An overview of MiMo2

The MiMo2 translation system combines several leading ideas in the areas of linguistics, computation and translation. In the area of translation we follow the ideas of Landsbergen (1987) by assuming that translation is symmetric; and combine these ideas with the advantages of a transfer approach. Computationally the system focuses on computability and declarativity. The linguistics of the system is based on a lexicalistic and sign-based approach to grammar.MiMo2 is a sideline project of the Dutch Eurotra team in Utrecht. An extended version of this article was presented at the machine translation workshop in Austin, Texas, June 1990.     

基于分层语块分析的统计翻译研究

本文描述了一个基于分层语块分析的统计翻译模型。该模型在形式上不仅符合同步上下文无关文法,而且融合了基于条件随机场的英文语块分析知识,因此基于分层语块分析的统计翻译模型做到了将句法翻译模型和短语翻译模型有效地结合。该系统的解码算法改进了线图分析的CKY算法,融入了线性的N-gram语言模型。目前,本文主要针对中文－英文的口语翻译进行了一系列实验,并以国际口语评测IWSLT(International Workshop on Spoken Language Translation)为标准,在2005年的评测测试集上,BLEU和NIST得分均比统计短语翻译系统有所提高。     

动态二进制翻译器CrossBit的性能分析与评估

动态二进制翻译是广泛应用于虚拟机系统的一种二进制代码的翻译技术。动态二进制翻译由于拥有代码缓存、本地执行、代码块链接、动态热路径生成等优化技术的支持,有着很高的性能。CrossBit是一个多元多目标的动态二进制翻译系统,通过对CrossBit二进制翻译器的性能进行的研究,分析动态二进制翻译器性能提升中所必须解决的若干问题,并通过定量的分析总结了一些二进制翻译系统的在不同的配置和负载下系统优化手段的执行时策略。     

机器翻译评测的新进展

机器翻译评测对机器翻译的研究和开发具有至关重要的作用,对其的研究一直是国内外机器翻译界的重点课题。本文首先全面地介绍了最近出现的而且受到极大关注的机器翻译评测技术,即IBM公司的BLEU机器翻译评测标准和NIST采用的机器翻译评测技术。实验表明,自动翻译评测技术能够接近人工评价,评测结果也是可接受的。因此,采用自动翻译评测技术能够给自然语言处理的研究人员和开发人员带来很大的便利性。本文还展示了一个开放式的可扩展的自动翻译评测的平台,完全实现了BLEU和NIST评测标准,并做出了一定的改进使得该系统具有良好的使用性和可扩展性。     

基于GAN的手绘草图图像翻译研究综述

目前基于GAN的手绘草图图像翻译工作处于起步阶段,分析了草图图像翻译面临的挑战性问题,从无控制和精细控制的草图图像翻译两个方面对基于GAN的草图图像翻译研究工作进行分析,并对生成图像的评估方法进行总结。基于对已有研究工作的总结归纳,对该领域未来可能的发展趋势进行了展望,为该领域研究人员拓展研究思路提供了线索。     

Transfer formalisms

We discuss some different ideas about the sort of formalism appropriate for translation (more specifically, transfer) that have been explored within Eurotra. The intention is not to evaluate these different ideas, but to outline the issues of general interest and relevance that they raise. Section 1 describes some common assumptions, section 2 describes the different ways they have been interpreted, and section 3 considers some further implications.     

Developing integrated environments with ASDL

The authors present the Abstract-Type and Scheme-Definition Language (ASDL), a new formalism and environment kernel. The main goal of this object-oriented kernel is to reconcile integration and extensibility. ASDL adopts the ideas underlying language-based editor generators but overcomes their extensibility deficiencies. ASDL combines an object-oriented type system with syntax-directed translation schemes and a target-language interface. Its power lies in utilizing inheritance for structure-driven computations. The authors discuss object types, translation rules, and semantic information in ASDL     

English-Thai Translation: Initial Experiments with a Multiphase Translation System

A multiphase machine translation approach, Generate and Repair Machine Translation (GRMT), is proposed. GRMT is designed to generate accurate translations that focus primarily on retaining the linguistic meaning of the source language sentence. GRMT presently incorporates a limited multilingual translation capability. The central idea behind the GRMT approach is to generate a translationcandidate (TC) by quick and dirty machine translation (QDMT), then investigate the accuracy of that TC by translation candidate evaluation (TCE), and, if necessary, revise the translation in the repair and iterate (RI) phase. To demonstrate the GRMT approach, a translation system that translates from English to Thai has been developed. This paper presents the design characteristics and some experimental results of QDMT and also the initial design, some experiments, and proposed ideas behind TCE and RI.     

