LANGUAGE RESEARCH IN SERVICE TO THE NATION BUREAU FOR INTERNATIONAL LANGUAGE CO-ORDINATION BUREAU DE COORDINATION LINGUISTIQUE INTERNATIONALE Defense Language.

LANGUAGE RESEARCH IN SERVICE TO THE NATION

BUREAU FOR INTERNATIONAL LANGUAGE CO-ORDINATION

BUREAU DE COORDINATION LINGUISTIQUE INTERNATIONALE

Defense Language Institute Foreign Language Center (DLIFLC)

October 18, 2011Richard Brecht, PhDExecutive DirectorUniversity of Maryland Center for Advanced Study of Language1

LANGUAGE RESEARCH IN SERVICE TO THE NATION

Unique UARC Mission/Role

CASL is charged to deliver independent, empirically-based research solutions that:• Improve the human-centric components of military and intelligence

operations that include language performance, analysis and culture• Underpin leadership decisions on language readiness for the

warfighter and the intelligence analysts who support them• Elevate language to the stature of math and science as an IC & DoD

core capability

CASL is the only University Affiliated Research Center devoted to the study of language

2

LANGUAGE RESEARCH IN SERVICE TO THE NATION

UARCs

• Essential research, development and engineering "core" capabilities

• Maintain long-term strategic relationships with their DoD sponsors

• Operate in the public interest, free from real or perceived conflicts of interest

3

LANGUAGE RESEARCH IN SERVICE TO THE NATION

CASL Core Competencies

• Foreign languages and dialects

• Linguistics

• Disciplines relevant to analysis and critical thinking (cognitive & neurological sciences)

• Disciplines relevant to the manipulation, use, and sharing of information of varying quantities & diverse forms (tools & technologies, Human-Computer Interaction)

4

LANGUAGE RESEARCH IN SERVICE TO THE NATION

CASL Research Staff

Cognitive Science38%61 researchers

Linguistics44%71 researchers

Second Language Acquisition18%28 researchers

Total CASL Researchers = 160

5

LANGUAGE RESEARCH IN SERVICE TO THE NATION

Featured CASL Products

Assessment:•Language Product Evaluation Tool (LPET)

Resources: •Learning Management System: STARTALK

•African Languages: ALIFA, Pasaglossa

•Under-resourced languages: Grammars, Dictionaries, Parsers, new Mouton de Gruyter Series

6

LANGUAGE RESEARCH IN SERVICE TO THE NATION

Summary Translation

7

LANGUAGE RESEARCH IN SERVICE TO THE NATION

Language Product Evaluation Tool (LPET)

• Designed to establish standards for evaluating language translations, complete summaries, gists, etc.

• Makes quality control more systematic

• Enables consistent, comprehensive feedback for skill growth

• Streamlines the reporting process

• Can be migrated into coursework

8

LANGUAGE RESEARCH IN SERVICE TO THE NATION9

LANGUAGE RESEARCH IN SERVICE TO THE NATION

Teaching Mandarin Chinese Online

• Learning foreign languages at a distance

• Prototype implementation of “blended” learning model

• Special challenges of LCTLs

• “Teaching Mandarin Chinese Online: A guide to implementation from theory to practice,” by Katharine B. Nielson

10

LANGUAGE RESEARCH IN SERVICE TO THE NATION

Intermediate Online Mandarin Chinese• Overview of course content

• Resources required for online implementation

• Instructions for Course Implementation

• Research requirements

11

LANGUAGE RESEARCH IN SERVICE TO THE NATION

STARTALK LMS

12

• Built custom one-stop language Learning Management System (LMS)

• Combines language related tasks with VoIP and latest social media technology

• Activities are targeted: based on needs-analysis

• Lessons are here: http://www.chineseforall.org/ch_2009/

LANGUAGE RESEARCH IN SERVICE TO THE NATION

African Language Identification

• Initiative to build core and surge capabilities for 2K+ languages & 1K+ dialects & pidgins

• Language mapping tool with language ID interface & database of available experts & resources

