Target-Text Mediated Interactive Machine Translation

The use of Machine Translation as a tool for professional or other highly skilled translators is for the most part currently limited to postediting arrangements in which the translator invokes MT when desired and then manually cleans up the results. A theoretically promising but hitherto largely unsuccessful alternative to postediting for this application is interactive machine translation (IMT), in which the translator and MT system work in tandem. We argue that past failures to make IMT viable as a tool for skilled translators have been the result of an infelicitous mode of interaction rather than any inherent flaw in the idea. As a solution, we propose a new style of IMT in which the target text under construction serves as the medium of communication between an MT system and its user. We describe the design, implementation, and performance of an automatic word completion system for translators which is intended to demonstrate the feasibility of the proposed approach, albeit in a very rudimentary form.     

Power shifts in web-based translation memory

Web-based translation memory (TM) is a recent and little-studied development that is changing the way localisation projects are conducted. This article looks at the technology that allows for the sharing of TM databases over the internet to find out how it shapes the translator’s working environment. It shows that so-called pre-translation—until now the standard way for clients to manage translation tasks with freelancers—is giving way to web-interactive translation. Thus, rather than interacting with their own desktop databases as before, translators now interface with each other through server-based translation memories, so that a newly entered term or segment can be retrieved moments later by another translator working at a remote site. The study finds that, while the interests of most stakeholders in the localisation process are well served by this web-based arrangement, it can involve drawbacks for freelancers. Once an added value, technical expertise becomes less of a determining factor in employability, while translators lose autonomy through an inability to retain the linguistic assets they generate. Web-based TM is, therefore, seen to risk disempowering and de-skilling freelancers, relegating them from valued localisation partners to mere servants of the new technology.     

Unit Completion for a Computer-aided Translation Typing System

This work is in the context of, a system thatwatches over the users as they type a translation andrepeatedly suggests completions for the text already entered.The users may either accept, modify, or ignore these suggestions. Wedescribe the design, implementation, and performance of aprototype which suggests completions of units of texts that arelonger than one word.     

A Methodology for Building Translator-oriented Dictionary Systems

Interaction between humans and translation tools has been deeply studied in the field of machine-aided translation. However, support tools for translation are often designed without the co-operation of human translators. The underlying idea is that human translators must adapt to the new technologies, and it seems that new computerised tools would not need to consider translators' practical use and experience. On the contrary, we argue that it is worthwhile and necessary to analyse the behaviour of translators in order to fit the tools to their needs. This paper presents an experiment to incorporate human translators' expertise into an already constructed lexical system. We focus on the design methodology that could be applied for the improvement of other similar tools.     

Maximum Entropy Modeling: A Suitable Framework to Learn Context-Dependent Lexicon Models for Statistical Machine Translation

Current statistical machine translation systems are mainly based on statistical word lexicons. However, these models are usually context-independent, therefore, the disambiguation of the translation of a source word must be carried out using other probabilistic distributions (distortion distributions and statistical language models). One efficient way to add contextual information to the statistical lexicons is based on maximum entropy modeling. In that framework, the context is introduced through feature functions that allow us to automatically learn context-dependent lexicon models.In a first approach, maximum entropy modeling is carried out after a process of learning standard statistical models (alignment and lexicon). In a second approach, the maximum entropy modeling is integrated in the expectation-maximization process of learning standard statistical models.Experimental results were obtained for two well-known tasks, the French–English Canadian Parliament Hansards task and the German–English Verbmobil task. These results proved that the use of maximum entropy models in both approaches, can help to improve the performance of the statistical translation systems.This work has been partially supported by the European Union under grant IST-2001-32091 and by the Spanish CICYT under project TIC-2003-08681-C02-02. The experiments on the Verbmobil task were done when the first author was a visiting scientist at RWTH Aachen-Germany.Editors: Dan Roth and Pascale Fung     

Constructive machine translation evaluation

When surveying the many methods currently employed in MT evaluation,¹ it is not immediately obvious that the methods used serve to increase the knowledge of the properties being measured. This report describes aconstructive machine translation evaluation method, aimed at addressing this issue.² Edited version of a presentation given to the International Working Group on the Evaluation of Machine Translation Systems, Vaud, Switzerland, April 1991.     

The Long-Term Forecast for Weather Bulletin Translation

Machine Translation (MT) is the focus of extensive scientific investigations driven by regular evaluation campaigns, but which are mostly oriented towards a somewhat particular task: translating news articles into English. In this paper, we investigate how well current MT approaches deal with a real-world task. We have rationally reconstructed one of the only MT systems in daily use which produces high-quality translation: the Météo system. We show how a combination of a sentence-based memory approach, a phrase-based statistical engine and a neural-network rescorer can give results comparable to those of the current system. We also explore another possible prospect for MT technology: the translation of weather alerts, which are currently being translated manually by translators at the Canadian Translation Bureau.     

