Chapter One: Introduction

1.1Research purpose and questions

Although machine translation is very popular at present, the readability of the target language texts is far from satisfaction. Despite the fact that in the past few decades, machine translation functions have been greatly improved, the ability of machine translation engines is still limited and wants optimization (魏长宏and 张春柏，2007).One of the main reasons for this is that machines cannot fully understand the meaning and structure of the source language texts. Hence, measures like pre-editing should be taken to make the texts generated by machine translation engines better.

First, this paper intends to find out the circumstances in which errors of target language texts generated by the Google machine translation engine will occur. Then, replicable pre-editing strategies will be put forward accordingly to avoid these errors so that people with only a limited knowledge of the target language can generate readable translation texts with the Google machine translation engine. Moreover, despite the fact that the thesis is only a case study of the Google machine translation engine, the author sincerely hopes that the pre-editing strategies put forward in this thesis will be of practical reference significance to other machine translation systems and that this thesis will make some contribution to the overall improvement of machine translation quality and efficiency. The research questions of the essay are as follows:

1. What are the deficiencies of the Google machine translation engine?

2. In what circumstances may machine translation errors occur?

3. Which strategies of pre-editing will improve the quality of target language texts?

4. To what extent can pre-editing improve the quality of target language texts?

1.2 Research significance

This research is both of theoretical and practical significance. Theoretically, it intends to fill the gap of practical pre-editing studies. As will be seen in the following literature review in Chapter Two, although there are lots of research studies about “machine translation” in China, there are still not many research studies on pre-editing. Worse still, almost all the research studies on pre-editing focused their attention on finding technological solutions. No one had attempted to find a way thats easy for every ordinary people with a limited knowledge of the English language to improve machine translation performance. These research blanks should be filled to warrant machine translation a comprehensive and healthy development. Practically, this thesis intends to find out a series of replicable pre-editing strategies that people with only a limited knowledge of target language can use to generate readable translation texts with the Google machine translation engine. Although this paper is only a case study of the Google machine translation engine, the pre-editing strategies proposed here will be of applicable significance to other machine translation systems, thus greatly enhance the performance of machine translation as a whole.

1.3 Organization of thesis

This study consists of seven chapters. Chapter One serves as the introduction of the whole paper and demonstrates the research purpose, questions and significance of the study, as well as the organization of this thesis. Chapter Two is the literature review of relevant studies on machine translation and pre-editing. In Chapter Three, the author introduces the research methodology, which deals with experiment design, sample choosing and research methods. Chapter Four presents the circumstances in which the Google machine translation errors tend to arise and probes into the causes of these deficiencies. Chapter Five covers the pre-editing strategies categorized based on the previous analysis of the former chapters and the sample study of the thesis. Chapter Six demonstrates the experiment, which includes data analysis and experiment results. In the last chapter, the author draws a conclusion of the major findings of the study, discusses the implications and applications of the present study, and suggests some possible issues for further study.

Chapter Two: Literature Review

2.1 Previous studies on machine translation

With the growth of international communication, the demand for translation has grown rapidly. People begin to realize that human translators can hardly satisfy their communication needs, thus an increasing number of people choose to turn to machine translation for help.

In 1954, Georgetown University and International Business Machines Corporation (IBM) successfully finished some translation work that goes beyond word to word translation, which demonstrated to the world that machine translation is practical. Shortly after, a book entitled Machine Translation of Languages (Locke amp;Donald, 1995) was published, which contained fourteen essays and was ; comprehensive demonstration of machine translation field. However, in 1964, the U.S Automatic Language Processing Committee (ALPAC) released a report and declared that:” There is no immediate or predictable prospect of useful machine translation (1966)”. Thereafter, many people withdrew from machine translation research.

In the 1970s, machine translation research came into a recovery and development stage. Two of the most representative study results during this period were: SYSTRAN and METEO. Hutchins W.J. published an essay “Machine Translation and Machine-aided Translation”(1978 ),which was a review of past development, an assessment of current situation, and a guess of the future in machine translation. Then in the 1980s, Europe started a huge project named Eurotra, which was the first open machine translation system and could accommodate new methods and technologies in later development. Meanwhile, in the year 1986, the first international journal entitled Machine

剩余内容已隐藏，支付完成后下载完整资料

第一章：简介

1.1研究目的和内容

虽然机器翻译目前非常流行，但目标语言文本的译文质量依然不能令人满意。 尽管机器翻译功能在过去的几十年中已经大大提高，机器翻译引擎的水平仍然有限，并且有望被不断优化（魏长宏，张春柏，2007）。其中一个主要原因是机器不能完全了解源语言文本的含义和结构。 因此，我们应该采取预先编辑等措施，优化机器翻译引擎生成的文本。