基于变分信息瓶颈的半监督神经机器翻译

变分方法是机器翻译领域的有效方法, 其性能较依赖于数据量规模. 然而在低资源环境下, 平行语料资源匮乏, 不能满足变分方法对数据量的需求, 因此导致基于变分的模型翻译效果并不理想. 针对该问题, 本文提出基于变分信息瓶颈的半监督神经机器翻译方法, 所提方法的具体思路为: 首先在小规模平行语料的基础上, 通过引入跨层注意力机制充分利用神经网络各层特征信息, 训练得到基础翻译模型; 随后, 利用基础翻译模型, 使用回译方法从单语语料生成含噪声的大规模伪平行语料, 对两种平行语料进行合并形成组合语料, 使其在规模上能够满足变分方法对数据量的需求; 最后, 为了减少组合语料中的噪声, 利用变分信息瓶颈方法在源与目标之间添加中间表征, 通过训练使该表征具有放行重要信息、阻止非重要信息流过的能力, 从而达到去除噪声的效果. 多个数据集上的实验结果表明, 本文所提方法能够显著地提高译文质量, 是一种适用于低资源场景的半监督神经机器翻译方法.     

Simulation of English translation text filtering based on machine learning and embedded system

The translation is generally assumed to distinguish it from the original language text in the same area. Therefore, say these constitute a unique multilingual translation, commonly referred to as translation; translation is also affected by the source language and exhibits different characteristics according to the source language. Therefore, claim that these variants constitute the same target language translated into additional "dialog." Using Machine Learning English and embedded technologies investigated the differences between the general characteristics of different translation sources and language translation. There is little research corpus between the complicated relationship between translation and the original text and translation itself for very different language types. May does not translate enough knowledge of the subject areas covered. There are many guidelines to help authors clearly express their ideas to promote translation; scientists and engineers find it difficult to apply the procedures required for a high degree of speech recognition. For financial reasons, non-essential text may be edited publication. In this case, as described in the article, the author can take some precautions to express terms of intelligibility. Machine Learning and embedded systems determine the importance of various features for data collection and the importance of its characteristics compared with previously reported studies in English. Also, the method allows us to add the grammatical function of translation studies rarely used. The results show that, even if only based on five features, the full translation can be reliably separated from the non-translated region.     

The engineering of a Translator Workstation

The design of a Translator Workstation is an engineering task involving technical aspects like dictionary organization/retrieval and natural language processing, as well as human aspects like friendly user interface and user accessibility. This paper describes the design and implementation of such a workstation for translating English texts into Malay. Although specifically developed for English-Malay translation work, most of the ideas and features presented are language independent. Important issues raised include translation modes and functions, dictionary and thesaurus organization, morphological processing, spelling check and special word processing features.     

面向神经机器翻译的枢轴方法研究综述

低资源语言的神经机器翻译（neural machine translation,NMT）一直是机器翻译领域研究的难点和热点,基于枢轴的方法为其性能的提升和改进提供了思路。针对枢轴思想在低资源语言神经机器翻译中的应用,从枢轴翻译、基于枢轴的伪平行数据生成和基于枢轴的模型构建三方面,对不同方法的国内外研究现状、主要问题和趋势进行了分析和比较,为该领域的研究提供参考和借鉴。     

