| { | |
| "paper_id": "1998", | |
| "header": { | |
| "generated_with": "S2ORC 1.0.0", | |
| "date_generated": "2023-01-19T10:27:13.727072Z" | |
| }, | |
| "title": "", | |
| "authors": [], | |
| "year": "", | |
| "venue": null, | |
| "identifiers": {}, | |
| "abstract": "", | |
| "pdf_parse": { | |
| "paper_id": "1998", | |
| "_pdf_hash": "", | |
| "abstract": [], | |
| "body_text": [ | |
| { | |
| "text": "The following describes the different translation technologies which are currently available in the Translation Service (SdT) of the European Commission. In order to exemplify their usage, they are described in a (maximalistic) workflow.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "INTRODUCTION", | |
| "sec_num": "1." | |
| }, | |
| { | |
| "text": "For almost any of the technologies described below, it is essential to have the source document available in electronic form. In order to give requesters an incentive to use electronic mail, the SdT has developed a simple and user-friendly interface, known as POETRY (Processing of Electronic Translation Requests), which allows users to send a translation request together with the document to be translated and, if possible, reference material. This request is then passed on to individual translators via WINSUIVI, a management tool which makes it possible for allowing the work to be allocated according to the target languages and products required.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "INTRODUCTION", | |
| "sec_num": "1." | |
| }, | |
| { | |
| "text": "If one abstracts away from organisational aspects (which parts are done by secretaries, which parts are done by translators etc), a difference has to be made between the preparation of a translation, and the translation itself. In the worst case, a translation memory has to be created from zero.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "INTRODUCTION", | |
| "sec_num": "1." | |
| }, | |
| { | |
| "text": "In the SdT, translation memories are increasingly often used in order to improve productivity as well as quality and coherence. If users want to use TM technology, possibly combined with machine translation or replacement tools, and if translation memories do not yet contain enough suitable data, they have to look for reference material which can be imported.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "BUILDING-UP A TRANSLATION MEMORY (TM)", | |
| "sec_num": "2." | |
| }, | |
| { | |
| "text": "The Commission's translators have a number of on-line full-text databases, which can be searched for reference material. SDTVISTA contains almost all the translations from 1994 onwards (except for confidential documents) plus many source texts and a number of reference documents. This allows users to check whether or not a document, or part of it, has already been translated and to retrieve pertinent source documents as well as their translations. Queries are typically made on the basis of search strings, but they can be refined by means of additional filters (e.g. requesting service, year, translation type etc.):", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Reference documents", | |
| "sec_num": "2.1." | |
| }, | |
| { | |
| "text": "If interesting material can be found, it can be viewed or downloaded for further treatment. During the translation process, translators might also consult SDTVISTA in order to solve terminological problems:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Reference documents", | |
| "sec_num": "2.1." | |
| }, | |
| { | |
| "text": "CELEX is the full text database of the Office for Official Publications of the European Communities. It offers multilingual coverage of a wide range of legal documents. CELEX offers a user-friendly Internet access with a large number of different query types.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Reference documents", | |
| "sec_num": "2.1." | |
| }, | |
| { | |
| "text": "The SdT provides an automatic batch retrieval where on the basis of references found in a document, CELEX identifiers are calculated and the corresponding documents are returned to the user. For example, the German sentence 1 ", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Reference documents", | |
| "sec_num": "2.1." | |
| }, | |
| { | |
| "text": "Interesting reference documents and their translations can be downloaded to the user's PC, and a sentence alignment request can be launched:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Alignment", | |
| "sec_num": "2.2." | |
| }, | |
| { | |
| "text": "The SdT uses its own aligner which is highly customised and therefore produces markedly better results than commercially available applications.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Alignment", | |
| "sec_num": "2.2." | |
| }, | |
| { | |
| "text": "For CELEX documents, a pre-processing is carried out (currently still on the client side so that users can still modify texts before they are aligned). This pre-processing makes it possible to iron out the language-specific differences which affect alignment quality. Example: in a number of languages (English, French, Spanish etc.), several \"whereas\"-clauses are frequently put in one sentence (separated by a semi-colon); for German, Danish, Swedish and Greek, new sentences are created in these cases; if many such differences occur in a short piece of text, this can lead to mis-alignment. The solution is to insert a paragraph marker if a semicolon is followed by \"whereas\" (or its equivalent in another language).", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Alignment", | |
