National Centre for Language Technologies

1
National Centre for Language Technologies

• conducts research into the processing of human
  language by computers
• includes speech, translation, treebanks, CALL,
  software localisation and globalisation
• interdisciplinary and has substantial economic
  implications and potential
• basic research, develops applications

2
National Centre for Language Technologies

• Enterprise Ireland Grants (2 Basic Research
  Grants)
• Research clusters
• Research collaborations

3
National Centre for Language Technologies

• Enterprise Ireland Basic Research Grant
• Deriving Linguistic Resources from Treebanks
  (BRG)
• Dates Oct 2001 - Sep 2004
• People Prof. J. Van Genabith, Dr A. Way,
• 2 PhD students A. Cahill, M.
  McCarthy
• Money Euro 130
• Overview Develop novel automatic annotation
  methods for generating new linguistic resources
  from treebanks.

4
National Centre for Language Technologies

• Enterprise Ireland Basic Research Grant
• Deriving Linguistic Resources from Treebanks
  (BRG)
• Background
• Many Natural Language Processing (NLP)
  applications require high quality training
  corpora.
• These corpora have to provide tree structures
  with meaning representations (e.g. who did what
  to whom).
• Such corpora are difficult and time-consuming to
  construct and hard to find.
• Our research involves the use of a simple, yet
  innovative approach to this problem.

5
National Centre for Language Technologies

• Enterprise Ireland Basic Research Grant
• Deriving Linguistic Resources from Treebanks (BRG)

Wall Street Journal Corpus 50, 000 sentences 1M
words
Machine Translation
Automatically Annotated Treebank
Automatic Annotation Tool
Parsing
Semantics
(42, 000 sentences annotated)
6
National Centre for Language Technologies

• Enterprise Ireland Basic Research Grant
• Deriving Linguistic Resources from Treebanks
  (BRG)
• Publications
• A. Cahill, M. McCarthy, J. van Genabith and A.
  Way Parsing with a PCFG and Automatic
  F-Structure Annotation LFG 2002, The Seventh
  International Conference on Lexical-Functional
  Grammar Athens, Greece, July 3-5, 2002
• A. Cahill, M. McCarthy, J. van Genabith and A.
  Way Automatic Annotation of the Penn-Treebank
  with LFG F-Structure Information LREC 2002,
  Third International Conference on Language
  Resources and Evaluation Las Palmas, Canary
  Islands, Spain, 27th May - 2 June, 2002
• A. Cahill and J. van Genabith TTS - A Treebank
  Tool Suite LREC 2002, Third International
  Conference on Language Resources and Evaluation
  Las Palmas, Canary Islands, Spain, 27th May - 2
  June, 2002

7
National Centre for Language Technologies

• Enterprise Ireland Basic Research Grant
• Integrating techniques from Computational
  Linguistics (CL) into CALL
• Dates Oct 2002 - Sep 2005
• People M. Ward, Dr A. Way,
• 2 PhD students
• Money Euro 140
• Overview
• To integrate techniques from Computational
  Linguistics into CALL (which currently
  under-utilises these techniques).

8
National Centre for Language Technologies

• Enterprise Ireland Basic Research Grant
• Integrating techniques from CL into CALL
• CALL is a multidisciplinary domain
• linguistics, pedagogy, computing
• The project will involve
• the development of a multi-dimensional,
  cross-classification model of CALL and CL
• the development of a low-level CL/CALL
  environment generation software system (1 PhD)
• the development of a high-level CL/artificial
  co-learner system for CALL (1 PhD)

9
National Centre for Language Technologies

• Enterprise Ireland Basic Research Grant
• Integrating techniques from CL into CALL

CALL
Beginner
Advanced Learner
Generate CALL materials
CL
High-level CL techniques
Low-level CL techniques (part of speech tagging,
morphology)
A multi-dimensional, cross-classification matrix
10
National Centre for Language Technologies

• Research clusters
• Finite State Technology (FST)
• Speech
• Machine Translation (MT)
• Corpora
• Computer Assisted Language Learning (CALL)
• Semantics
• Machine Learning
• Virtual Reality

11
National Centre for Language Technologies

• Finite State Technology (FST)
• 2-level morphology for Irish (Elaine)
• uses Xerox Finite State Technology to implement a
  2-level morphology of Irish for the inflected
  parts of speech i.e verbs, nouns and adjectives.
• E.g. analysis (raibh, bí) conjucation (bí -gt
  all tenses, all forms)
• FST chunking for English (Patricia)
• Developing a chunking grammar for unrestricted
  English text using the Xerox Incremental Parser
  (XIP).
• Advantages
• fast
• works with real languages
• unrestricted
• input into EI CL/CALL project

12
National Centre for Language Technologies

• Speech
• speech generation (Ronan)
• speaker characterisation (John, Michelle)
• modelling voice source (John)
• Irish speech synthesis group (Ronan, John,
  Michelle, Monica)
• multi-modal interfaces (Donal)

13
National Centre for Language Technologies

• Applications of Speech Technology and Multi-modal
  interfaces

Internal Format
Document
Very structured information Uses visually
impaired people, PDAs, text msgs,
mobile phones
Spoken text
14
National Centre for Language Technologies

• Machine Translation (MT)
• constrained language (Sharon)
• example based machine translation (Andy, Nano)
• combination of statistical and rule-based
  translation - LFG-DOT (Andy, Mary)
• Forthcoming conference
• May 2003 Controlled Translation (EAMT, CLAW)

15
National Centre for Language Technologies

• Corpora
• automatic treebank annotation (Mairead, Andy,
  Josef)
• stochastic parsing with treebank grammar (Aoife,
  Andy, Josef)
• translation corpora (Dorothy, Gabi, Marion)
• aligned bilingual corpora (Nano, Andy)
• e.g. Belgium
• test our translation systems with internationally
  recognised reference corpora

16
National Centre for Language Technologies

• CALL
• Learner Autonomy (Francoise)
• Tandem email (Christine)
• CALL and CL (Monica, Andy, Josef)
• CALL for Minority and Endangered Languages
  (Monica, Andy)

17
National Centre for Language Technologies

• Other areas
• Semantics
• Machine Learning
• Virtual Reality

18
National Centre for Language Technologies

• Active Research Events
• Dublin Computational Linguistics Seminar Series
• held weekly with TCD and UCD (initiated by DCU 5
  years ago)
• Human Language Technology reading group
• informal weekly seminar series (in its 3rd year)
• Recent Awards
• 3 researchers won Albert College Fellowships