基于改进GLR算法的智能识别英语翻译模型设计

GLR算法模型翻译识别结果存在数据点重合的情况,精确度无法得到有效保障。为了准确的识别短语,设计了基于改进GLR算法的短语智能识别算法,该算法构建标记规模约74万个英汉单词的短语语料库,使短语具备可搜索功能,通过短语中心点构建短语结构,可获得词性识别结果,依据解析线性表的句法功能校正词性识别结果中的英汉结构歧义,最终获得识别的内容。实际测评结果显示,该算法克服了GLR的弊端,相对统计算法和动态记忆算法提高了运算速度和处理性能,更加适合机器翻译任务,为在智能机器翻译领域提供了新的思路。     

A Case Study in Program Transformation: Translation into Polish

Program transformation is used to show the correspondence between two known algorithms for translating parenthesized expressions into their polish counterparts. The first algorithm is a syntax-directed translator which produces output upon the recognition of predefined syntactic rules. The second one, suggested by Dijkstra, is of empirical nature: it performs the translation by simulating the operation of a railyard shunt. The second algorithm is more efficient than the first, but less general. This work enables one 1) to assert the correctness of the empirical algorithm, 2) to compare the relative efficiencies of the two algorithms, and 3) to gain insight into the circumstances in which syntax-directed translation may be improved by generalizing the ideas involved in Dijkstra's algorithm.     

Validating More Loop Optimizations

Translation validation is a technique for ensuring that a translator, such as a compiler, produces correct results. Because complete verification of the translator itself is often infeasible, translation validation advocates coupling the verification task with the translation task, so that each run of the translator produces verification conditions which, if valid, prove the correctness of the translation. In previous work, the translation validation approach was used to give a framework for proving the correctness of a variety of compiler optimizations, with a recent focus on loop transformations. However, some of these ideas were preliminary and had not been implemented. Additionally, there were examples of common loop transformations which could not be handled by our previous approaches.This paper addresses these issues. We introduce a new rule Reduce for loop reduction transformations, and we generalize our previous rule Validate so that it can handle more transformations involving loops. We then describe how all of this (including some previous theoretical work) is implemented in our compiler validation tool TVOC.     

Formal Languages: Origins and Directions

Origins of the theory of formal languages and automata are surveyed starting from 1936 with the work of Turing and Post. Special attention is given to the machine translation projects of the 1950s and early 1960s and associated work in mathematical linguistics. The development of the Chomsky hierarchy of grammars, machines, and languages from 1956 to 1964 is traced. It is observed that the same important ideas emerged independently for the automatic analysis and translation of both natural and artificial languages. Since 1964, formal language theory is part of theoretical computer science. A few of the directions since 1964 are considered: restrictions and extensions of context-free grammars and pushdown store automata, unifying frameworks, and complexity questions.     

理性主义与经验主义相结合的机器翻译研究策略

主要介绍了基于规则、基于实例和基于统计等3种主流机器翻译方法,探讨了自然语言处理技术和机器翻译中基于规则的理性主义方法和基于统计的经验主义方法的优缺点,结合机器翻译研究的现状和发展方向,提出了规则和统计相结合的机器翻译方法的基本思路,阐述了词义消歧中的理性主义方法和经验主义方法相结合的发展方向,对机器翻译的发展趋势进行了探讨。     

Towards an analyzer (parser) in a machine translation system based on ideas from expert systems

GETA (Groupe d'études pour la traduction automatique) is a research team working basically in the domain of machine translation. GETA's software system, ARIANE-78, has been tested over various pairs of relatively unrelated languages. Being a product of the late seventies, the system misses out on some of the artificial intelligence technology, particularly that of the eighties. Nevertheless, GETA carries out some research related to artificial intelligence within a general effort to bring improvements to the current system. This paper reports on an effort to embed such work within the framework of an entirely new system based on ideas from expert systems, significantly departing from the methodology of the current system (and that of other currently implemented machine translation systems). The proposed architecture aims for total modularity and flexibility and some degree of intelligence.     

TURN服务器原型系统的设计与实现

TURN协议是一种穿越对称NAT的技术。以RFC3489技术为基础,对TURN协议草案进行了深入研究与分析,并在此基础上,改进和简化了草案中存放动态分配地址的地址映射表结构,设计了TURN技术的工作方式和应用模型。然后,借鉴了STUN方式的设计思想,设计并实现了TURN服务器原型系统,解决了SIP UA在STUN等方式下不能穿越对称性NAT问题。