Axioms for control operators in the CPS hierarchy

A CPS translation is a syntactic translation of programs, which is useful for describing their operational behavior. By iterating the standard call-by-value CPS translation, Danvy and Filinski discovered the CPS hierarchy and proposed a family of control operators, shift and reset, that make it possible to capture successive delimited continuations in a CPS hierarchy. Although shift and reset have found their applications in several areas such as partial evaluation, most studies in the literature have been devoted to the base level of the hierarchy, namely, to level-1 shift and reset. In this article, we investigate the whole family of shift and reset. We give a simple calculus with level-n shift and level-n reset for an arbitrary n>0. We then give a set of equational axioms for them, and prove that these axioms are sound and complete with respect to the CPS translation. The resulting set of axioms is concise and a natural extension of those for level-1 shift and reset. This article is a revised version of the paper presented at 18th International Workshop on Computer Science Logic (CSL’04), September, 2004.     

融合被动和可能态模型的日汉统计机器翻译

日语中谓词语态有不同的词尾变形,其中被动态和可能态具有相同的词尾变化,在统计机器翻译中难以对其正确区分及翻译。因此,该文提出一种利用最大熵模型有效地对日语可能态和被动态进行分类,然后把日语的可能态和被动态特征有效地融合到对数线性模型中改进翻译模型的方法,以提高可能态和被动态翻译规则选择的准确性。实验结果表明,该方法可以有效提升日语可能态和被动态句子的翻译质量,在大规模日汉语料上,最高翻译BLEU值能够由41.50提高到42.01,并在人工评测中,翻译结果的整体可理解度得到了2.71%的提升。     

Head-Transducer Models for Speech Translation and Their Automatic Acquisition from Bilingual Data

This article presents statistical language translation models,called dependency transduction models, based on collectionsof head transducers. Head transducers are middle-out finite-state transducers which translate a head word in a source stringinto its corresponding head in the target language, and furthertranslate sequences of dependents of the source head into sequencesof dependents of the target head. The models are intended to capturethe lexical sensitivity of direct statistical translation models,while at the same time taking account of the hierarchical phrasalstructure of language. Head transducers are suitable for directrecursive lexical translation, and are simple enough to be trainedfully automatically. We present a method for fully automatictraining of dependency transduction models for which the only inputis transcribed and translated speech utterances. The method has beenapplied to create English–Spanish and English–Japanese translationmodels for speech translation applications. The dependencytransduction model gives around 75% accuracy for an English–Spanishtranslation task (using a simple string edit-distance measure) and70% for an English–Japanese translation task. Enhanced with targetn-grams and a case-based component, English–Spanish accuracy is over76%; for English–Japanese it is 73% for transcribed speech, and60% for translation from recognition word lattices.     

EXTRA: a system for example-based translation assistance

In this paper we present EXTRA (EXample-based TRanslation Assistant), a translation memory (TM) system. EXTRA is able to propose effective translation suggestions by relying on syntactic analysis of the text and on a rigorous, language-independent measure; the search is performed efficiently in large amounts of bilingual texts thanks to its advanced retrieval techniques. EXTRA does not use external knowledge requiring the intervention of users and is completely customizable and portable as it has been implemented on top of a standard DataBase Management System. The paper provides a thorough evaluation of both the effectiveness and the efficiency of our system. In particular, in order to quantify the benefits offered by EXTRA assisted translation over manual translation, we introduce a simulator implementing specifically devised statistical, process-oriented, discrete-event models. As far as we know, this is the first time statistical simulation experiments have been used to face the nontrivial problem of evaluating TM systems, particularly for comparing the time that could be saved by performing assisted translation versus “manual” translation and for optimally tuning the system behaviour with respect to differently skilled users. In our experiments, we considered three scenarios, manual translation with one or two translators and assisted translation with one translator. The time needed for one translator to do an assisted translation is significantly closer to that of a team of two translators than to that of the single translator. The mean sentence translation time is by far the lowest for this scenario, corresponding to the highest per translator productivity. We also estimate the total translation time when the number of query sentences, the maximum number of suggestions to be read, and the probability of look up are varied: the best trade-off is given by reading (and presenting) four or five suggestions at the most.     

统计机器翻译综述

本文综述了基于信源信道思想和基于最大熵思想的统计机器翻译方法并介绍了统计机器翻译的评测方法。基于信源信道的方法将翻译概率表示为一个语言模型和一个翻译模型。而基于最大熵的方法则是利用一系列实数值特征函数的线性组合来求解最优的译文。基于最大熵的统计机器翻译方法比基于信源信道的方法更具有一般性,后者可以看做前者的一个特例。     

Automated Derivation of Translators From Annotated Grammars

In this paper we propose a technique to automate the process of building translators between operations languages, a family of DSLs used to program satellite operations procedures. We exploit the similarities between those languages to semi-automatically build a transformation schema between them, through the use of annotated grammars. To improve the overall translation process even more, reducing its complexity, we also propose an intermediate representation common to all operations languages. We validate our approach by semi-automatically deriving translators between some operations languages, using a prototype tool which we implemented for that purpose.     

