Machine Translation Summit (1991)
This paper outlines a new architecture for a NLP/MT development environment for the EUROTRA project, which will be fully operational in the 1993-94 time frame. The proposed architecture provides a powerful and flexible platform for extensions and enhancements to the existing EUROTRA translation philosophy and the linguistic work done so far, thus allow- ing the reusability of existing grammatical and lexical resources, while ensuring the suitability of EUROTRA methods and tools for other NLP/MT system developers and researchers.
IBM is engaged in advanced research and development projects on various aspects of machine translation, between several language pairs. The activities reported on hero are all parts of a rather large-scale, international effort, following Michael McCord’s LMT approach. The paper focuses on seven selected topics: recent enhancements made in the Slot Grammar formalism and the specific analysis components; specification of a semantic type hierarchy and its use for verb sense disambiguation; incorporation of statistical techniques in the translation process; anaphora resolution; linkage of target morphology modules; methods for the construction of large MT lexicons; and interactive disambiguation.
ULTRA (Universal Language TRAnslator) is a multilingual, interlingual machine translation system currently under development at the Computing Research Laboratory at New Mexico State University. It translates between five languages (Chinese, English, German, Japanese, Spanish) with vocabularies in each language based on approximately 10,000 word senses. The major design criteria are that the system be robust and general purpose with simple to use utilities for customization to suit the needs of particular users. This paper describes the central characteristics of the system: the intermediate representation, the language components, semantic and pragmatic processes, and supporting lexical entry tools.
We describe an interlingua-based approach to machine translation, in which a DRS representation of the source text is used as the interlingua representation. A target DRS is then created and used to construct the target text. We describe several advantages of this level of representation. We also argue that problems of translation mismatch and divergence should properly be viewed not as translation problems per se but rather as generation problems, although the source text can be used to guide the target generator. The system we have built relics exclusively on monolingual linguistic descriptions that are also, for the most part, bi-directional.
The ArchTran English-Chinese Machine Translation System is among the first commercialized English-Chinese machine translation systems in the world. A prototype system was released in 1989 and currently serves as the kernel of a value-added network-based translation service. The main design features of the ArchTran system are the adoption of a mixed (bottom-up parsing with top-down filtering) parsing strategy, a scored parsing mechanism, and the corpus-based, statistics-oriented paradigm for linguistic knowledge acquisition. Under this framework, research directions are toward designing systematic and automatic methods for acquiring language model parameters, and toward using preference measure with uniform probabilistic score function for ambiguity resolution. In this paper, the underlying probabilistic models of the ArchTran designing philosophy will be presented.
The METAL system which originally evolved from a cooperation between the University of Texas and Siemens became a product in 1988. METAL is implemented on multi-user worksta- tions with a LISP server in the background. It is integrated into the office environment and permits automatic deformatting and reformatting of documents. METAL is characterized by recursive grammars, best paths parsing and a modular lexicon structure. Recent changes in system design have focussed both on internal structure and on user interface. Experiences with productive use have proven METAL’s cost-effectiveness but have also shown the need for increased cooperation between developers and end-users.
This presentation outlines the implementation of a machine translation system for avalanche warning bulletins in natural language, using a unification-based formalism developed at ISSCO, which will be introduced at the same occasion. Concrete examples taken from this project exemplify a modern approach to ma- chine translation: a rich representation of the semantic content of a sentence, the use of a sin- gle grammar for parsing and generating as well as generation and transfer based exclusively on the semantic representation of a sentence. Simultaneously, the limits of bidirectional trans- fer are being tested.
This paper shows that, there are a number of common concepts which are used to define a class of nouns in standard, monolingual English and Spanish dictionaries. An experiment is described to show how a small sot of such con- cepts was derived semi-automatically by automatically analysing the definitions in each language and then matching equivalent definitions manually. Also, some of the benefits of constructing such sets are described, together with the problems encountered while carrying out the experiment.
Knowledge-based interlingual machine translation systems produce semantically accurate translations, but typically require massive knowledge acquisition. This paper describes KANT, a system that reduces this requirement to produce practical, scalable, and accurate KBMT applications. First, the set of requirements is discussed, then the full KANT architecture is illustrated, and finally results from a fully implemented prototype are presented.
