Site Report from Carnegie Mellon University

1
Site Report from Carnegie Mellon University

• Sixth workshop on interlinguas
• September 23, 2003
• Teruko Mitamura and Lori Levin

2
Outline

• Background technology
• Merge and consolidate language-faithful
  interlinguas into conceptual graphs
• Examples from annotation exercise
• Mismatches due to differences in text planning
  (number of events, objects, clauses, etc.)
• Proposal for annotation of mismatches
• Language-faithful annotation
• Merged interlingua

3
Reference

• Eric Nyberg, Teruko Mitamura, Dave Svoboda,
  Kathryn Baker, Brian Peterson, Jennifer Williams
  (2002). Deriving Semantic Knowledge from
  Descriptive Texts Using an MT system,
  Proceedings of AMTA 2002

4
KANT(OO) Multilingual MT

• Knowledge-based, interlingual MT
• Multilingual translations of technical documents
• Domain includes heavy equipment documentation,
  computer manuals, automotive documentation, etc.
• Documents are written in Controlled Language
• KANTOO (1998) is the reimplementation of the KANT
  MT system (1991)

5
KANT Architecture
Text
KANT Analyzer
Feedback
Author
Grammar Checking
Translation
KANT Generator
Target Texts
6
KANT Controlled Language Checker

• Thin-client checker program runs on authors PC
  (Java)
• Accesses KANT analyzer software running on a
  network server
• Features
• Dynamic checking (while the author is typing)
• Automatic and interactive disambiguation
• Pronoun resolution
• Interactive rewriting to controlled English
• Grammar Diagnostics

7
ControlledLanguageChecker
8
CL Input to Knowledge Interchange Format (KIF)
Text
KANT Analyzer
Feedback
Author
Grammar Checking
Generation
Merged Interlingua
Interlingua Merger
KIF
Output
9
CL for Paragraphs

• Paragraph-level writing guidelines
• To promote isomorphism
• To look for shared elements from sentence to
  sentence
• Work from graph to text
• Develop a set of key words for mapping
• in order to Purpose
• by using Method

10
Merging Interlinguas (IRs) into CG

• Preprocess the IRs
• Construct an IR Index for every concept
• Find concepts appearing in more than one IR
• Create unique identifiers for multiple (unique)
  instances
• Create unified IR for each concept
• Merge the unified IRs
• Remove features which dont contain concepts

11
Examples for Merged Interlingua

• A prime mover drives a generator rotor.
• A prime mover is part of the generator.
• A generator rotor is part of the generator.
• A generator is a type of power generation system.

12
Merged Interlingua Output

• (A-DRIVE
• (agent
• (O-PRIME-MOVER-c
• (part-of_OF
• (O-GENERATOR-b
• (type_OF (O-POWER-GENERATION-SYST
  EM))))))
• (theme
• (O-GENERATOR-ROTOR
• (part-of_OF (O-GENERATOR-b)))))

13
MergedInterlingua
14
KIFOutput
15
Annotation of French and English

• Original French text annotated by Simon Fung
• Original English text annotated by Teruko
  Mitamura and Lori Levin
• English translation from French Simon Fung
• Annotation of Translated English Simon Fung

16
Mismatches caused by text planning

• Different number of clauses
• Differences in which clause is the main clause
  and which is a subordinate clause.
• Resulting in different numbers of events, states,
  and objects

17
Example 1

• Original English
• In its first five years of operation, PRODEM
  financed loans to over 13,300 micorentrepreneurs,
  77 per cent of whom were women, disbursing over
  27 million in loans averaging 273.
• Original French
• Au bout de cinq ans, le programme avait consenti
  plus de 27 millions de dollars de prets d'un
  montant moyen de 273 dollars, a plus de13 300
  entrepreneurs, dont 77 de femmes ....
• English Translation from French
• At the end of five years, the program had granted
  more than 27 million dollars in loans with an
  average amount of 273 dollars, to more than 13
  300 entrepreneurs, of which 77 were women,....

18
Example 1

• Original English
• financed
• loans
• to over 13,300 micorentrepreneurs,
• disbursing
• over 27 million
• in loans
• Original French
• consenti
• plus de 27 millions de dollars
• de prets
• a plus de 13 300 entrepreneurs,
• English Translation from French
• granted
• more than 27 million dollars
• in loans
• to more than 13 300 entrepreneurs

19
Example 2

• Original English
• Its network of eighteen independent organizations
  in Latin America has lent ..
• Original French
• le reseau regroupe dix-huit organisations
  independantes qui ont debourse ..
• English Translation from French
• the network comprises eighteen independent
  organizations which have disbursed ..

20
Example 2

• Original English
• has lent
• Its network
• of eighteen independent organizations
• ..
• Original French
• regroupe
• le reseau
• dix-huit organisations independantes
• ont debourse
• English Translation from French
• comprises
• the network
• eighteen independent organizations
• have disbursed

21
Proposal for multilingual corpus annotation

• Language faithful annotation
• Reflects the text planning and predicate argument
  structure of a particular sentence
• Can be annotated reliably
• Useful for many NLP applications
• Merged interlingua
• Input one or more language faithful annotations
  for the same sentence
• Align the language faithful annotations
• Identify co-reference of entities and events
• Use a limited set of well-defined rules to
  re-write the language faithful interlinguas

22
Interlingua Merging

• Language-faithful interlinguas
• Original English
• financed
• loans
• to over 13,300 micorentrepreneurs
• disbursing
• over 27 million
• in loans
• Original French
• consenti
• plus de 27 millions de dollars
• de prets
• a plus de 13 300 entrepreneurs
• English Translation from French
• granted
• more than 27 million dollars
• in loans

• Merged Interlingua
• TRANSFER-MONEY
• over 27 million
• to over 13,300 micorentrepreneurs
• SOME-RELATION
• over 27 million
• loans

23
Interlingua Merging

• Original English
• has lent
• Its network
• of eighteen independent organizations
• Original French
• regroupe
• le reseau
• dix-huit organisations independantes
• ont debourse
• English Translation from French
• comprises
• the network
• eighteen independent organizations
• have disbursed

• Merged Interlingua
• HAS-AS-PART
• the network
• eighteen independent organizations
• TRANSFER-MONEY
• the network
• ..

24
Proposal for multilingual corpus annotation

• Language faithful annotation
• Reflects the text planning and predicate argument
  structure of a particular sentence
• Can be annotated reliably
• Useful for many NLP applications
• Merged interlingua
• Input one or more language faithful annotations
  for the same sentence
• Align the language faithful annotations
• Identify co-reference of entities and events
• Use a limited set of well-defined rules to
  re-write the language faithful interlinguas

25
Example 3

• Original English
• Three of the most advanced institutions in the
  ACCION network started their programmes as
  non-profit organizations and have, in the last
  five years, converted into
• Original French
• Trois des institutions les plus performantes
  rattachees a ACCION International qui etaient au
  depart des organisations a but nonlucratif sont
  devenues ces cinq dernieres annees
• English Translation from French
• Three of the most successful institutions
  connected to ACCION International, which were
  non-profit organizations in the beginning, have
  become, in these last five years,

26
Example 3

• Original English
• Started
• their programmes
• Institutions
• as non-profit organizations
• Converted
• Institutions
• ..
• Original French
• sont devenues
• Institutions
• relative-clause etaient au depart
• institutions
• English Translation from French
• Have become
• Institutions
• Relative-clause Were in the beginning
• institutions