| { | |
| "paper_id": "1996", | |
| "header": { | |
| "generated_with": "S2ORC 1.0.0", | |
| "date_generated": "2023-01-19T10:30:53.989865Z" | |
| }, | |
| "title": "Translator's Workbenches: A Practical Application", | |
| "authors": [ | |
| { | |
| "first": "Adriane", | |
| "middle": [], | |
| "last": "Rinsche", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "The Language Technology Centre Ltd", | |
| "location": { | |
| "addrLine": "27 Cotswold Close, GB-Kingston", | |
| "postCode": "KT2 7JN", | |
| "settlement": "Surrey" | |
| } | |
| }, | |
| "email": "" | |
| } | |
| ], | |
| "year": "", | |
| "venue": null, | |
| "identifiers": {}, | |
| "abstract": "The paper presents an example of a solution designed for a multinational client involving translator's workbenches. As a first stage, the user requirements were assessed, as concerns contents and form of documents, the technological base, human resources and volume and cost of translation. The range of possible solutions was then evaluated, in terms of budget, technological base again, language combinations, user friendliness, customisability and linguistic features. Recommendations were then made in terms of either setting up an in-house language business unit or an outsourcing solution. The client opted for an outsourcing solution. We converted, aligned and imported existing information into the workbench solution and resolved organisational issues. We have offered MT as an integral part of Workbench solutions to a variety of clients. We will discuss the financial viability and our own views of integrating MT into a translation technology environment.", | |
| "pdf_parse": { | |
| "paper_id": "1996", | |
| "_pdf_hash": "", | |
| "abstract": [ | |
| { | |
| "text": "The paper presents an example of a solution designed for a multinational client involving translator's workbenches. As a first stage, the user requirements were assessed, as concerns contents and form of documents, the technological base, human resources and volume and cost of translation. The range of possible solutions was then evaluated, in terms of budget, technological base again, language combinations, user friendliness, customisability and linguistic features. Recommendations were then made in terms of either setting up an in-house language business unit or an outsourcing solution. The client opted for an outsourcing solution. We converted, aligned and imported existing information into the workbench solution and resolved organisational issues. We have offered MT as an integral part of Workbench solutions to a variety of clients. We will discuss the financial viability and our own views of integrating MT into a translation technology environment.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Abstract", | |
| "sec_num": null | |
| } | |
| ], | |
| "body_text": [ | |
| { | |
| "text": "Translation technology offers solutions to a variety of multilingual documentation problems which may vary considerably from one prospective user or user organisation to the other.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "The increasing number of tools and systems becoming available on the market place requires a careful evaluation as a basis for any far-reaching decision, because the investment envisaged is not restricted to the cost of the software solution chosen. In order to provide the best result, any off the shelf product or product configuration requires considerable customisation. The time and cost involved in designing and implementing an appropriate solution should not be underestimated. Very well defined and well organised procedures need to be introduced in order to achieve a cost-effective, high quality and high speed solution to the translation problem. Today, many international corporations do not organise the processes involved in the most appropriate way yet.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "An example of a solution we designed for one of our multinational clients will be discussed below. It illustrates the most crucial steps, embracing \u2022 an analysis of the user requirements", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u2022 an evaluation of the relevant technology", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u2022 recommendations and an implementation plan.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "As our evaluation was based on the requirements of a specific client, it cannot be generalised. Although we developed global evaluation criteria for translation technology, which were, for instance, applied as a basis of our recent \"OVUM Evaluates\" report, the choice of a particular solution depends on the requirements of a given company. There is no global \"best solution\" or \"best product\" on the market.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "Our client is an American multinational manufacturer of machine tools. The European Headquarters are based in the UK. The management within one department of this organisation asked us to provide a technology based solution to their translation requirements. We gathered the following information as a basis for further research and our final recommendations:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Towards integrated multilingual document processing", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "Step 1: Evaluating User Requirements (May 1995)", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Towards integrated multilingual document processing", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "The company handles, in principle, four text types within the technical/engineering domain, three of them highly repetitive, the last has a more informal and chatty style.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Text types:", | |
| "sec_num": "1.1." | |
| }, | |
| { | |
| "text": "\u2022 Repair Instructions, repetitive", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Text types:", | |
| "sec_num": "1.1." | |
| }, | |
| { | |
| "text": "\u2022 Consumer Manuals, repetitive", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Text types:", | |