| "sec_num": "2.2." | |
| }, | |
| { | |
| "text": "Alignment results can be corrected by a special editor and later imported into a translation memory:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Alignment", | |
| "sec_num": "2.2." | |
| }, | |
| { | |
| "text": "If users want to work exclusively with their own local translation memory, they can start translating at this point. They can however request additional information coming from other sources via the EURAMIS client interface which gives access to a number of batch services.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "PREPARATION OF WORK", | |
| "sec_num": "3." | |
| }, | |
| { | |
| "text": "The SdT has developed server translation memories whose design facilitates data sharing and the combination with other products. Central translation memories are provided for each of the seven thematic groups of the SdT. In addition, there are a number of consolidated topical translation memories (e.g. most important European legislation and Court rulings, the monthly Bulletin etc.). For each of these translation memories, a coordinator grants write access, sees to a coherent use of attributes etc. In addition to common TMs, there are personal TMs for every translator: there are no restrictions concerning read access, everybody can retrieve from everybody's TM. The following shows a typical TM retrieval request:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "TM", | |
| "sec_num": "3.1." | |
| }, | |
| { | |
| "text": "EURAMIS TM queries can be fine-tuned by means of positive and negative filters (\"only these\" or \"all but these\"). The following filters are possible: requesting service (e.g. Directorate-General XXII), domain (e.g. agriculture), document type (e.g. letter) and individual translator (unique login).", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "TM", | |
| "sec_num": "3.1." | |
| }, | |
| { | |
| "text": "MT is currently used in the Commission for 17 language pairs with a quality which varies considerably between the different languages: French-Spanish offers the best quality, followed by French-English, French-Italian and English-French; language pairs with German produce less acceptable results.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Machine translation (MT)", | |
| "sec_num": "3.2." | |
| }, | |
| { | |
| "text": "MT is available not only to translators, but to all Commission officials who are equipped with either a terminal or a PC connected to the internal network. The number of pages translated by the system has increased considerably in recent years: 170 000 in 1995, 231 000 in 1996, 260 000 estimated for 1997. MT is mainly used for", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Machine translation (MT)", | |
| "sec_num": "3.2." | |
| }, | |
| { | |
| "text": "\u2022 the fast translation of short, repetitive texts with a standardised structure and terminology (mail, minutes of meetings, parliamentary questions, reports etc.); \u2022 the browsing of texts written in a language the user does not know;", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Machine translation (MT)", | |
| "sec_num": "3.2." | |
| }, | |
| { | |
| "text": "\u2022 drafting purposes: users write a text in their mother tongue and request a machine translation in order to have a document drafted in something other than their native or main language.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Machine translation (MT)", | |
| "sec_num": "3.2." | |
| }, | |
| { | |
| "text": "MT is appreciated by translators because of its speed, its capacity to keep the original format, and as a terminology aid.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Machine translation (MT)", | |
| "sec_num": "3.2." | |
| }, | |
| { | |
| "text": "TMan replaces pre-defined strings (from words up to paragraphs) in the source document so that the resulting document is a mix of source and target language items. TMan replacements are based on a repetition analysis of the document type in question. This means that the use of TMan is only worthwhile if a document type is quite frequent and if it contains a large number of repetitive elements. This approach is taken for a number of master documents and regular publications (e.g. the Bulletin) where many expressions can be found in every issue, and consequently the analysis of previous texts yields large expression databases to be used for subsequent issues of the same publications. For example, the following French text (from a prototype contract) ", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "TMan", | |
| "sec_num": "3.3." | |
| }, | |
| { | |
| "text": "It is possible to combine TM results with MT (the same integration with TMan is foreseen): on the basis of the user requirements, a EURAMIS TM retrieval is carried out, and the remaining gaps are filled with MT results.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Combination of products", | |
| "sec_num": "3.4." | |
| }, | |
| { | |
| "text": "EURAMIS provides automatic terminology extraction of pertinent EURODICAUTOM entries, together with indirect access to TIS and EUTERPE, the terminology databases of the European Council of Ministers and the European Parliament. Since the Commission's MT system is used as a supporting application, this service is restricted to documents in English, French, German and Spanish. The user can choose between the following output formats: EURODICAUTOM import format (for data correction), a sequential RTF file, and MULTITERM input format.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Document-related terminology", | |
| "sec_num": "3.5." | |
| }, | |
| { | |
| "text": "Queries can be restricted in a number of ways (e.g. by indicating domain). Output can be restricted by selecting specific EURODICAUTOM fields:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Document-related terminology", | |
| "sec_num": "3.5." | |
| }, | |
| { | |
| "text": "As a first step towards migrating EURODICAUTOM to a relational database management system, a new database structure has been designed recently: among other things, formal restrictions are imposed on values and synonyms are put in separate fields (the latter sometimes leads to some overgeneration, which can be seen in the notes of the German entries of the example below). A conversion to this new data model (including some clean-up and normalisation) is already used for output in MULTITERM:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Document-related terminology", | |
| "sec_num": "3.5." | |
| }, | |
| { | |
| "text": "In principle, users can translate in two different working environments; depending on their own preferences and on the type of the document to be translated, they can choose between TRADOS' Translator's Workbench and a simpler treatment completely inside word processing.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "TRANSLATION PROPER", | |
| "sec_num": "4." | |
| }, | |
| { | |
| "text": "With TWB, the products mentioned above can be used in parallel: retrieval from central TM is imported together with MT output; the latter receives a special attribute in order to warn users. Terminology retrieval can be imported directly to MULTITERM.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Translator's Workbench", | |
| "sec_num": "4.1." | |
| }, | |
| { | |
| "text": "The main advantages of Translator's Workbench (as compared to a word processing approach) are the following:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Translator's Workbench", | |
| "sec_num": "4.1." | |
| }, | |
| { | |
| "text": "\u2022 it is interactive, i.e. document-internal repetition can be exploited immediately;", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Translator's Workbench", | |
| "sec_num": "4.1." | |
| }, | |
| { | |
| "text": "\u2022 it is fully integrated with MULTITERM so that terminology can be consulted on the fly; \u2022 it provides a number of additional features such as concordance and coverage analysis.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Translator's Workbench", | |
| "sec_num": "4.1." | |
| }, | |
| { | |
| "text": "Its main disadvantage is that several applications have to share the limited space on the screen and that many people find it too complicated to use:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Translator's Workbench", | |
| "sec_num": "4.1." | |
| }, | |
| { | |
| "text": "If the result is requested in native Word format, the formatting of the source text is preserved to a very large extent 1 . Since colours are rare in Commission texts, they are used to convey information on result type and have to be reset after editing, e.g. blue for TM full matches, red for TM fuzzy matches, and magenta for MT.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Word processing only", | |
| "sec_num": "4.2." | |
| }, | |
| { | |
| "text": "Although the applications at hand make it possible to produce more coherent translations in less time, there are still two main disadvantages:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "FUTURE DEVELOPMENTS", | |
| "sec_num": "5." | |
| }, | |
| { | |
| "text": "\u2022 too much interaction is still needed in order to prepare work (e.g. downloading reference documents, aligning them, importing alignments etc.); \u2022 a certain degree of experience is necessary in order to determine which application(s) should be used under which circumstances. Contrary to MT (which \"knows\" what is translated by what), TM cannot establish a 1:1 correspondence between words. It is therefore not possible for TM to reproduce correct character formatting below sentence level (e.g. bold, italic etc.). The following solution has been implemented instead:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "FUTURE DEVELOPMENTS", | |
| "sec_num": "5." | |
| }, | |
| { | |
| "text": "\u2022 if some character formatting is switched on AND off inside the same sentence, this formatting information is not taken over in the translation \u2022 if some character formatting is only switched on OR off inside the same sentence, the switch is moved to the front of the sentence, i.e. the formatting information is taken up from the beginning of the sentence.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "FUTURE DEVELOPMENTS", | |
| "sec_num": "5." | |
| }, | |
| { | |
| "text": "The first problem can partly be solved by combining existing modules. One example for this is the automatic retrieval of CELEX titles and documents, where a longer chain could be implemented:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "FUTURE DEVELOPMENTS", | |
| "sec_num": "5." | |
| }, | |
| { | |
| "text": "\u2022 calculation of CELEX identifiers from references in source document;", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "FUTURE DEVELOPMENTS", | |
| "sec_num": "5." | |
| }, | |
| { | |
| "text": "\u2022 file transfer of identified documents from CELEX server;", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "FUTURE DEVELOPMENTS", | |
| "sec_num": "5." | |
| }, | |
| { | |
| "text": "\u2022 clean-up and normalisation of CELEX documents at server level;", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "FUTURE DEVELOPMENTS", | |