We wrote this report in Japanese and translated it by NEC's machine translation system PIVOT/JE.) IBS (International Business Service) is the company which does the documentation service which contains translation business. We introduced a machine translation system into translation business in earnest last year. The introduction of a machine translation system changed the form of our translation work. The translation work was divided into some steps and the person who isn't experienced became able to take it of the work of each of translation steps. As a result, a total translation cost reduced. In this paper, first, we report on the usage of our machine translation system. Next, we report on translation quality and the translation cost with a machine translation system. Lastly, we report on the merit which was gotten by introducing machine translation.
The demand for personal use of a translation system seems to be increasing in accordance with the improvement in MT quality. A recent portable and powerful engineering workstation, such as AS1000 (SPARC LT), enables us to develop a personal-use oriented MT system This paper describes the outline of ASTRANSAC (an English-Japanese/Japanese- English bi-directional MT system) and the extensions related to the personalization of ASTRANSAC, which have been newly made since the MT Summit II.
The Translator’s Workbench provides the user with a set of computer-based tools for speeding up the translation process and facilitate multilingual text processing and technical writing. The tools include dictionaries, spelling, gram- mar, punctuation and style checkers, text pro- cessing utilities, remote access to a fully auto- matic machine translation system and to termi- nological data bases, an on-line termbank, and a translation memory in an integrated framework covering several European languages.
ULTRA (Universal Language Translator) is a multi-lingua] bidirectional translation system between English, Spanish, German, Japanese and Chinese. It employs an interlingua] structure to translate among these five languages. An interlingual representation is used as a deep structure through which any pair of these languages can be translated in either direction. This paper describes some techniques used in the Chinese system to solve problems in word ordering, language equivalency, Chinese verb constituent and prepositional phrase attachment. By means of these techniques translation quality has been significantly improved. Heuristic search, which results in translation efficiency, is also discussed.
This paper describes a memory-based machine translation system developed for the Semantic Net- work Array Processor (SNAP). The goal of our work is to develop a scalable and high-performance memory-based machine translation system which utilizes the high degree of parallelism provided by the SNAP machine. We have implemented an experimental machine translation system DMSNAP as a central part of a real-time speech-to-speech dia- logue translation system. It is a SNAP version of the ΦDMDIALOG speech-to-speech translation system. Memory-based natural language processing and syntactic constraint network model has been incorporated using parallel marker-passing which is directly supported from hardware level. Experimental results demonstrate that the parsing of a sentence is done in the order of milliseconds.
Recently, several types of Japanese to English MT (machine translation) systems have been developed, but prior to using such systems, they have required a pre-editing process of re-writing the original text into Japanese that could be easily translated. For communication of translated information requiring speed in dissemination, application of these systems would necessarily pose problems. To overcome such problems, a Multi-Level Translation Method based on Constructive Process Theory had been proposed. In this paper, the benefits of this method in ALT-J/E will be described. In comparison with the conventional elementary composition method, the Multi-Level Translation Method, emphasizing the importance of the meaning contained in expression structures, has been ascertained to be capable of conducting translation according to meaning and context processing with comparative ease. We are now hopeful of realizing machine translation omitting the process of pre-editing.
All human languages are open and complex communication systems. No machine translation system will ever be able to automatically translate all possible sentences from one language to another in high quality. One way to combat complexity and openness of language translation is to decompose the task into well-defined sequential subtasks and solve each using declarative, modular rules. This paper describes such an MT system. A language-independent MT Machine has been designed for the transformation of linguistic trees in a general fashion. A full MT system is composed of a sequence of instances of that machine. Finnish-English implementation is discussed.
We present JANUS, a speech-to-speech translation system that utilizes diverse processing strategies including connectionist learning, traditional AI knowledge representation approaches, dynamic programming, and stochastic techniques. JANUS translates continuously spoken English utterances into Japanese and German speech utterances. The overall system performance on a corpus of conference registration conversations is 87%. Two versions of JANUS are compared: one using an LR parser (JANUS-LR) and one using a neural-network based parser (JANUS-NN). Performance results are mixed, with JANUS-LR deriving benefit from a tighter language model and JANUS-NN benefiting from greater flexibility.