| "sec_num": "1.1." | |
| }, | |
| { | |
| "text": "\u2022 Machinery manuals", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Text types:", | |
| "sec_num": "1.1." | |
| }, | |
| { | |
| "text": "\u2022 Training materials, non-repetitive", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Text types:", | |
| "sec_num": "1.1." | |
| }, | |
| { | |
| "text": "A considerable amount of terminology (parts lists) and parallel text in seven languages is stored in machine readable form on a mainframe computer, all in capital letters . The information stored on the mainframe will not be supported beyond then end of 1995. It is therefore vital to retrieve the information and convert, store and reuse it in an appropriate environment.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Information available in machine readable form:", | |
| "sec_num": "1.2." | |
| }, | |
| { | |
| "text": "English is the source language, at the time of the evaluation 6 target languages need to be covered, to be extended to 12 target languages later.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Languages:", | |
| "sec_num": "1.3." | |
| }, | |
| { | |
| "text": "Formatting and tagging requirements are practically non-existent in the text type \"Repair Instruction\", with the exception of occasional tables and statistics. Machinery and consumer manuals, on the other hand, are produced in Pagemaker.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Formatting and tagging:", | |
| "sec_num": "1.4." | |
| }, | |
| { | |
| "text": "Several departments use Ami Pro as their word processing environment throughout, on networked PCs and workstations.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "The technology base", | |
| "sec_num": "2." | |
| }, | |
| { | |
| "text": "The graphics department uses one Macintosh powerstation and a number of smaller MACs with Pagemaker as their major input and output medium for linguistic data and QuarkXpress as their major medium for pictorial data. Some employees use other packages such as Word for Windows.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "The technology base", | |
| "sec_num": "2." | |
| }, | |
| { | |
| "text": "A group of technical authors, illustrators, project managers is responsible for documentation.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Human resources", | |
| "sec_num": "3." | |
| }, | |
| { | |
| "text": "Translations are carried out:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Human resources", | |
| "sec_num": "3." | |
| }, | |
| { | |
| "text": "\u2022 partially in-house by trainees or technical staff with language skills", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Human resources", | |
| "sec_num": "3." | |
| }, | |
| { | |
| "text": "\u2022 partially by various translation agencies", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Human resources", | |
| "sec_num": "3." | |
| }, | |
| { | |
| "text": "\u2022 partially by low level bilingual staff in the various national markets. This is a fragmented and unorganised way of organising multilingual documentation, with a high degree of inconsistency, at low speed, resulting in low quality output.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Human resources", | |
| "sec_num": "3." | |
| }, | |
| { | |
| "text": "The volume of text material translated in the past in one department was comparatively low. The source text material consisted of approximately 20 000 words a year and was translated into 6 target languages, with an overall annual cost of \u00a314,000. If additional requirements are met without technical support the cost of translation will rise to \u00a3250,000 for traditional external translation services in one department alone (calculated on the basis of \u00a3120 per 1000 words translation).", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Volume and cost of translation", | |
| "sec_num": "4." | |
| }, | |
| { | |
| "text": "Other departments within the organisation have similar translation requirements in the same subject domains.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Volume and cost of translation", | |
| "sec_num": "4." | |
| }, | |
| { | |
| "text": "A limited budget for introducing translation technology is available to the department with very urgent translation requirements, and although we would have preferred a top to bottom approach where we would have aimed to introduce an all-embracing solution, we decided to adopt an initial \"pilot\" approach in order to help our client with their immediate needs and in order to support their unusual interest and motivation to dedicate themselves to a technologically oriented approach to multilingual documentation. The department is obviously aware of the fact that a centralised and unified approach to multilingual document production and management is not only more cost-effective and consistent, but that it provides them with critical influence on the company's international corporate image.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Volume and cost of translation", | |
| "sec_num": "4." | |
| }, | |
| { | |
| "text": "Step 2: Evaluating the range of possible solutions (June 1995)", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Volume and cost of translation", | |
| "sec_num": "4." | |
| }, | |
| { | |
| "text": "Based on the information obtained from the client, we used the following strategy for evaluating the translation technology products available on the market:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Volume and cost of translation", | |
| "sec_num": "4." | |
| }, | |
| { | |
| "text": "In calculating the cost of supplying and implementing a computer-assisted translation solution, the following costs had to be taken into account: a)", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Budget", | |
| "sec_num": "1." | |
| }, | |
| { | |
| "text": "Purchasing price for hardware and software b) Customisation cost (e.g. converting data)", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Budget", | |
| "sec_num": "1." | |
| }, | |
| { | |
