced (to write an advanced
search)search)◦Using variables ◦Re-defining a relation
Ex : (ATTSUJ,#2,#1) = (ATTS,#3,#1) AND (SUJ,#3,#2)
Step Step 3 : Display3 : DisplayKWIC display, can be customised
Displaying a wider contextDisplaying a wider contextDisplay of a wider context
Displaying syntactic Displaying syntactic structurestructure
Displaying syntactic Displaying syntactic structurestructure
Statistical calculations & Statistical calculations & displaydisplay of resultsof results
Graphic display ofGraphic display of results: results: “we”“we”
2nd Year: 728 occs. 3rd Year: 1607 occs.
ConclusionConclusion
Av ailable on S c ientext Web s ite
F rench scientificcorpus , 5 million
E ng lishacademic/learnercorpus ,1.1 million
ConclusionConclusion
Project still running (through early 2010)◦ Constitution of corpus & tagging : LONG … & fastidious◦ Interface still being developed◦ Linguistic model still needs finalising ◦ More grammars need to be developed◦ Teaching materials need developing & piloting
Issues: interface between lexis & rhetorical functions
Future Research◦ Linguistic study of markers :
“positioned” citations markers of scientific affiliation
◦ Teaching materials need piloting & evaluating
Thank you!!
(and please try it out)
Publications & resources Publications & resources linked to Scientext projectlinked to Scientext project
Boch F., Grossmann F. (2002). “Se référer au discours d’autrui : quelques éléments de comparaison entre experts et néophytes”. L’écrit dans l’enseignement supérieur. Enjeux,:Brussels, pp41-51.
Boch F., Grossmann F. , Rinck (2007). “Conformément à nos attentes ...” ou l’étude des marqueurs de convergence/divergence dans l’article scientifique”. Revue Française de Linguistique Appliquée. Voll. XII-2, pp109-122.
Bourigault D. (2007). SYNTEX, analyseur syntaxique opérationnel. Mémoire d’habilitation à diriger des recherches, Université Toulouse Le Mirail.
Chavez I. (2008). La démarcation dans les écrits scientifiques - Les collocations transdisciplinaires comme aide à l’écrit universitaire auprès des étudiants étrangers, Mémoire de Master 2 Français Langue Etrangère Recherche, C. Cavalla (supervisor), Université Stendhal-Grenoble3: Grenoble.
Garcia P.P. (2008). Etude des marques de la filiation dans les écrits scientifiques. Master 1 thesis, Université Stendhal-Grenoble3: , F. Grossmann and A. Tutin (supervisors).
Grossmann F., Tutin A. (2008). “Evidential Markers in French Scientific Writing: the Case ofthe French Verb voir. Evidentiality Workshop, Bamberg, 27-29 February 2008.
Henderson, A . & R. Barr (2009), “Corpus-based L2 Writing Instruction : Raising Awareness of Authorial Stance”, Journal of Writing Research, (forthcoming).
Rinck, F. (2006). L’article de recherche en Sciences du Langage et en Lettres, Figure de l’auteur et approche disciplinaire du genre. Doctoral thesis, Sciences du Langage, F. Boch and F. Grossmann (supervisors), Université Stendhal-Grenoble3: Grenoble.
Rinck, F., Boch, F., Grossmann, F. (2007). “Quelques lieux de variation du positionnement énonciatif dans l’article de recherche”, in Lambert P., Millet A., Rispail, M.Trimaille C. (eds). Variations au coeur et aux marges de la sociolinguistique. L’Harmattan, Espaces Discursifs, Paris.
Tutin A. (2008). “Evaluative adjectives in academic writings”. Interpersonality in written academic discourse: perspectives across languages and cultures, 11-13 December, Zaragoza, Spain.
Tutin A. (2007a) (ed) “Lexique et écrits scientifiques”. Revue Française de Linguistique Appliquée, volume XII-2, December 2007.
Tutin, A. (2007b). “Modélisation linguistique et annotation des collocations : application au lexique transdisciplinaire des écrits scientifiques”, in S. Koeva, D. Maurel, M. Silberztein (eds). Formaliser les langues avec l’ordinateur. Presses universitaires de Franche-Comté: Besançon.
Williams G & Millon C. (2009). “The General and the Specific: Collocational resonance of scientific language”. Proceedings Corpus Linguistics 2009, University of Liverpool. (forthcoming)
Williams G & Millon C. (2009.) “Les verbes et la science: la construction d’un dictionnaire organique”. Actes des Journées de la Linguistique de Corpus 2009. Texte et Corpus. (forthcoming)
Williams G. (2008). “Le Corpus et le dictionnaire dans les langues de spécialité”, in Maniez et al (eds). Corpus et dictionnaires de langues de spécialité. Presses Universitaires de Grenoble, pp 135-151.
Wiliams G. (2008) “Verbs of Science and the Learner’s Dictionary”. Proceedings Thirteenth EURALEX International CongressBarcelona, Spain. 15-19 July 2008, pp797-806.