| |
| |
Front Page |
| |
|
| |
|
But I am pleasantly surprised at reading Lumeras's statement that in some areas of study the importance of contextualizing texts is recognized, and „meaning is not content any more, but is relativized, negotiated and remodeled according to external factors... etc."
Yet the most common repertory of meaning limited to that of a word only is usually a dictionary of some kind. I am not going to evaluate dictionaries here; I will only risk one important remark: the fact that the core of most dictionaries is a headword is a serious drawback to the further development of knowledge in linguistics and translation technology. The association between one word and a string of other words taken as meaning has the unsolved problem of sorting multi-item headwords rather than providing a correct solution to enumerating distinct concepts. Putting that subject aside for the moment, let me come up with my definition of translation with a view to one of Lumeras's references that I am happy to buy where she says that „communication is first of all negotiation... etc." On the other hand, I do not agree with her when she claims that „equivalence is a dangerous notion...", because equivalence of two texts in L1 and L2 is the common goal that any translator or user wants to arrive at, even though the output may be amply annotated to explain in what sense the two texts in L1 and L2 may be considered equivalent.
The word translation means at least two different things (making such a distinction is often called disambiguation):
Disambiguation is needed because dictionary entry words are produced by decontextualization, which is now considered counterproductive to learning and translating: this is why corpus linguistics and concordances are so popular today.
Translation is a complex mental and physical activity of creation and authorship usually performed under a number of constraints, an output operation, the result of which is a translation, the output itself, in practice a text in a human (natural) language deemed to be equivalent to another text used as input for the operation—by a competent translator and a competent client. Translation2 is also a business service provided, similarly to appraisals, another business service with which translation seems to share some core similarities. Appraisers work on the basis of fair market values.3 The analogies must be clear with emphasis here on competent players, i.e. a competent client and a competent translator in terms of knowing the translator trade, and the text, the context, and the relevant chunks of realia (concrete chunks of reality, the whole range of contexts) in both languages and cultures, etc. In operative terms translation is transformation, transfer, recursion or rewriting to meet certain new criteria for a text message to work in another language and environment, including culture. For instance, it is nearly always normal not to keep the original title of a book, film, poem or a model name of a product, etc. when the product is to be introduced elsewhere. Most of these titles are not generated by considering the words in the original title; instead, they follow the rules of name-giving, the last exercise in the process of localization. This is not the only example that shows how unfortunate is the practice of producing paper-based dictionaries with single-word entries, based on the assumption that a word is the smallest meaningful unit of a language. In fact, this is not true, and certainly it is a very deplorable way of translating texts believing that a translator needs to find a single word to fit in some missing sense. And most embarrassingly, the fact that a thing is called what you find in a dictionary is a very poor excuse. But this issue is another broad subject that I am not going to discuss here both for lack of space and for the sake of concentrating on the original topic of meaning. So coming back to the practice of translating, changing, or transcribing some part of the input text is done by following a number of conventions that are or may be written down for a particular producer of a particular class of texts. Examples include News Agencies, EU GD, Translation Agencies, Publishers, Pharmaceutical companies, On-line magazines, etc. If meeting such conventions are a prerequisite to submitting any text for publication, then they are reasonably called standards, which may be industry-wide, meaning the publishing industry rather than the translating industry, which is too diverse to appear as one block of market players of identical interests. Therefore a recent publication of Translation Standards (for Agencies) seems to be far from that type of document. (Note that people outside the translation trade manage to claim that they know what this business is about and publish quite irresponsible texts as in the latter example.)
So instead of following the standards that are compiled in the recent EU Standards, and which are practically the same as those in any other service business, it would be better to remind the trade of what is going on in practice. It is quite likely, for example, that some other rules govern the work of professional translators whether they are aware of them or not. They may be expressed as below: Rule 1: Every translator delivers according to the best of his ability considering the available time and other constraints Rule 2: When a translator is in doubt, he will use authentic sources (dictionaries, previous translations, versions, anything already checked for quality), or Rule 3: The translator will team up with a better (native) speaker and/or specialist (from client, etc.), or Rule 4: The translator will split up work to ensure compliance with the delivery times. Rule 5: He will observe the required consistency (over time, across document or client, market players, etc.) and Rule 6: He will keep context in focus. Rule 7: When it doubt, he will consult the client. As it has already been claimed above, context is seen as the wide world starting with collocations and ending with real-life experience represented by proven high-level abstractions or categories. It is important to highlight that Natural Languages are context-sensitive; this is why you have ambiguities that you want to get rid of, or sometimes on the contrary, if you enjoy or aim at double entendre, a silly trend in contemporary marketing texts and business mumbo-jumbo.
