论文总字数：35584字

摘 要

目前世界上已经有几十个机器翻译系统。其中法国、加拿大和美国成果较多。我们开展机器翻译研究已经几十年了，机器翻译的应用也在人们的生活中广泛应用，例如天气预报翻译，说明书翻译等。但是机器翻译仍然有些方面不尽人意，本文以《商务英语》教材为语料来源，利用现代翻译软件有道，对比机器翻译译文与书本上的人工翻译译本，分析机器翻译词汇错误类型，并提出可能处理方法。

关键词: 机器翻译; 错误分析; 词汇错误

Contents

1. Introduction 1

2. Literature Review 2

2.1 Studies on Machine Translation 2

2.2 Studies on Error Analysis 3

3. Analysis of Lexical Errors in Machine Translation -- A Case Study on the Sentences from Business English 5

3.1 Term 6

3.2 Conjunction 7

3.3 Part of Speech 8

3.4 Abbreviation 8

3.5 Replacement. 9

3.6 Missing Translation. 9

4. Suggestions on Machine Translation Errors 10

5 . Conclusion 11

Works Cited 12

1. Introduction

“Since A.D. Booth and W. Weaver put up first time the idea of using computers to translate one language into another language automatically in 1947,” (Feng Zhiwei, 2004:14) “machine translation has made great changes to the world. It is no exaggeration to say that nearly everyone’s life is linked in some way to machine translation directly or indirectly in the era of information age. So it is necessary to understand the basic knowledge of machine translation.” (Feng Zhiwei, 2004:2)

Machine Translation is known as automatic translation, which is an important aspect in the field of science and artificial. It means that people use computers to turn one natural language into another. Natural language generally refers to the natural language sentence and the text translation, which is a branch of Natural Language Processing. There is a close relationship between computational linguistics and natural language understanding. Mastering this technology is extremely significant to English research and learning.

“Machine translation means to use computers for natural language translation by new experimental disciplines. The subject arose in the early 1950s, in the mid - 1960 - s, once down, in the late 1960s to thrive again, now is still in development.”(Philipp Kehn, 2012: 3).

Machine translation studies not only relate to the study of language cognition of human, but also refer to the computer science, mathematics, information theory and knowledge of subjects which concerned with Machine Translation of Linguistics. The core of Machine Translation is the language analysis techniques.

Machine translation can divide into rule-based machine translation and corpus-based machine translation. Although machine translation has been put into practice and is playing a more and more important role in people’s life and learning, its translation quality are still suffering from the disease.

As a matter of fact, in order to overcome this major barrier and further improve the quality of machine translation. This paper will compare the translated sentences by You Dao online translation system and those by professional translators. Based on the analysis of translation errors of machine translation, some advices can be given on how to improve the quality of machine translation.

2. Literature Review

2.1 Studies on Machine Translation

“In 1993, famous British scholar W.J. Hitchins in MT Summit IV gave a speech and indicated that since 1989, the development of machine translation goes into a new age of translation, but knowledge system does not complete before translation corpus being used during translation”(ALPAC, 1966: 14). Huang Heyan and Chen Zhaoxiong in The field of basic machine translation study pointed out that “the research of machine translation is involved in linguistics, computer software, artificial intelligence and so on, and the requirement of knowledge is related with the basic field in artificial intelligence research. Some basic problems need to be further researched, including the disposal of basic linguistics and knowledge, and computer processing.”(Huang Heyanamp;Chen Zhaoxiong , 1997:2) In the literature web-based machine translation, Wang Haifeng and Wu Hua they dedicated one chapter to describing the “characteristics, challenges of web-based machine translation and the features described above, proposing the hybrid machine translation method, corpus mining and filtering methods, and approach to support the personalization of machine translation via the combination of translation technology and search technology.”(Wang Haifengamp; Wu Hua, 1990:11) Machine Translation has been widely applied, but not all people know the real meaning of machine translation until the publication of Machine Translation by Gao Li. He points out “machine translation system means the translation of one language into another language by computer software system. The real machine translation system should include two parts: dictionaries and grammars. And the system should be in control according to the algorithm design.”(Gao Li, 1991:1) At the same time more and more people put their mind on machine translation evaluation such as Zhang Jian and Wu Ji’s The improvement of Automatic Machine Translation Evaluation “Evaluation play a critical role in the machine translation. The research of automatic machine translation evaluation is an urgent need for the natural language processing researcher and developers.” (Zhang Jianamp;Wu Ji, 1998:1)

