Language and Intelligence

1
Language and Intelligence
Natural Language Processing (NLP), Machine
Translation (MT), Computer Assisted Language
Learning (CALL), Speech
Artificial Intelligence, World Wide Mind
Language
Intelligence
Language Intelligence
Language Evolution, Semantics, 3D Worlds, Neural
Networks, Speech and Multi-Modal Interfaces
School of Computer Applications
2
Language and Intelligence
Staff Postgrad Students Dr D. Fitzpatrick A.
Cahill, N. Gough, J. Hayes S. Harford, M.
Hearne, Dr M. Humphrys M. Mc Carthy, C.
OLeary, J. Kelleher M. Tooher Dr J. Mc
Kenna Prof J. Van Genabith Affiliated
Researcher R. Walshe D. OConnor M. Ward Dr
A. Way
3
Language and Intelligence

• NCLT
• National Centre for Language Technologies
• computing.dcu.ie/nclt
• World Wide Mind
• w2mind.org

4
Language Research Areas

• Example-Based Machine Translation (EBMT)
• People Dr A. Way
• M. Hearne Hybrid (Stats rule-based)
  Machine Translation
• N. Gough Web-Based Machine Translation
• Overview
• We are currently investigating two approaches to
  MT which can broadly be described as EBMT
• a) Marker-based EBMT
• b) DOT and LFG-DOT

School of Computer Applications
5
Language Research Areas
Example Based Machine Translation
Given John went to school Jean est allé à
lécole. The butchers is next to the bakers La
boucherie est à côté de la boulangerie.
Isolate useful fragments John went to
Jean est allé à the bakers la
boulangerie We can now translate John went to
the bakers as Jean est allé à la boulangerie.
School of Computer Applications
6
Language Research Areas

• Speaker Characterisation
• People Dr J. McKenna
• M. Tooher Machine Learning of Speaker
  Characteristic
• Speech Dynamics and Interactions
• Overview
• Our research aims to separate the linguistic
  content of speech from that containing
  speaker-specific information.

School of Computer Applications
7
Language Research Areas

• Speaker Characterisation

Machine Translation
Separate Linguistic Data from
Speaker Characteristics
Hello
New Language
Bonjour
School of Computer Applications
8
Language Research Areas

• CALL
• use of XML technologies
• specific requirements for Endangered Languages
• e.g. computing.dcu.ie/mward/nawat.html
• interest from UNESCO, European Bureau of Lesser
  Used Languages
• working with projects in Siberia and Togo/Benin
• VOCALL (Vocationally oriented CALL)

School of Computer Applications
9
Intelligence

• World Wide Mind project
• People Dr M. Humphrys, R. Walshe. C. OLeary,
  D. OConnor
• Overview
• This is a new idea for decentralising the work in
  AI by putting agent mind and worlds online as
  reusable servers
• This work proposes that the construction of
  advanced artificial minds may be too difficult
  for any single lab
• No easy system exists whereby a working mind can
  be made from the components of two or more labs
• Our system aims to change this and accelerate the
  growth of AI

School of Computer Applications
10
Intelligence

• Society of Mind constructed from Multiple servers
• 1. client talks to
• 1. MindM, which talks to
• 1. Mind
• 2. MindM, which talks to
• 1. Mind
• 3. MindAS, which talks to
• 1. Mind
• 2. MindM, which talks to
• 1. Mind
• 3. Mind
• 2. WorldW, which talks to
• 1. World

School of Computer Applications
11
Language and Intelligence

• World Wide Mind

State
Mind
Mind Server
Mind
Action
Action
State
World (problem to solve)
State
Client (do some task)
State
Mind
Action
Action
Uses World Wide Web and cgi-bin/perl for
communication
School of Computer Applications
12
Language and Intelligence

• Dr D Fitzpatrick Applications of Speech
  Technology and Multi-modal interfaces

Force Feedback/ (Haptic) Device
Information Analysis
Map
Purpose to convey spatial information
non-visually i.e. using sensors other than vision
School of Computer Applications
13
Language and Intelligence

• R. Walshe Evolution of Early Language

Go Left
World
State
State of the world
Grrraahhh ???
Grrraahhh
Action
Action
Agent (Speaker, Hearer, Learner) Reinforcement
Learning Network (Neural Network)
Agent (Speaker, Hearer, Learner) Reinforcement
Learning Network (Neural Network)

• Unique features
• No master
• No prior language knowledge

School of Computer Applications
14
Language and Intelligence

• J. Kelleher Natural Language interface to 3D
  world - Situated Language Interpreter

Natural Language Understanding
Visual Context
Natural Language Interface
School of Computer Applications
15
Language and Intelligence

• J. Hayes Semantics - computational modelling of
  nominal compounds

Linguistic Level Compounding Cognitive
Process Concept Combination
?
Computer
Wizard

Computer Wizard
Generate an Interpretation (form a meaning)
Interpretation
School of Computer Applications
16
Language and Intelligence

• S. Harford A Neural Network model of Melodic
  Memory

Output
Feedback
Learning, Feedforward
Input
Neural Networks
Processing
School of Computer Applications