2005统计机器翻译研讨班研究报告

2005年7月13日至15日,中国科学院自动化研究所、计算技术研究所和厦门大学计算机系联合举办了我国首届统计机器翻译研讨班。本文主要介绍本次研讨班参加单位的测试系统和实验结果,并给出相应的分析。测试结果表明,我国的统计机器翻译研究起步虽晚,但已有快速进展,参评系统在短期内得到了较好的翻译质量,与往年参加863评测的基于规则方法的系统相比性能虽还有差距,但差距已经不大。从目前国际统计机器翻译研究的现状和发展趋势来看,随着数据资源规模的不断扩大和计算机性能的迅速提高,统计机器翻译还有很大的发展空间。在未来几年内,在基于短语的主流统计翻译方法中融入句法、语义信息,必将成为机器翻译发展的趋势。     

An object-oriented operation-based approach to translation between MOF metaschemas

This paper proposes a new approach to the schema translation problem. We deal with schemas whose metaschemas are instances of the OMG’s MOF. Most metaschemas can be defined as an instance of the MOF; therefore, our approach is widely applicable. We leverage the well-known object-oriented concepts embedded in the MOF and its instances (object types, attributes, relationship types, operations, IsA hierarchies, refinements, invariants, pre- and postconditions, etc.) to define metaschemas, schemas and their translations.The main contribution of our approach is the extensive use of object-oriented concepts in the definition of translation mappings, particularly the use of operations (and their refinements) and invariants, both of which are formalized in OCL. Our translation mappings can be used to check that two schemas are translations of each other, and to translate one into the other, in both directions. The translation mappings are declaratively defined by means of pre- and postconditions and invariants, and they can be implemented in any suitable language. From an implementation point of view, by taking a MOF-based approach we have a wide set of tools available, including tools that execute OCL. By way of example, we have defined all schemas and metaschemas in this paper and executed all the OCL expressions in the USE tool.     

A process study of computer-aided translation

We investigate novel types of assistance for human translators, based on statistical machine translation methods. We developed the computer-aided tool Caitra that makes suggestions for sentence completion, shows word and phrase translation options, and allows postediting of machine translation output. We carried out a study of the translation process that involved non-professional translators that were native in either French or English and recorded their interaction with the tool. Users translated 192 sentences from French news stories into English. Most translators were faster and better when assisted by our tool. A detailed examination of the logs also provides insight into the human translation process, such as time spent on different activities and length of pauses.     

一种基于分类的平行语料选择方法

大规模高质量双语平行语料库是构造高质量统计机器翻译系统的重要基础,但语料库中的噪声影响着统计机器翻译系统的性能,因此有必要对大规模语料库中语料进行筛选。区别于传统的语料选择排序模型,本文提出一种基于分类的平行语料选择方法。通过少数句对特征构造差异较大的分类器训练句对,在该训练句对上使用更多的句对特征对分类器进行训练,然后对其他未分类句对进行分类。相比于基准系统,我们的方法不仅缩减40%训练语料规模,同时在NIST测试数据集合上将BLEU值提高了0.87个百分点。     

基于数据扩充的翻译记忆库与神经机器翻译融合方法

神经机器翻译是目前机器翻译领域的主流方法,而翻译记忆是一种帮助专业翻译人员避免重复翻译的工具,其保留之前完成的翻译句对并存储在翻译记忆库中,进而在之后的翻译过程中通过检索去重用这些翻译。该文基于数据扩充提出两种将翻译记忆与神经机器翻译相结合的方法:(1)直接拼接翻译记忆在源语句后面;(2)通过标签向量拼接翻译记忆。该文在中英与英德数据集上进行了实验,实验表明,该方法可以使翻译性能获得显著提升。     

Statistical Translation of Text and Speech: First Results with the RWTH System

In this paper, we describe a first version of a system for statisticaltranslation and present experimental results. The statistical translationapproach uses two types of information: a translation model and a languagemodel. The language model used is a standard bigram model. The translationmodel is decomposed into lexical and alignment models. After presenting the details of the alignment model, we describe the search problem and present a dynamic programming-based solution for the special case of monotone alignments.So far, the system has been tested on two limited-domain tasks for which abilingual corpus is available: the EuTrans traveller task (Spanish–English,500-word vocabulary) and the Verbmobil task (German–English, 3000-wordvocabulary). We present experimental results on these tasks. In addition to the translation of text input, we also address the problem of speech translation and suitable integration of the acoustic recognition process and the translation process.