Romanian Online Dialect Atlas© 2003 Embleton, Uritescu, Wheeler1 Romanian Online Dialect Atlas An exploration into the management of high volumes of complex.

Romanian Online Dialect Atlas

© 2003 Embleton, Uritescu, Wheeler

1


An exploration into the management of high volumes of complex knowledge in the social sciences and humanities.

Sheila M. Embleton

Dorin Uritescu

Eric S. Wheeler



2

Romanian Online Dialect Atlas Sheila M. Embleton Department of Languages, Literatures and

Linguistics, York University

Dorin Uritescu co-editor of source atlas: Noul Atlas lingvistic

român. Crisana.

Department of French, Glendon College, York University

Eric S. WheelerITEC program, York University,

Managing partner, Wheeler and Young Inc.



3


Supported (2003-2006) by a grant from:

Social Sciences and Humanities Research Council (Canada)



4

Agenda

The problem of high-volume, complex data in social sciences and humanities.

Predecessor projects: English, Finnish dialect data

Use of Multidimensional Scaling (MDS) to consolidate data

Interactive, media-rich presentation



5

ProblemIn social sciences/humanities, data is

often characterized by: high volume multiple variables or dimensions no a priori model

Dialectology provides a good exemplar



6

Dialectology Explain the variations in linguistic

usage across geography Simple example:

“church” vs. “kirk” (< OE cirice) More realistic problem:

169 features in 313 locations (SED)

213 features in 400+ locations (Finnish)



7

Dialect atlases Record the details in maps Many maps needed to make an

atlas Recovery of individual facts is

possible but... Global understanding of the

situation is lost in the volume of details



8

English

Survey of English Dialects (SED) 169 features at 313 locations

Computer Developed Linguistic Atlas of English

Applied MDS to already computerized data



9

English: results

2-D map of dialect locations No geographic information used Close correspondence to

geography (as expected) Highlighted further problems of

handling and understanding high-volumes of data



10

English Dialect Map Northern counties at

top Mid and southern

counties below Somerset, Devon

(South-west) is out of place (in East)

Star-bursts, colours, dotted lines all help interpret map data



11

Finnish



12

Kettunen (1940) The Dialect Atlas of Finland 213 maps x 530 locations Up to 16 features per map Typically 1-3 features per location ~120,000 data items

Project: data computerization (largely done)

Stage II: application of MDS (not yet done)



13

Map 1 (parts)



14

•Special software to facilitate accurate data entry



15

Ambiguity

?



16

Resolution Make Editorial decision:

“X, not Y”

Mark as “AMBIGUOUS”

“X or Y”

Get more input

“X (says expert)”



17

Lesson

In transforming data from one medium to another, even well-structured data will have unexpected pitfalls:

Design data-transformation carefully Prototype your system; Find the

problems early Plan to work iteratively



18

Romanian Online Dialect Atlas: Crisana Apply innovative contemporary

methods in dialect geography to an online set of Romanian dialect data.



19

Romanian language

Key to understanding the evolution of all Romance languages Early branch, distinct from French-

Spanish-Italian line Exemplar of non-hierarchical,

dialect variation, and linguistic continua Transition areas contain mixtures

of dialect features and specific features



20

RODA: Part 1

Create online version of The New Romanian Linguistic Atlas. Crisana (Stan & Uritescu. 1996)

Available on internet and CD Default interpretations Interactive interface to data

custom select data for a map Add audio clips to illustrate data



21

RODA Prototype 1



22

RODA: Part 2

Allow plug-in applications and other analyses of data, e.g.