| "text": "We considered the following systems: Due to the high estimated degree of repetitiveness of 60% applicable to two out of three text types covering approximately two thirds of the documents involved, it was therefore appropriate to consider a translation memory based solution. All translation workbenches available on the marketplace offered the required language combinations, except the Eurolang Optimizer, which does not cover Greek. We retained the other four systems for further evaluation.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Budget", | |
| "sec_num": "1." | |
| }, | |
| { | |
| "text": "MT", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Budget", | |
| "sec_num": "1." | |
| }, | |
| { | |
| "text": "At the time of the evaluation, it was impossible to resolve all issues regarding layout and formatting, which ideally should be retained throughout the multilingual production process. Pagemaker and QuarkXpress were not supported by any of the translation products. The client requires further consultancy at a later stage in order to adopt an integrated and considerably more cost-effective strategy at a later stage.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Formatting requirements", | |
| "sec_num": "3." | |
| }, | |
| { | |
| "text": "We distinguished between two categories of Workbench solutions: We considered the Type A products more suitable in this context, the Type B products were therefore ruled out.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Overall design", | |
| "sec_num": "4." | |
| }, | |
| { | |
| "text": "Our client employs no linguistic staff in-house and does not intend to do so in the future. For the application planned it was therefore not feasible for the corporation to employ translators. It was our intention, however, to make the documentation process as transparent as possible to the client. We wanted them to be able to control the process from their premises to a certain extent by keeping one copy of the software on site. In order to do so, the system introduced had to be particularly user-friendly, with little training required, an adequate help system and a practical user-interface. Some features of the two systems which \"survived\" the evaluation to this final stage were:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "User-friendliness", | |
| "sec_num": "5." | |
| }, | |
| { | |
| "text": "IBM Translation Manager: System specific editor (at the time of the evaluation) ", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "User-friendliness", | |
| "sec_num": "5." | |
| }, | |
| { | |
| "text": "Our decision in this specific context was in favour of the Trados Translation Workbench. We expected a cost reduction for translation of approximately 60% for the repetitive text material, with considerably increased quality, consistency and speed. The client requires further consultancy in terms of designing source document production with a view to multilingual documentation.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusion", | |
| "sec_num": "8." | |
| }, | |
| { | |
| "text": "Step 3: Recommendations and implementation plan (August 1995)", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusion", | |
| "sec_num": "8." | |
| }, | |
| { | |
| "text": "1. Because of the highly repetitive structure of three out of four text types we recommended the above translation memory based approach with a terminology database both for terminology coding and update processing.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusion", | |
| "sec_num": "8." | |
| }, | |
| { | |
| "text": "2. One multilingual terminology database and one copy of all bilingual translation memory databases required are held on site by our client. The reasons for this are several:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusion", | |
| "sec_num": "8." | |
| }, | |
| { | |
| "text": "a) Data security is increased and risk management improved by having the data on more than one site. Although LTC holds copies of the databases on two powerful PCs and on tape, fire or burglary could result in the loss of the databases. b) Any new text is submitted to the Workbench at the client's site to estimate the cost of a given translation cycle beforehand.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusion", | |
| "sec_num": "8." | |
| }, | |
| { | |
| "text": "c) The percentage of identical matches is estimated to be 60%. The client is, in principle, able to increase this percentage by modifying very similar matches, using the database contents as their standard, and eliminating stylistic, syntactic and terminological variants. By this means an even higher percentage of the new source text becomes identical with source text already stored. This leads to a highly effective and standardised use of both language and storage capacities. In practise, with one year operational experience (August 1996) the client has relied on us for any source text standardisation.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusion", | |
| "sec_num": "8." | |
| }, | |
| { | |
| "text": "3. The structure of the multilingual database and the translation memories is designed in such a way that a more comprehensive and integrated approach to language engineering within the overall corporate business processes can be implemented step by step.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusion", | |
| "sec_num": "8." | |
| }, | |
| { | |
| "text": "4. LTC converts the contents of the existing mainframe database into Workbench format which provides instant reusable text and terminology in 7 languages and adds the relevant information for the additional 5 target languages in a second phase.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusion", | |
| "sec_num": "8." | |
| }, | |
| { | |
| "text": "5. LTC manages the translation process on behalf of the client, as in-house linguistic staff is not available. LTC trains specialist translators in the use of the workbench and sends one fully updated copy of the databases to the client after each translation cycle. The client pays only for the translation of new text.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusion", | |
