Sale!
Placeholder

An Evaluation of the GOOGLE Translation Software

10,000 3,000

Topic Description

SUMMARY
This aim of this project was to produce an evaluation of the automatic translations, of combinations of
adjectives and nouns from French to English, produced by the GOOGLE Language Tool. In other words,
it was undertaken in order to assess how well GOOGLE’s machine translation system works for this
specific subset of the French language. The evaluation was from the perspective of the person proficient
in both English and French, interested in whether, or how well, GOOGLE ’gets the translation right’ based on the original combination, but similarly included the perspective of the monolingual user – English native speaker – whose standard is based on what the translation should have been.

TABLE OF CONTENTS
_____________________________________________________________________________________
SUMMARY ii
_____________________________________________________________________________________
ACKNOWLEDGEMENTS iii
________________________________________________________________________
Chapter One INTRODUCTION 1
1.1 Project Aim and Scope 1
1.2 Project Methodology and Limitations 1
1.3 Project Objectives 2
1.4 Project Deliverables 2
1.5 The Minimum Requirements 2
________________________________________________________________________
Chapter Two BACKGROUND RESEARCH 3
2.1 Machine Translation 3
2.1.1 About Machine Translation 3
2.1.2 MT Evaluation: Methods and Challenges 4
2.1.3 GOOGLE’s Translation Software 6
2.2 Linguistics 8
2.2.1 Natural Language 8
2.2.2 Nouns, Adjectives and Collocations 9
2.3 Using Corpora 10
2.4 Statistics 11
2.4.1 Statistically representing populations 11
2.4.2 Comparative tests for populations 12
2.5 Conclusion 12
_____________________________________________________________________________
Chapter Three DERIVING THE GOLD STANDARD
AND THE GOOGLE TRANSLATIONS 13
3.1 Building the Corpus 13
3.1.1 About the Source of the Corpus 13
3.1.2 Selecting the Texts in the Corpus 14
3.2 Annotating the French Adjective-Noun Combinations 14
3.2.1 Annotation Criteria 14
3.2.2 Mark-up Scheme 16
An Evaluation of the GOOGLE Translation Software
Marisa Alexander
Final Year Project
v
3.3 Annotating the English Gold Standard Translations 16
3.3.1 Classifications 16
3.3.2 Mark-up Scheme 17
3.4 Annotating the English GOOGLE Translations 17
_______________________________________________________________________
Chapter Four GOOGLE against GOLD STANDARD:
QUANTITATIVE RESULTS 18
4.1 Methodology 18
4.2 Statistics on GOOGLE Classifications 18
4.3 Statistics on Gold Standard Classifications 19
4.4 Quantitative Evaluation 19
_______________________________________________________________________
Chapter Five CLASSIFYING THE ERRORS 21
5.1 Methodology 21
5.2 Error Classification 22
_________________________________________________________________________
Chapter Six GOOGLE against GOLD STANDARD:
QUALITATIVE RESULTS 23
6.1 Questionnaires: Obtaining Human Judgement 23
6.1.1 Developing the questionnaires 23
6.1.2 Distribution and Returns 24
6.1.3 Results 25
6.2 Qualitative Evaluation and Ideas for Improvements 26
_______________________________________________________________________
Chapter Seven VERIFYING MY EVALUATION 28
7.1 Reasons for Not Calculating the Edit Costs 28
7.2 Pros and Cons of the Methodology 28
_______________________________________________________________________
Chapter Eight CONCLUSION 30
_____________________________________________________________________________________
APPENDICES 31
_____________________________________________________________________________________
BIBLIOGRAPHY 44

GET COMPLETE PROJECT