Apply Multidimensional Scaling to dialect data

Statistical technique Consolidate large amounts of data Complement to traditional analyses

of small amounts of data



23

Multidimensional Scaling



24

Multidimensional Scaling Statistical technique (Torgerson

1952) Used in sociology, psychology,

marketing Reveals the scales along which

data varies; gives a data-space Uses distances [(dis)similarities]

among responses of subjects



25

MDS

10

15

250

1210

A B C D EA 0B 25 0C 10 11 0D 12 17 11 0E 24 10 15 26 0

Axioms of metric d(X,X) = 0 d(X,Y) = d(Y,X) d(X,Y) > 0 if XY d(X,Y)

d(X,C) + d(C,Y) for all points C

Matrix reflects these rules



26

MDS

n+1 points generate an n-dimensional space

MDS can reduce that high-dimensional space to 2 (or 3) dimensions

Result: complex data can be viewed as a “map”



27

MDS

Can use MDS to consolidate data English 312 dimensions reduced

to 2 All 169 features included (and

taken in relevant subsets) Finnish, Romanian provide large

data sets that can do the same



28

Interactive, media-rich presentationObjectives Make data accessible, useful to a

wide research audience

Methods Interactive selection of data Constructive presentation of data Addition of audio and other media

Online is much more than a book!



29

Framework and App’ns

Online atlas provides a framework for accessing and presenting data

Other applications can work within the framework to transform or process the data, such as:

MDS data consolidation Tools to analyze dialect variants of

phonemes (proposed) Others



30

Summary

Humanities and Social Sciences deal with large, complex data sets

Explore methods to access, process, present this kind of data

Solutions include: MDS type processing Online, interactive, rich presentation

Example: Romanian Online Dialect Atlas



31

References Embleton, Sheila M. and Eric S. Wheeler (2000). Computerized Dialect Atlas of Finnish: Dealing with

Ambiguity. J. of Quantitative Linguistics 2000. 7.3. pp 227-231. Embleton, Sheila M. and Eric S. Wheeler (1997a). Multidimensional Scaling and the SED Data. in

Wolfgang Viereck and Heinrich Ramisch. The Computer Developed Linguistic Atlas of England 2. Tuebingen: Max Niemeyer Verlag.

Embleton, Sheila M. and Eric S. Wheeler (1997b). Finnish Dialect Atlas for Quantitative Studies. J. of Quantitative Linguistics 1997. 4.1-3. pp 99-102

Schiffman, Susan S. , M. Lance Reynolds, Forrest W. Young (1981). Introduction to Multidimensional Scaling. Theory, Methods, and Applications. New York: Academic Press. 411pp.

Torgerson, W. S. 1952. Multidimensional scaling: 1. theory and method. Psychometrika. 17. 401-419. Stan, Ionel & Uritescu, Dorin. 1996. Noul Atlas lingvistic român. Crisana. Vol. I. Bucharest: Romanian

Academy Press. (2003. Vol. II. Bucharest: Romanian Academy Press) Uritescu, Dorin. 1983. “Asupra repartiţiei dialectale a graiurilor dacoromâne. Graiul din Oaş" / "On the

Dialect Structure of Daco-Romanian. The Dialect of Oaş”/, in Materiale si cercetari dialectale II, Cluj-Napoca: The University of Cluj- Napoca, pp. 231 - 246.

Uritescu, Dorin. 1984a. “Subdialectul crisean.” In: V. Rusu (ed.), Tratat de dialectologie româneasca. Craiova: Scrisul românesc, 284-320, 916-930.

Uritescu, Dorin. 1984b. “Graiul din Tara Oasului.” In: V. Rusu (ed.), Tratat de dialectologie româneasca. Craiova: Scrisul românesc, 390-399, 964-967.

Wheeler, Eric S. (2002). Zipf's Law and Why It Works Everywhere. Glottometrica 4, 45-48. Wheeler, Eric S. (2003). Multidimensional Scaling to Visualize Text Separation. Glottometrica 6

forthcoming. Wheeler, Eric S. (nd). Multidimensional scaling. chapter in Reinhard Koehler. (ed) forthcoming

Handbook in Quantitative Linguistics.

Romanian Online Dialect Atlas© 2003 Embleton, Uritescu, Wheeler1 Romanian Online Dialect Atlas An exploration into the management of high volumes of complex.

Documents

dialect atlas of finland

maps n

crisana n

dimensions n

expected n

n somerset

results n

n available