There are many different evaluation criteria of machine translation system. “It is generally acknowledged that the evaluation of machine translation system should pay more attention to translation quality or efficiency of translation during practical using”(AMTA, 1992: 56). However, if only considering these two factors, many other factors will be ignored. “One thing has to be noticed that many of the factors are the unique internal characteristics of machine translation system; they cannot be evaluated in a general standard. Usually, the evaluation criteria of machine translation system includes seven factors such as translation quality, efficiency during practical use, working style, operating environment, maintainability and expansibility, cost performance of machine translation system and robustness”(Arnold. D, et al, 1993: 95). For the moment, “machine translation system can only deal with simple sentences. Once it translates sentences with complex structure, the translation will have more or less problems.”（Zhou Jiongliangamp;Zhou Changle, 2011:34）

2.2 Studies on Error Analysis

Mistake is a kind of phenomenon in second language learning, and it is inevitable in the process of language. The error is a deviation of the standard of the target language. Carl James defines error as “unsuccessful language”. Errors in English are resulted from learner’s lack of target language knowledge.

There are various studies on machine translation. In 70"s，there was an error analysis regarding that mother tongue interference was the only source of errors, however, from the target language itself. The author thinks that Main causes of error is result from mother tongue. But now more scholars have error have given full play to their professional knowledge and skill, and tried to adopt different theories, frameworks and methods in error analysis study so as to make a detailed analysis of mother tongue interference.

“The error rate of syntax is lower than that of lexicon. Syntax refers to the interrelationships between elements in sentence structure. Syntax includes three relationships, namely, “positional relations, relations of substitutability, and relations of co-occurrence.”（Hu Zhuanglin, 2011:7）Of them all, positional relations are the most important one. If language wants to fulfill its communicative functions, it must have a way to mark the grammatical roles of the various phrases that can occur in a clause. (Hu Zhuanglin, 2011:7）Zi Weili in Error Analysis Theory and Error Classification puts forward the following domain errors in second language learning.( Zi Weili, 1998:2)

1. Negative transfer. This kind of the error expression generally exist in written and oral English.

2. Overgeneralization. Language learners summarized some rules of grammar because of the inappropriate generalization.

3. Hypercorrection. The discipline used in correcting one error is application into all the conditions.

4. Ignorance of rule restrictions. Some rules are applied to the inappropriate context.

5. The error of hypothetical concept. In the process of foreign language learning. Learners know a language rule which has made a hypothesis, but the hypothesis is false. The learner has formed a relatively complete language rule in the process of language learning and not have a correct language rules to use it correctly. This error occurs when people forget the rules temporarily.

Corded, the advocator of error analysis in Introducing Applied Linguistics, has put forward three simplified error types( Corded, 2005: 122).

1. Pre-systematic error. The errors are made by learners in communicative activities before their language system is formed.

2. Systematic errors. Learners can not fully understand or misunderstand the target language rules. Of course, in addition to the above common mistakes, some other errors called induced errors are worthy of paying attention .

3. Post-systematic error. The errors are resulted from negligence of the language rule they have mastered correctly after they have formed a relatively complete language system.

Influenced by other scholars, Li Mei and Zhu Qiming also began to make research on error analysis. In Error Patterns and statistical analyses of an English-Chinese machine translation corpus “The research identified the MT error patterns of a text corpus of 10,000 sentence pairs of a car maintenance manual and did statistical analysis of MT errors. Examples are used in this paper for illustrations of the error patterns. Statistical analysis reveals that the MT error reaches as high as over 80%, and among the MT errors lexical error stand at 70%.” (Li Mei amp;Zhu Qiming, 2013:9) At the same time, Xiao Wei also did statistical analysis on machine translation errors, Chinese-English machine translation is currently inseparable from human edition. (Xiao Wei, 2007:1)

In Translation Errors of Terms in Machine Translation, Luo Jimei offers a comprehensive analysis of the machine translation which reveals the statistical result of the translation errors of technical terms in machine translation output, and explains the three typical forms of errors in misidentifying the syntactical functions of technical terms serving as noun, verb or adjective and mistranslating abbreviations of technical terms. The research is useful to improve the translation system by narrowing down the usage of technical terms in the restricted area of auto-mobile technical documents.(Luo Jime, 2009:1)

Except for the error analysis, we also have known some English translation error types such as lexical errors, mistranslation of term, omission of translation, mistranslation of part of speech, mistranslation of conjunction, mistranslation of replacement of vocabulary.

