METIS-II: Low-Resource MT for German to English

In this article the principles of the METIS Machine Translation system are presented. METIS employs an extensive tagged and lemmatised corpus of texts in the target language, coupled with bilingual lexica covering the desired pairs of source-target languages. To generate a high-quality translation, the METIS system is provided with statistical tools enabling it to extract linguistic knowledge from the annotated corpus of the target language. The advantage of this approach is that, although no grammars need to be provided explicitly, grammatical translations are retrieved from the corpus by using pattern matching techniques.

2016

1999

Is adaptation of English NLP applications the right way to go multilingual? Should one prefer "language-independent" systems with a view to applying them to a large number of different languages? Experience from the processing of Portuguese in several different areas (part-of-speech tagging, corpus tools, lexical decomposition, machine translation, etc.) suggests that neither of these offers a satisfactory solution. This paper argues for a thorough study of the way individual languages work in order to develop applications suited for the language in question, i.e., "language-dependent" systems.

This paper describes a small, structured English corpus that is designed for translation into Less Commonly Taught Languages (LCTLs), and a set of re-usable tools for creation of similar cor- pora. 1 The corpus systematically explores meanings that are known to affect morphology or syntax in the world's languages. Each sentence is associated with a feature structure showing the elements of meaning that are represented in the sentence. The corpus is highly structured so that it can support machine learning with only a small amount of data. As part of the REFLEX program, the corpus will be translated into multiple LCTLs, resulting in parallel corpora can be used for training of MT and other language technologies. Only the untranslated English corpus is described in this paper.

Proceedings of XVIII FIT …, 2008

2006 saw the start of a project for compiling a multifunctional parallel corpus with Dutch as a pivotal language: the Dutch Parallel Corpus (DPC). Among other things, parallel corpora can be a useful tool in translation business. They can help to improve idiomatic language usage, provide translation suggestions or serve for filling up translation memory with high-quality data. The advantages of parallel corpora over multilingual comparable corpora or dictionaries/ glossaries is search speed and reliability. They contain a great amount of aligned data, examples from which can be viewed in the surrounding context. Besides, parallel corpora offer their user the benefit of metadata with additional information allowing for a finer-tuned search of the corpus. The corpus design and text typology are crucial for the usability of the corpus. Insights from cognitive linguistics on basic-level categories have proven to be useful for elaborating such a design and typology assuring (i) text type diversity containing translation samples from different areas of expertise; (ii) high translation quality providing reliable translation solutions; (iii) a well-structured taxonomy for prompt data retrieval.

SIGART newsletter, 1985

Machine Translation, 1991

In this article, we outline the contents of the linguistic specifications of the Eurotra machine translation system. We start in sections 1 and 2 from some of the requirements placed by multilingual MT on the overall design of the linguistic components. We then move on to a characterization of the Eurotra interface structure (section 3), the nature of transfer (section 4), and trends towards more interlingual representations within the project (section 5). Thereafter, we concentrate on the contents of the various levels beside the interface structure (section 6) before giving a brief survey of word structure (section 7) and outlining some areas for further research (section 8).' KEYWORDS: Eurotra, linguistic specifications, translation theory 'As it happens, crane1 and crane2 are distinguished m some EC languages (e.g. Spanish grulfa vs. grua and Dutch kraanvogel vs. kraan).

Journal of Computing and Information Technology, 2004

With more and more text being available in electronic form, it is becoming relatively easy to obtain digital texts together with their translations. The paper presents the processing steps necessary to compile such texts into parallel corpora, an extremely useful language resource. Parallel corpora can be used as a translation aid for second-language learners, for translators and lexicographers, or as a data-source for various language technology tools. We present our work in this direction, which is characterised by the use of open standards for text annotation, the use of publicly available third-party tools and wide availability of the produced resources. Explained is the corpus annotation chain involving normalisation, tokenisation, segmentation, alignment, word-class syntactic tagging, and lemmatisation. Two exploitation results over our annotated corpora are also presented, namely a Web concordancer and the extraction of bi-lingual lexica.

2003

In this paper, we combine techniques from rule-based and corpus-based MT in a hybrid approach. We only use a dictionary, basic analytical resources and a monolingual targetlanguage corpus in order to enable the construction of an MT system for lesser-resourced languages. Statistical and example-based systems usually do not involve a lot of linguistic notions. Cutting up sentences in linguistically sound subunits improves the quality of the translation. Demarcating clauses, verb groups, noun phrases, and prepositional phrases restricts the number of possible translations and hence also the search space. The sentence chunks are translated using a dictionary and a limited set of mapping rules. By bottom-up matching the different translated items and higher-level structure with the database information, one or more plausible translated sentences are constructed. A search engine ranks them using the frequencies of occurence and the matching accuracy in the target-language corpus.