Natural languages are context-sensitive as, opposed to machine or programming languages, which are not. In other words: the vocabulary and grammar rules of context-free languages are unambiguous to the processor which parses, translates, and produces an output—at another level of a machine language. We have seen above that the enrichment of dictionary entries consists of including more and more of the context that is related to any particular entry. Thus, in addition to statistics on word use frequency, definitions, or relations as in a thesaurus, we have ontology for providing more context. But such elements of contextual information are no longer required to be physically next to an entry word, or grammatically linked to it; they may be linked via hypertext, taking the reader to another dimension from the original surface, a process that may be repeated by infinite chaining. Context research as such, has just started and is in an incipient stage. Obviously a word may occur in multiple contexts, and it is not practical to list all of them under an entry word of everyday use. Instead, we have specialized dictionaries showing specialized context and usage. If we do not have them, it is for various reasons such as competition, envy, and conflict of interests. So there is a ray of hope there. In fact, since most of the words are either identifiers or descriptors of some sort, we should not be interested in the verbal phrases per se, but in what they denote, the objects in reality regardless of how we have experienced this reality.
The names of objects, concepts, and other entities that we need to give names to are just titles, headings or labels, or clusters of words. We have other representations of such items that seem to be taken for granted or known, such as visual, audio, or complex representation. Yet they are more difficult to sort into an accessible classification system, while books and printed matter have a long-standing, albeit imperfect, system of classification. They are an excellent resource for secondary utilization; for example, the UDC (Universal Decimal Codes) system is an excellent identifier the documents in a way that reflects our current knowledge of reality. This is usual with all sciences where an inventory of the subject matters is created, either based on a morphological (alphabetical) classification, or on the sorting criteria resulting in nomenclatures of all sorts. All we need to keep in mind is that to identify anything in this living world or the universe we need spatial and temporal identifiers, usually numbers or coordinates, and the same is true about man-made artifacts, whether real and tangible or imagined and intangible. Therefore whether words or numbers, they are just names or pointers to locate where the given object or property, in short, knowledge, is to be recalled from. But that calls for a better clarification of what meaning really is and what really words stand for. And let us remember: we have a pattern to search any information, and this pattern happens to be a single alpha string on the Internet, where in fact all the index records are ultimately identified as numbers. So the lesson is that you either have an alpha order that makes no sense, or a numeric order that may suggest some relationship as in a thesaurus, but none of them are useful enough for the time being for translation purposes. WordNet and similar visual representations are also inadequately designed for that application, because they do not fit the way we think in encountering a problem in translation and trying to find a solution. The solution is not finding a word, but understanding that bit of reality after discovering or exploring the relationship of a particular concept that is still unclear to us.
The translation as a finished product is subject to evaluation, assessment and criticism, all deploying some form of ideal model to compare the actual work delivered. Therefore we need to define quality first, then the standards used in the comparative operation themselves.
Quality4 in general is a judgement resulting from comparing something done to something desired to be done. With respect to translation, we work with two significant factors: speed of delivery, and accuracy of the finished product—defined as the equivalence of the two texts, and meaning the ratio of "translation errors" in comparison with an ideally well-formed (faithfully rendered) product.
The meaning for the term quality has developed over time. Various interpretations5 suggest that it can be a numeric indicator, especially with so many text processing software tools available. Just as the size of a translation or the analysis of the level of difficulty of any text may be well supported by the use of linguistic programs, including concordance programs and other statistical analyses of corpora. It is a composite indicator that needs to have its elements defined.
The elements of quality are derived from various checks performed on the final text by comparing the two texts at the same time. In this respect the text as a standalone product must also satisfy the requirements set for any information product, or product meant to be read, understood and used as information. So here is what is to check for: Is the work Complete, Timely, True, Reliable, Authentic, Relevant, Faithful, Valid, Fit for purpose, Suitable for occasion, Acceptable by client or standards, To the point, Professional, Equivalent in terms of, .... Wholly, partly, hardly... etc. Now, none of these checks is trivial or may be performed by a machine. Just as there are no machines readily available with rules to check a text against the realia:
At this junction I must emphasize that there is no such thing as a perfect synonym—what we have is a list of phrases that can be used in place of each other in a single context subject to certain limitations (or agreement on the exact nature of the context). This may surprise some people, but is a very good start to start thinking differently about meaning and the fallacies of meaning bound to one word (dictionary entry). Here, we have a list of problems, including the problem of using a word for something else than what it identifies, the problem of using a different word for something that is identified by another word, the problem of using an old word for a new concept, and the problem of using a new word for an old concept, etc. Word, context, realia and user may equally be abused. All that will lead to the biggest problem of all: the problem of the dictionary of synonyms and thesauri (high level ontology). High-level, upper, or core ontology is a field used in AI to identify the components of an artificially created world or piece of equipment with the aim of making it work and to be able to understand/describe how it works. All that is about an artificially created reality.