首先，本文将例举Google机器翻译引擎生成错误译文的的实例。然后提出可行的预编辑策略以避免这些错误，以便目标语言知识有限的人能够通过Google机器翻译引擎生成正确的译文。 最后，尽管本文只是对Google机器翻译引擎的个案研究，但作者衷心希望提出的预编辑策略对其他机器翻译系统具有实际的参考意义，本文将为机器翻译质量和效率的整体提高做出一些贡献。 本文的研究问题如下：

1.谷歌机器翻译引擎的不足之处是什么？

2.在哪些情况下可能会出现机器翻译错误？

3.哪些预编辑策略可以提高目标语言文本的质量？

4.预编辑能够在多大程度上提高目标语言文本的质量？

1.2研究意义

这项研究既具有理论意义，也具有现实意义。理论上，它打算填补实际编辑前研究的空白。从第二章的以下文献综述中可以看出，虽然在中国有很多关于“机器翻译”的研究，但对于预编辑的研究还很少。更糟糕的是，几乎所有关于预编辑的研究都集中于寻找技术解决方案。没有人试图找到一种对于英语知识有限的普通人来说很容易提高机器翻译性能的方法。这些研究空白应该填补，以保证机器翻译的全面和健康的发展。实际上，本文旨在找出一系列可复制的预编辑策略，可以帮助目标语言知识有限的人能使用谷歌机器翻译引擎生成正确的译文。虽然本文只是谷歌机器翻译引擎的案例研究，但这里提出的预编辑策略对于其他机器翻译系统具有重要意义，能够有效提高了机器翻译的整体性能。

1.3框架结构

本研究由七章组成。第一章简述本文的研究目的，研究内容，研究意义以及组织结构。第二章是机器翻译和预编辑相关研究的文献综述。第三章介绍研究方法，涉及实验设计，样本选择和研究方法。第四章介绍了Google机器翻译错误倾向发生的情况，并探讨了这些缺陷的原因。第五章基于前几章的分析和样本研究分类，简述预编辑策略。第六章以实验为例，展示了包括数据分析和实验结果。在最后一章，作者对本研究的主要结果作了总结，展现了本研究的意义和应用，并提出了一些可供进一步研究的问题。

第二章：文献综述

2.1过去机器翻译的研究

随着国际交流的发展，翻译需求迅速增长。人们开始意识到，译员很难满足他们的沟通需求，越来越多的人选择使用机器翻译。

1954年，乔治敦大学和国际商用机器公司（IBM）成功完成了一些超越单词到词语翻译的翻译工作，向全世界证明机器翻译的实用性。不久后，《机器翻译》（Locke＆Donald，1995）出版，书中的十四篇文章全面详尽的介绍了机器翻译。然而，1964年，美国自动处理咨询委员会（ALPAC）发布了一份报告，并称：“有用的机器翻译（1966年）没有的前景”。此后，许多人退出机器翻译研究。

在20世纪70年代，机器翻译研究进入了一个复苏和发展阶段。在此期间最有代表性的两项研究成果是系统翻译和报告翻译。Hutchins W.J.发表了一篇文章《机器翻译和机器辅助翻译》（1978），该文章回顾了机器翻译的发展，对其发展现状的分析以及对未来的展望。

在20世纪80年代，欧洲开始了一个大型项目：Eurotra，这是第一个开放的机器翻译系统，可以在后期开发中容纳新的方法和技术。在1986年，第一本题为《机器翻译》的国际期刊成立。 研究机器翻译理论和方法论问题会议和机器翻译峰会定期举行。国际语言学会和各种关于自然语言处理的国际会议都把机器翻译作为其重要议题。在机器翻译发展期间出现的散文包括“机器翻译的最新发展：过去五年的回顾（Hutchins，1988）”，“Bemard Vauqois对构建MT系统的理论和实践的贡献：历史透视（Boitet，1988）“，《机器翻译：过去，现在，未来》（Hutchins，1988）。在机器翻译发展期间出现的论文包括《机器翻译的近期发展：对过去五年的回顾》（Hutchins，1988）， “Bemard Vauqois《对构建MT系统理论和实践的贡献：历史观点》（Boitet，1988）”，《机器翻译：过去，现在，未来》（Hutchins，1988）。

剩余内容已隐藏，支付完成后下载完整资料

资料编号：[24655]，资料为PDF文档或Word文档，PDF文档可免费转换为Word