• Lexical comparison generator

• Links to additional online language resources

13

LANGUAGE RESEARCH IN SERVICE TO THE NATION

African Language Identification and Familiarization Aid

14

LANGUAGE RESEARCH IN SERVICE TO THE NATION15

LANGUAGE RESEARCH IN SERVICE TO THE NATION

Lexical Comparison Generator

16

LANGUAGE RESEARCH IN SERVICE TO THE NATION

Under-Resourced Languages• Grammar series: Pashto, Bangla, Dhivehi,

Punjabi, Swahili, Javanese, Burmese, Bambara

• Morphological parsers: Grammars in Bangla, Urdu, and Pashto already imported into morphological parser and being used by USG professionals nation-wide

• Automatic tools for language analysis: Did You Mean? In Urdu and two dialects of Arabic

• Compendium of key words in 19 dialects of Arabic

17

LANGUAGE RESEARCH IN SERVICE TO THE NATION

Cognitive Science for Improved Candidate Selection, Diagnostic Profiling and Curriculum Adaptation

Defense Language Institute Foreign Language Center (DLIFLC)

October 18, 2011Amy Weinberg, PhDDeputy Executive DirectorUniversity of Maryland Center for Advanced Study of Language

18

LANGUAGE RESEARCH IN SERVICE TO THE NATION

Presentation Structure

• Predicting Language Success and Diagnosing Individual Differences:– Executive Function– Hi-LAB: Cognitive predictors of high level language

aptitude– ALAB: Applications to assessing aptitude for learning

English– Cognitive Profiling

• Improving Readiness to Learn:– Working memory abilities can be improved through

training• Individualized Curricula

LANGUAGE RESEARCH IN SERVICE TO THE NATION20

Predicting Language Success

LANGUAGE RESEARCH IN SERVICE TO THE NATION

A Causal Model of Language Proficiency

Cognitive and perceptual components associated with language aptitude and proficiency

the working memory system

21

LANGUAGE RESEARCH IN SERVICE TO THE NATION

Executive Function: Working Memory• Active information storage that allows

attention to be focused for the purpose of higher level combination, reasoning, or other subsequent processing.

• Decays quickly

• Varies among individuals

• Affects language processing (e.g. error recovery)

22

LANGUAGE RESEARCH IN SERVICE TO THE NATION

Instructions:Monitor the identity of a series of letters and

indicate when the current letter is the same as the one presented 2 trials previously

U

T

R

T

Example:

Demo: N-Back Training Task

23

LANGUAGE RESEARCH IN SERVICE TO THE NATION

NBVBXLYDYCQC

24

LANGUAGE RESEARCH IN SERVICE TO THE NATION25

N B V B X L Y D Y C Q C

LANGUAGE RESEARCH IN SERVICE TO THE NATION

Results: Composite α = 0.90Cognitive construct Reliable Stable

Jointly acceptable

Short-term memory 2 of 2 1 of 2 1 of 2

Updating 3 of 3 2 of 3 2 of 3

Inhibition 3 of 3 2 of 3 2 of 3

Task-switching 2 of 2 1 of 2 1 of 2

Novel word memory 3 of 3 1 of 3 1 of 3

Perceptual acuity 3 of 3 2 of 3 2 of 3

Speed 2 of 2 1 of 2 1 of 2

Semantic primability 1 of 3 1 of 3 1 of 3

Implicit induction 1 of 3 1 of 3 1 of 3

26

LANGUAGE RESEARCH IN SERVICE TO THE NATION

Afghan-LAB: Aptitude For EnglishNATO and DLIELC Requirement Need aptitude instrument to better predict language learning success in the DLIELC-Afghan program

RelevanceNeed better readiness for follow-on training and moreDLIELC graduates meeting graduation requirements

SolutionQuick fix for immediate implementation followed by more rigorously developed test27

LANGUAGE RESEARCH IN SERVICE TO THE NATION28

Executive Function: Induction

Induction

Ability to infer a general conclusion from particular instances.