Zhong Shangli pointed in The Human Intervention and Reliability of Machine Translation that “As an important technical way to break through the global language barrier, machine translation research and its software development already attract more attention, but what must be mentioned is that the target language is not readable after machine translation. This paper points out that human intervention is necessary to enhance the credibility of machine translation. Contextual constraints determine the purpose of machine translation and Machine-assisted translation.” (Zhong Shangli, 2004:5)

3. Analysis of Lexical Errors in Machine Translation -- A Case Study on the Sentences from Business English

The English Sentences and their translation in this research are taken from the textbook of Business English. Language materials in this book are all extracted from professional journals and books and have a high reliability and authenticity.

In Error Patterns and Statistical Analyses of English Chinese Machine Translation Corpus, Li Mei and Zhu Xing firstly classified 100000 E-C sentences (including phrases). Source text and standard translation were listed in pair in the Excel document. The Chinese version were given by professional translator who has long been engaged in translating materials on the car. So these versions were regarded as standard translation. Then, the source text were translated by machine. Therefore, it forms a parallel Comparison among sources text, machine translation and standard translation. Before the error analysis, 100 pair of sentences were selected from 100000 pairs of sentences as the example to be analyzed to get categories of machine translation errors which were listed in Table 1 .( Li Mei amp;Zhu Xing, 2013:2)

Table1.categories of MT Errors

Vocabulary errors is divided into 6 categories, including terms, conjunctions, part of speech, abbreviations, placement, missing translation etc. Sentences are made up by lexicon. The translation of lexicon plays an important role in machine translation. The quality of the translation of lexicon directly decides whether the translation is readable or not. Lexical errors takes the largest proportion of machine translation errors, So lexical errors translation is analyzed. The following are different types of lexical errors appearing in machine translation of sentences from Business English.

3.1 Term

Term: mistranslation of the term into the vocabulary of general meaning.( not limited to nouns).

Example 1：This Bill of Lading is issued in a negotiable form.

Machine translation: 这是议论形式的提单。

Standard Translation: 这是一份可转让提单。

In Business English, Bill of Lading it should be translated into a Business term with the meaning. So in translation, it should be translation into“可转让提单”.

Example 2: Our products have been endorsed by the National Quality Inspection Association.

Machine translation: 我们的产品已在国家质量检验协会背书。

Standard Translation: 我们的产品已通过国家质量检验协会推荐。

In the original sentence, “endorsed” just means “推荐”.The meaning of “背书”as a business term.

3.2 Conjunction

Conjunction: mistranslation of conjunction such as “and”, “but”, “however”

Example 3: Although your price is below our level, we accept your order in view of our initial business.

Machine translation: 虽然你们的价格低于我们的标准，我们在最初的业务接受你的订单。

Standard Translation: 虽然你的报价低于我们的额标准，鉴于第一次合作我们将接受你的订单。

In the above translation, machine translation failed to translate the intended meaning of the conjunction “although”. Potential content can not be accurately translate conveyed.

Example 4: The conclusion of the dealing is certainly not the ending, it is only the beginning and a good one, of the long and friendly business relationship between us.

Machine translation: 交易的结果当然不是结束，它仅仅是一个开始，一个好的，我们之间长期友好的业务关系。

Standard Translation: 这次达成的交易肯定不是结束，它不仅仅是一个开端，而是我们只见一个良好的，长期的，友好的贸易关系的开始。

In this sentence, machine translation did not display the function of “and”.

3.3 Part of Speech

Part of speech: mistranslation of the part of speech, such as noun into verb and noun into adjective.

Example 5: we have heard you are a good exporter textiles.

Machine translation ：我方了解您是一位好的纺织品出口商。

Standard Translation: 我方获悉贵方擅长纺织品出口。

In this sentence, “good” is translated into “好”, not just“擅长”.

Example 6 : we are interested to hear you are looking for a UK distributor for your teaching aids.

Machine translation:我们有兴趣听到你正在寻找你的英国教具分销商。

剩余内容已隐藏，请支付后下载全文，论文总字数：35584字