The concept of reality is central to the issue of translating non-fiction, non-essay, scientific or technical prose from L1 (e.g. English) into L2 (e.g. Hungarian). In my mind, reality is always a serious business where it is assumed that a language is used to pin down something vital for man in order to be able to act upon such linguistic description of life in good faith. Ambiguity, falsification, deliberate lies, hoaxes, metaphors and other gadgetry of brainwashing are outside my concern of reality, no matter how widespread they are. In this context one must make the texts acceptable in terms of truth, reason, reality and reliability tests applied. This means, among other things, that you must get the names, the numbers and the ranks/titles and measurements right as your priority concern. In the majority of cases in my experience, we often need to "translate" something from L1 that does not exist in L2. And if we do not translate reality, we end up with words borrowed from L1 likely to eventually ruin the reality in L2. Most people believe that reality is the same as existence or living, of which we all have a subjective experience to share. And as soon as we share such experience in the form of some "knowledge representation" we have an objective product, something that we are not free to interpret subjectively (on our own) only, but in conjunction and in accordance with the understanding of the rest of the world as incompletely reflected in bilingual dictionaries, for example. This is why a professional translator is required not only to speak two languages, but also to be an expert at the subject matter of the text at hand in both languages. However, contrary to current followers of MT research, who start out from huge volumes of TMs and other aligned corpora, our knowledge of reality should not start from bottom to top, but from top to bottom. This must be obvious from the day-to-day practice. When we cannot name or identify an object, we take a shortcut to the next higher concept level and will use a class term which includes the object. Having said that, the current practice of statistical analysis of paired sentences or passages is nothing more then EDI (Electronic Data Interchange) already in use for many years. The technology relies on numerical identifiers and a reality check if it is to be correct. But Machine Translation Tools fail to give us that verification option and will produce garbage, save a few exceptions that are in fact EDI applications. As a conclusion I should say that my and your problem as a translator is that we do not translate a word, but the longest sequence or cluster of words that makes sense when checked against two realities, in L1 and L2. Then we realize whether there is a similar or equivalent construct in the target language describing or identifying the relevant chunk of reality in the country or context of L2. And of course, reality will also cover our mental constructs that need to be systematically described in terms of a high-level ontology language to show how they are related, and those relations will make sure that the descriptors of objects, properties, etc, in L1 will match their counterparts in L2.
Notes: 1 See: Ferenc Kiefer, Hungary: Jelentéselmélet (Theory of Meaning), Corvina Egyetemi Könyvtár, 2. revised ed., 383 pages. Budapest, 2007, ISBN 978 963 13 5682 3. 2 Note: The activity translation is further divided into operations, that collectively or individually make up the broad term translation that has specific variations depending on the ways it is done. The process is similar to thinking, which at this time is not being discussed in terms of minor logic operations despite the fact that it is already called for in a need identified for reflective thinking (by the EU in PISA) for critical thinking and informal logic (in the form of a school subject in the UK) and for a disciplined mind or disciplined thinking (see: J. Dewey). 3 Fair Market Value is a term in both law and accounting to describe an appraisal based on an estimate of what a buyer would pay a seller for any piece of property. It is a common way of evaluating the value of property when assessing damages to be awarded for the loss of, or damage to the property, generally in a claim under tort or a contract of insurance. 4 A classic formulation of the definition of "fair market value" is found (in the specific context of U.S. tax law) in the United States Supreme Court decision in the Cartwright case: "The fair market value is the price at which the property would change hands between a willing buyer and a willing seller, neither being under any compulsion to buy or to sell and both having reasonable knowledge of relevant facts." It is also important that they should be knowledgeable and unencumbered by undue pressure, each acting in his own best interest, the definition of which shall be very enlightening to those trying to identify the essence of translation as a professional activity pursued in a professional way. 5 Many different techniques and concepts have evolved to improve product or service quality, including SPC, Zero Defects, Six Sigma, quality circles, TQM, Theory of Constraints(TOC),Quality Management Systems (ISO 9000 and others), and continuous improvement.)
(However, the American Society for Quality defines "quality" as "a subjective term for which each person has his or her own definition." ) Source: http://www.asq.org/glossary/q.html. | |
|
|
his paper has been encouraged by the publication of Maite Aragonés Lumeras: Meaning: ThePhilosopher's Stone of the Alchemist Translator? (Translation Journal, Volume 12, No. 3 July 2008