| "sec_num": "5." | |
| }, | |
| { | |
| "text": "\u2022 alignment (very reliable due to normalisation of CELEX documents);", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "FUTURE DEVELOPMENTS", | |
| "sec_num": "5." | |
| }, | |
| { | |
| "text": "\u2022 creation of an ad-hoc TM;", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "FUTURE DEVELOPMENTS", | |
| "sec_num": "5." | |
| }, | |
| { | |
| "text": "\u2022 combined search with source document in ad-hoc and other TMs.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "FUTURE DEVELOPMENTS", | |
| "sec_num": "5." | |
| }, | |
| { | |
| "text": "The second problem can partly be solved by providing an expert system which suggests the most suitable treatment for a given text. As far as TM treatment is concerned, a recommendation could be based on the following analyses:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "FUTURE DEVELOPMENTS", | |
| "sec_num": "5." | |
| }, | |
| { | |
| "text": "\u2022 calculate the degree of internal repetition of the source document (including fuzzy sentences): high value favours treatment with Translator's Workbench;", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "FUTURE DEVELOPMENTS", | |
| "sec_num": "5." | |
| }, | |
| { | |
| "text": "\u2022 create CELEX ad-hoc TM if possible, find most pertinent existing server TMs, calculate overall coverage of source document: high value favours TM treatment in general.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "FUTURE DEVELOPMENTS", | |
| "sec_num": "5." | |
| }, | |
| { | |
| "text": "The next step would then be to integrate such an expert system in the production management application WINSUIVI, so that the queries necessary for the preferred treatment can already be launched and the results can be saved into a working directory before the translation request even arrives on the translator's desk.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "FUTURE DEVELOPMENTS", | |
| "sec_num": "5." | |
| } | |
| ], | |
| "back_matter": [], | |
| "bib_entries": { | |
| "BIBREF0": { | |
| "ref_id": "b0", | |
| "title": "\u00dcbersetzungswissenschaft im Umbruch. Festschrift f\u00fcr Wolfram Wilss zum 70. Geburtstag", | |
| "authors": [ | |
| { | |
| "first": "A", | |
| "middle": [], | |
| "last": "Blatt", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "A", | |
| "middle": [], | |
| "last": "Lauer", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "H", | |
| "middle": [], | |
| "last": "Gerzymisch-Arbogast", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Haller", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "E", | |
| "middle": [], | |
| "last": "Steiner", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1996, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "131--134", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Blatt, A. (1996): The EURAMIS Project. In: Lauer, A., Gerzymisch-Arbogast, H., Haller, J., Steiner, E.: \u00dcbersetzungswissenschaft im Umbruch. Festschrift f\u00fcr Wolfram Wilss zum 70. Geburtstag. T\u00fcbingen 1996, pp. 131-134.", | |
| "links": null | |
| }, | |
| "BIBREF1": { | |
| "ref_id": "b1", | |
| "title": "EURAMIS, The European Advanced Multilingual Information System", | |
| "authors": [ | |
| { | |
| "first": "A", | |
| "middle": [], | |
| "last": "Blatt", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "P", | |
| "middle": [], | |
| "last": "Martins", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1997, | |
| "venue": "The ELRA Newsletter", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "3--5", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Blatt, A., Martins, P. (1997): EURAMIS, The European Advanced Multilingual Information System. The ELRA Newsletter 1997-2, pp. 3-5.", | |
| "links": null | |
| }, | |
| "BIBREF2": { | |
| "ref_id": "b2", | |
| "title": "Integrierte \u00dcbersetzungssysteme. Betrachtungen zu \u00dcbersetzungsproze\u00df", | |
| "authors": [ | |
| { | |
| "first": "U", | |
| "middle": [], | |
| "last": "Reinke", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1997, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "97--106", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Reinke, U. (1997): Integrierte \u00dcbersetzungssysteme. Betrachtungen zu \u00dcbersetzungsproze\u00df, \u00dcbersetzungsproduktivit\u00e4t, \u00dcbersetzungsqualit\u00e4t und Arbeitssituation. Lebende Sprachen 1997-3, pp. 97-106.", | |
| "links": null | |
| } | |
| }, | |
| "ref_entries": { | |
| "FIGREF0": { | |
| "type_str": "figure", | |
| "num": null, | |
| "uris": null, | |
| "text": "4a) La Communaut\u00e9 europ\u00e9enne, repr\u00e9sent\u00e9e par la Commission des Communaut\u00e9s europ\u00e9ennes, ici repr\u00e9sent\u00e9e par Monsieur ..., Directeur g\u00e9n\u00e9ral,d'une part, et la firme \"...\", dont le si\u00e8ge social est \u00e0 ... ci-apr\u00e8s d\u00e9nomm\u00e9e \"le fournisseur\", repr\u00e9sent\u00e9e par Monsieur ... en vertu de la d\u00e9l\u00e9gation lui conf\u00e9r\u00e9e par ladite soci\u00e9t\u00e9 would be translated as (4b) where colours (put in bold here) show what remains to be translated: (4b) La Communaut\u00e9 europ\u00e9enne, repr\u00e9sent\u00e9e par la Commission de las ComunidadesEuropeas, ici repr\u00e9sent\u00e9e par Monsieur ..., Director General, y, de otra, y la firme \"...\", con domicilio social en ... denominada en lo sucesivo \"el proveedor\", representada por el Sr ... seg\u00fan el poder otorgado por la citada sociedad," | |
| }, | |
| "FIGREF1": { | |
| "type_str": "figure", | |
| "num": null, | |
| "uris": null, | |
| "text": "1" | |
| } | |
| } | |
| } | |
| } |