| "sec_num": "8." | |
| }, | |
| { | |
| "text": "6. During a later stage in 1997 source document control will be increased by providing an English to English translation memory database. Standard text is stored in the database, new segments are accepted only if they are not variants of existing translation units.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusion", | |
| "sec_num": "8." | |
| }, | |
| { | |
| "text": "After one year... the solution is fully operational.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusion", | |
| "sec_num": "8." | |
| }, | |
| { | |
| "text": "Statistics show that our estimate of an average reduction of 60% of the volume of translation and cost is correct. The example below was taken from our August \u00b496 repair instruction update, German. There are 49% of identical text (cf. 100% below).", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusion", | |
| "sec_num": "8." | |
| }, | |
| { | |
| "text": "Repetitions are repeated untranslated elements, 50% of repeated words are charged to the client. We have an open strategy in terms of read and write access to the translation memory database. All translation work is, at present, carried out on site, therefore each translator can take advantage of maximum information, resulting in high motivation.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusion", | |
| "sec_num": "8." | |
| }, | |
| { | |
| "text": "Each translator is responsible for the status of his/her database, which increases motivation further. We are in the fortunate position to be able to agree any deadlines several weeks beforehand with the client to ensure that the same specialised and trained translators are available for each translation cycle.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusion", | |
| "sec_num": "8." | |
| }, | |
| { | |
| "text": "We believe that machine translation plays and overall important role in information scanning and multilingual information retrieval, and will be increasingly relevant in value added and customised internet and intranet applications. In the past, considerable success was achieved with Machine translation and controlled language in technical documentation (cf. Caterpillar and Rank Xerox, to name only two examples).", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "The possible role of MT", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "With the emergence of translation workbenches, machine translation loses relevance for repetitive documentation in limited domains where a considerable amount of 100% matches can be expected through translation memory. However, machine translation could support the initial process of populating a translation memory database where parallel text is not available. Once a workbench is operational, it seems more practical to give translators terminology and translation memory information rather than having them post-edit machine output.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "The possible role of MT", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "If machine translation is integrated into a workbench solution, design, implementation and production procedures should be as follows:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "The possible role of MT", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "The first step is obviously to create or update the terminology database. Then all terms from the term bank are copied and imported into machine translation system dictionaries, at the same time adding the required linguistic coding. For the above client, this would have been possible, but expensive. The cost would have been GBP 2 per term = GBP 4000 per 2000 terms = GBP 48 000 for 12 languages, and the benefit would have been limited and restricted to only some target languages. The client is therefore at present not interested in machine translation integrated into the current solution.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "The possible role of MT", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "After updating the system dictionaries, the relevant source text can be submitted to a batch translation routine, with all identical segments automatically replaced by translation memory database content and all non-100% matches are machine translated.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "The possible role of MT", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "In an interactive revision process the translator checks and post-edits the flagged machine translated text material and enters the correct sentences into translation memory. We would recommend this strategy to companies with a comparatively low repetition rate in their documentation.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "The possible role of MT", | |
| "sec_num": null | |
| } | |
| ], | |
| "back_matter": [], | |
| "bib_entries": {}, | |
| "ref_entries": { | |
| "TABREF0": { | |
| "html": null, | |
| "type_str": "table", | |
| "num": null, | |
| "text": "The very advanced IBM terminology system Translexis is multilingual, but the translation system LMT with which it is integrated offers only two commercially available language pairs at present: English to German and German to English. The solution was not further evaluated for this reason.", | |
| "content": "<table><tr><td>systems considered: METAL, LOGOS, DP/TRANSLATOR, SYSTRAN</td></tr><tr><td>all too expensive, therefore ruled out</td></tr><tr><td>Power Translator Professional</td></tr><tr><td>cheap, not ruled out for financial reasons</td></tr><tr><td>LMT</td></tr><tr><td>not ruled out for financial reasons</td></tr><tr><td>Translation Workbenches considered: IBM Translation Manager</td></tr><tr><td>Eurolang Optimizer</td></tr><tr><td>Trados Translation Workbench</td></tr><tr><td>Star Transit</td></tr><tr><td>XL8</td></tr><tr><td>investment in principle acceptable</td></tr><tr><td>2. Language combinations</td></tr></table>" | |
| } | |
| } | |
| } | |
| } |