Induction in the language domain

Ability to infer the rules generating language patterns (word or sentence structure)

LANGUAGE RESEARCH IN SERVICE TO THE NATION

Content of the ALAB

29

Non-linguistic tests assess aptitude for learning

Linguistic tests assess aptitude for language learning

LANGUAGE RESEARCH IN SERVICE TO THE NATION30

Content of the ALAB: Raven’s Test

Choose the picture that completes the large pattern

LANGUAGE RESEARCH IN SERVICE TO THE NATION

Content of the ALAB

31

Non-linguistic tests assess aptitude for learning

Linguistic tests assess aptitude for language learning

LANGUAGE RESEARCH IN SERVICE TO THE NATION

Linguistic Tests

• ALAB has two linguistic tests: Language Analysis and Words in Sentences

• In addition to the ALAB instructions, language and cultural concepts must be adapted and translated.

• Language Analysis: required transliteration of an artificial language from Roman to Arabic

• Direct assessment of linguistic ability

LANGUAGE RESEARCH IN SERVICE TO THE NATION

Example: Language Analysis Test• Word list:

zorit..............................the farmervolip.............................the workerzorit volipu pigom........The farmer pushed the worker.

•Exercise:

33

The worker pushed the farmer.

A. zoritu volip pigom

B. volip zoritu pigom

C. zorit volip pigom

D. volip zorit pigom

LANGUAGE RESEARCH IN SERVICE TO THE NATION34

Cognitive Profiling

LANGUAGE RESEARCH IN SERVICE TO THE NATION

LANGUAGE RESEARCH IN SERVICE TO THE NATION

Improving Readiness to Learn

36

LANGUAGE RESEARCH IN SERVICE TO THE NATION

Behavioral Working Memory Training Study

Goal: Test if training gains transfer to untrained measures of attention, language, and reasoning

Pre-testing Training (N = 47)(20 hours training)

Post-Testing

Follow-up

Training Condition

Control Condition

5 Weeks 3 Months

Pre-testing No Contact (N = 45) Post-testing

Follow-up

37

LANGUAGE RESEARCH IN SERVICE TO THE NATION

Improvements on Training Tasks

Do training gains transfer to measures of sentence reinterpretation?

n-back task

38

LANGUAGE RESEARCH IN SERVICE TO THE NATION

Accuracy on Comprehension Questions:“Did the thief hide himself?”

Pretest Posttest

Pro

port

ion

of E

rror

s

*

ns

Ambiguous

Unambiguous

ns

ns

*

39

LANGUAGE RESEARCH IN SERVICE TO THE NATION

fMRI activation: PreTraining

40

Pre-training brain activation in the high memory load condition for the vsWM and the placebo control group. Panel A shows the vsWM group, and panel B shows the placebo control group. Panel C shows the β values for the right and left IPL. 

No difference in activation prior to training

LANGUAGE RESEARCH IN SERVICE TO THE NATION

fMRI activation: PostTraining

41

Less diffuse activation following working memory training

Post-training minus pre-training BOLD brain activation in the high memory load condition for the vsWM and the placebo control group. Panel A shows the Post-Pre for the vsWM group, and panel B shows the Post-Pre for the placebo control group. Panel C shows Placebo control minus the vsWM Post-Pre. Panel D shows the β values for the right and left IPL. 

LANGUAGE RESEARCH IN SERVICE TO THE NATION

Individualized Curricula

42

LANGUAGE RESEARCH IN SERVICE TO THE NATION

Aptitude-by-Treatment Interaction

Problem

Individuals who are already selected for intelligence and motivation and have attained functional ability need to acquire higher levels of language

Solution

Match cognitive aptitude profile to language training procedures to optimize training and boost outcome levels

LANGUAGE RESEARCH IN SERVICE TO THE NATION44

ATI Cycle

LANGUAGE RESEARCH IN SERVICE TO THE NATION

For More Information:

www.casl.umd.edu

Richard Brecht

rbrecht@casl.umd.edu

Amy Weinberg

aweinberg@casl.umd.edu

Research Contacts

LPET

Erica Michael, PhD

emichael@casl.umd.edu

STARTALK

Katharine Nielson, MA

knielson@casl.umd.edu

PASSAGLOSSA

Christopher Green, PhD

cgreen@casl.umd.edu

Under-Resourced Languages

Michael Maxwell, PhD

mmaxwell@casl.umd.edu

Hi-LAB, ATI

Cathy Doughty, PhD

cdoughty@casl.umd.edu

ALAB, Working Memory

Michael Bunting, PhD

mbunting@casl.umd.edu

For more information on individual research projects presented here, please contact: