Automatic Translation System Research Articles

MARS: A Statistical Semantic Parsing and Generation-Based Multilingual Automatic tRanslation System

We present MARS (Multilingual Automatic tRanslation System), a research prototype speech-to-speech translation system. MARS is aimed at two-way conversational spoken language translation between English and Mandarin Chinese for limited domains, such as air travel reservations. In MARS, machine translation is embedded within a complex speech processing task, and the translation performance is highly effected by the performance of other components, such as the recognizer and semantic parser, etc. All components in the proposed system are statistically trained using an appropriate training corpus. The speech signal is first recognized by an automatic speech recognizer (ASR). Next, the ASR-transcribed text is analyzed by a semantic parser, which uses a statistical decision-tree model that does not require hand-crafted grammars or rules. Furthermore, the parser provides semantic information that helps further re-scoring of the speech recognition hypotheses. The semantic content extracted by the parser is formatted into a language-independent tree structure, which is used for an interlingua based translation. A Maximum Entropy based sentence-level natural language generation (NLG) approach is used to generate sentences in the target language from the semantic tree representations. Finally, the generated target sentence is synthesized into speech by a speech synthesizer. Many new features and innovations have been incorporated into MARS: the translation is based on understanding the meaning of the sentence; the semantic parser uses a statistical model and is trained from a semantically annotated corpus; the output of the semantic parser is used to select a more specific language model to refine the speech recognition performance; the NLG component uses a statistical model and is also trained from the same annotated corpus. These features give MARS the advantages of robustness to speech disfluencies and recognition errors, tighter integration of semantic information into speech recognition, and portability to new languages and domains. These advantages are verified by our experimental results.

Recognition of gestures in Arabic sign language using neuro-fuzzy systems

Hand gestures play an important role in communication between people during their daily lives. But the extensive use of hand gestures as a mean of communication can be found in sign languages. Sign language is the basic communication method between deaf people. A translator is usually needed when an ordinary person wants to communicate with a deaf one. The work presented in this paper aims at developing a system for automatic translation of gestures of the manual alphabets in the Arabic sign language. In doing so, we have designed a collection of ANFIS networks, each of which is trained to recognize one gesture. Our system does not rely on using any gloves or visual markings to accomplish the recognition job. Instead, it deals with images of bare hands, which allows the user to interact with the system in a natural way. An image of the hand gesture is processed and converted into a set of features that comprises of the lengths of some vectors which are selected to span the fingertips' region. The extracted features are rotation, scale, and translation invariat, which makes the system more flexible. The subtractive clustering algorithm and the least-squares estimator are used to identify the fuzzy inference system, and the training is achieved using the hybrid learning algorithm. Experiments revealed that our system was able to recognize the 30 Arabic manual alphabets with an accuracy of 93.55%.

Dictionary-Based Cross-Language Information Retrieval: Problems, Methods, and Research Findings

This paper reviews literature on dictionary-based cross-language information retrieval (CLIR) and presents CLIR research done at the University of Tampere (UTA). The main problems associated with dictionary-based CLIR, as well as appropriate methods to deal with the problems are discussed. We will present the structured query model by Pirkola and report findings for four different language pairs concerning the effectiveness of query structuring. The architecture of our automatic query translation and construction system is presented.

The Tower of Babel vs the Power of Babble

Since the dawn of human history, language differences have served as a barrier to full intercultural and international communication. The recent advent of synchronous, automatic translation systems (SATS), incorporated into the internet, are but the first sign of a communications revolution as profound as the invention of print. This article briefly surveys the following: (1) text/speech recognition and translation problems; (2) current developments in machine translation (MT) and artificial intelligence (AI) devoted to resolving these problems; and (3) potential future SATS technological developments and uses. The second half of the article is devoted to a wide-ranging analysis of several potentially profound future political, economic and cultural consequences of SATS. The main conclusion points to a paradox: although (perhaps because) SATS will lead to greater linguistic (and external-cultural) differentiation, the overall impact will be centripetal - greater integration among the world's peoples, more international peace, and a general higher level of agreement regarding norms and values.

Speech understanding and speech translation by maximum a-posteriori semantic decoding

This paper describes a domain-limited system for speech understanding as well as for speech translation. An integrated semantic decoder directly converts the preprocessed speech signal into its semantic representation by a maximum a-posteriori classification. With the combination of probabilistic knowledge on acoustic, phonetic, syntactic, and semantic levels, the semantic decoder extracts the most probable meaning of the utterance. No separate speech recognition stage is needed because of the integration of the Viterbi-algorithm (calculating acoustic probabilities by the use of Hidden-Markov-Models) and a probabilistic chart parser (calculating semantic and syntactic probabilities by special models). The semantic structure is introduced as a representation of an utterance's meaning. It can be used as an intermediate level for a succeeding intention decoder (within a speech understanding system for the control of a running application by spoken inputs) as well as an interlingua-level for a succeeding language production unit (within an automatic speech translation system for the creation of spoken output in another language). Following the above principles and using the respective algorithms, speech understanding and speech translating front-ends for the domains ‘graphic editor’, ‘service robot’, ‘medical image visualisation’ and ‘scheduling dialogues’ could be successfully realised.

Towards speech technology for South African languages: automatic speech recognition in Xhosa

This article reports on the first application of ‘state-of-the-art’ automatic speech recognition (ASR) technology to Xhosa, an indigenous South African language. The project, aimed at developing the technology required to implement a spoken-language front-end for an automatic translation system, is part of a collaborative effort in speech technology research for South African languages. The system we propose to perform this task consists of a speaker independent, isolated word, automatic speech recogniser. During system development speech data from 40 native. Xhosa speakers was recorded, annotated and organised into a database. Two systems were evaluated, one using hidden Markov word models and another implemented with hidden Markov phone models. Recognition rates of up to 95% were achieved.

It‘s Still the Proper Place

In the sixteen years that have elapsed since this paper was written, little has happened to change the overall assessment we should make of the proper place of humans and machines in language translation. The Eurotra project began shortly after that writing as a massive attempt to build a practical fully automatic translation system incorporating the best linguistic technology of the day. It ended lamely attempting to justify the expenditure of close to a hundred million ECU on the grounds that researchers even in the poorest European nations had learnt something of computational linguistics. As always, translation was treated as an entirely linguistic problem, and the parts of the project read like the chapters of an elementary linguistics textbook. In 1992, the German Federal Ministry for Research and Technology, presumably encouraged by recent achievements in translation and speech recognition, initiated the Verbmobil project, a program of research that would depend on major advances in both of these fields. The first of the two phases planned for this project, involving some 1 30 researchers from Germany and abroad, is just coming to an end and the second phase, involving only 100 researchers, is about to begin. At a cost of some 60 million marks, the project has reached the point where it can interpret a small number of short sentences chosen from a predetermined corpus. The design of the project is essentially identical to that of its predecessors and it would be truly remarkable if it produced interestingly different results. Government and bureaucracy do indeed appear to be imbued with a sad fatalism that forces upon them a future that repeats the follies of the past. Verbmobil is especially disheartening for the great opportunities that it offers but that are not being pursued. The aim is to provide interpretation services for the participants in a face-to-face conversation so that the originators of the material it must translate can be assumed to be available to help clarify their intentions. However, only a vanishingly small proportion of the enormous resources of the project are devoted to designing ways in which this might be achieved. While the differences between past and present translation projects are minuscule relative to what would be required to affect substantially the value of the resulting system, it is of course by no means the case that they all start on an even footing. Computers are faster; the art of programming advances; storage is cheaper; largexorpora of text are more readily available; new algorithms are constantly

Feature-based & model-based semantics for English, French and German verb phrases

This paper considers the relative merits of using features and formal event models to characterise the semantics of English, French and German verb phrases, and considers the application of such semantics in machine translation. The feature based approach represents the semantics in terms of feature systems, which have been widely used in computational linguistics for syntactic representation of complex systems. The paper shows how a simple intutive semantics of verb phrases may be encoded as a feature system, and how this can be used to support modular construction of automatic translation systems through feature look-up tables. One such translator, from English to French and German, is described. The paper continues to formalise the feature-based approach via a model-based, Montague semantics, which extends previous work on the semantics of English verb phrases. In so doing, repercussions of and to this framework in conducting a contrastive semantic study are considered. The model-based approach also promises to provide support for a more sophisticated approach to translation through logical proof, the paper indicates further work required for the fulfilment of this promise.

A low-cost automatic translation and autofocusing system for a microscope

In this design note we describe a low-cost system for scanning the translation stage and automatically maintaining the correct focal position of a microscope. The image signal is derived from a video camera and the focal position determined by measuring the energy in the edge-detected image. The microscope is a conventional laboratory microscope with stepping motors attached to actuate the stage and control is effected through a laboratory personal computer.

The Fortran‐P Translator: Towards Automatic Translation of Fortran 77 Programs for Massively Parallel Processors

Massively parallel processors (MPPs) hold the promise of extremely high performance that, if realized, could be used to study problems of unprecedented size and complexity. One of the primary stumbling blocks to this promise has been the lack of tools to translate application codes to MPP form. In this article we show how applications codes written in a subset of Fortran 77, called Fortran‐P, can be translated to achieve good performance on several massively parallel machines. This subset can express codes that are self‐similar, where the algorithm applied to the global data domain is also applied to each subdomain. We have found many codes that match the Fortran‐P programming style and have converted them using our tools. We believe a self‐similar coding style will accomplish what a vectorizable style has accomplished for vector machines by allowing the construction of robust, user‐friendly, automatic translation systems that increase programmer productivity and generate fast, efficient code for MPPs.

Ma?innyj perevod i prikladnaja lingvistika: Problemy sozdanija sistemy avtomati?eskogo perevoda

This work, the title of which means 'Machine Translation and Applied Linguistics: Problems in the Development of Automatic Translation Systems,' and which is published by the Pedagogical Institute for Foreign Languages in Moscow, is a collection of papers on the French-Russian MT project of the Center for Translation in Moscow. The project is known by its Russian acronym frap, for Francuzsko-Russkij Avtomatides/dj Perevod ('French-Russian Automatic Translation'). The volume is a fairly thorough discussion of theoretical and methodological issues related to this system which, until now, has been described in the literature in only very general terms. Eight papers in the volume discuss various aspects of FRAP. The introductory article by N.N. Leont'eva, Senior Research Supervisor of the project, describes the frap information processing and translation system and the need for a component in MT systems that allows for a transition to a different semantic domain during transfer. A paper by E.E. Lovtskij and L. Ju. Korostelev emphasizes the programming rather than the linguistic point of view in the initial analysis of the source language. E.G. Sokolova discusses formal syntactic representation in terms of functional grammar. Initial and final syntactic representations of a sentence are discussed, as are some questions of syntactic ambiguity. Three of the papers focus on semantics, and they are the most intriguing in this volume. I.M. KundrjaSova describes the semantic dictionary of the system and its role in semantic analysis. The semantic dictionary is also the subject of a short paper by A.E. GureviC, who discusses the computational semantics of parametric nouns. The third, by Leont'eva, KundrjaSova, and O.B. MaleviC, discusses sentence-bound semantic analysis and summarizes some experimental work in semantic interpretation of syntactic structures. The authors posit that the aggregation of semantic representations of the sentences in a text can be regarded

Traduction automatique et recherche cognitive

ABSTRACT : Machins translation (MT) already has a history of about forty years. The well-known alpac report, which slowed down MT Research and Development (R and D) was preceded by a long period after the enthousiastic period (before alpac) and followed by another long period prior to the return to R and D (post alpac period). Despite considerable progress, the present automatic translation (AT) systems have their limits. The effect of the alpac report has been to isolate traditional AT from other sectors of doing research in natural language (NL) and especially artificial intelligence (AI). For AI researchers, a good quality AT is possible only by a cognitive approach : they develop small experimental systems within limited range to test their theories of comprehension of NL.

Converting text into speech in real time with microcomputers

Speech synthesizers have led to the evolution of many talking machines, which are now being used as automated text readers for the blind, communication aids for the nonvocal, and as a means of transmitting information from a database to remote locations. Where machine operators are fully using their eyes and hands to perform delicate operations, synthetic speech has proved invaluable in providing feedback from the machine to the operator. In this paper, the design method of a modest system for automatic translation of unlimited-vocabulary English text to synthetic speech is described in detail, as well as suggestions on possible applications of the system.

Ontwikkelingen en Problemen Bij Automatisch Vertalen

In the fifties and sixties a great deal of money and time was spent on the development of automatic translation systems. Because of the meagre results and sombre perspectives all large scale research projects were abandoned. The eighties show a renewed interest in the field of automatic translating. It is, therefore, a valid question whether old problems have been solved and new perspectives have been uncovered. A critical analysis shows that progress has been made in the domains of hardware (speed of data processing and data storage) and the organisation of data. The basic linguis-tic problems have not, however, been solved. Linguistic research that is not first and foremost directed at descriptive (instead of representational) problems and at fundamental problems such as the relation between syntax, semantics and knowledge of the world is subject to the same fate as the research into automatic translating in the fifties.

A telemetry automatic reduction system — Take II

The word TARE II is coined from the words Telemetry Automatic Reduction Equipment. TARE II is essentially an automatic translation system developed to convert raw telemetry data into a form which can be read into an electronic digital computer for subsequent computations. Specifically, the digital output format of TARE II was designed for compatibility with the Florida Automatic Computer (FLAG), the central component of the AFMTC data reduction facility. The computer can then be programmed to separate, reduce, linearize, scale, and tabulate the data, in addition to preparing input tapes for automatic plotting and presentation of the data.