This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.
One of my goals in this assignment was to change the weather forecast texts in order to make them more suitable for use with one of most popular translation pro-grammes for computers - SDL Trados Studio 2011. This software variant is highly effective but SL texts need to be re-organized for better results. A translation act is difficult for us humans and even more so for computer programmes. Different kinds of matches in TM are possible while working with SDL Trados: fuzzy matches (when similar segments are found), full or 100%-matches and context matches (Dennett 1995: 29). The first step would be to read the given text, in my case the weather forecast file with the first five reports.
Prior to this work I have red about theories of machine translation in regards to my topic (e.g. Arnold 2001, Dennett 1995), which will be discussed later. Using these theories I could realize the best way to analyse all the texts. First of all, my strategy was not to proceed step by step, changing the text bit by bit to discover insignificant changes. I tried to change as much as possible. I was certain that this method would affect my texts to a high degree. Of course, it was important to find synonyms or any sentences with similarities. Obviously, synonyms like "dawn" and "morning", "sho-wers" and "rain" decrease the degree of repetition. My intention here was to shrink the use of words with the same or similar meaning. I decided to constantly use one of the variants to accomplish this. The numbers for highest and lowest temperatures or tem-perature changes were very repetitive. Various sentences, and sentences with numbers in particular, could well be re-structured without distorting their meaning. I noticed that the numbers were linked with each other in three ways: with conjunctions "and" and "to" and with a dash "-" ('telegraphic style of writing' - Arnold 2001: 152). I knew that a higher degree of conformity could be achieved here - by leaving only one of them.
Further, in sentences with numerical information, the numbers were preceded by few indicating words, for example adjectives like "highest", "lowest", "top", "mi-nimum" and verbs like "range" and "reach". Likewise, similar information about weather conditions tended to be linked to directions and day time (e.g. "from the north", "in the south and southeast", "through the afternoon", "by dawn"). Here the repetition rate could also be increased by reducing synonyms. Other peculiarities, which caught my eye, were adjectives describing the force and the direction of the wind ("moderate southwest winds...") and the fact that information referring to the winds usually appears next to the mention of highest temperature, whereas clouds or cool weather are usually linked to lowest temperature (compare Hutchins 1992: 212). Rains, in particular heavy rains, seem also to happen on the coast more often than otherwise, which is displayed in the text. Besides, there was a difference in writing between "degrees Celsius", "Celsius" or simply "C", which had to be avoided.
After looking through theories and thinking about my topic, I discovered more methods to increase numbers of repetition in my text in terms of structure. Firstly, I tried to rephrase sentences, giving them a form according to a selected pattern. In ad-dition to it, I could break up long sentences in two, sometimes three or even four pie-ces, repeating some information purposely (in some cases I could do it parallel with rephrasing). Secondly, it was a good idea to keep sentences easy and short, either ori-ginal ones or newly formed. At the same time I thought about keeping the same word order in similar segments, which is not only effective for purpose of repetition rate but also make possible a more precise machine translation. For example, a substantive (weather, rain, wind) could be placed at the beginning of the sentence, whereas infor-mation about day time, temperature or directions - at its end. Verbs will usually fill the second place.
Linguistically it seemed appropriate to implement following restrictions: use of adjectives and adverbs - they must rather be converted into a substantive form; use of pronouns, which can be avoided by repeating the substantives; and use of the pre-sent continuous tense, then it will simplify the machine translation process (e.g. to use a future simple form "will continue" instead of a present continuous form "continu-ing").
An attempt to make a text more TM-friendly can only succeed at the cost of style and accuracies. Their meaning is secondary for a machine translation, most important is that information can be processed by the machine and transferred from one language into another. However, the text must remain accessible for human rea-ders as well, i.e. the logical sequence of it could be understood. It is very well pos-sible to come round to the golden mean. In order to do that, some of the sentences in the text have to stay unchanged or slightly changed, thus easing restrictions of the language for a machine translation ("controlled language" (Nyberg et al. in Somers 2003: 245)).
Having prepared myself in this way, I applied these methods to the first text. The analysis did not show high repetition results - only five repetitions. Of course, it was mainly due to the fact that my new translation memory was empty. Besides, my first text still had a few longer sentences and a broader vocabulary usually tends to shrink the repetition results as well. The text with the next five forecasts, however, displayed a better coefficient of repetition - as did few other weather forecast texts. I studied them and came to the conclusion that they had more TM-friendly structures already. On the other hand, my first text had one 100 % -match, one 85-94 %-match and nine matches between 75 and 84 %. Thus, my changes have had an effect on the text. After all the steps I was certain to be able to achieve better results. I improved my first text and edited my second text. After that I translated the first one.
Running the analysis tool again displayed clear distinctions between the first two texts. After the translation of the first text I had filled my TM with some vocabu-lary and it had an impact on the analysis. I could register 13 repetitions and numerous matches: eleven 100 % - matches, twelve matches between 95 - 99 % and between 85 - 94 %, five matches between 75 - 84 % and six 50 - 74 % - matches. However, my new vocabulary record still was over 50 %. I changed my third text (weather forecast file 11 - 15), which did not have any repetitions on first analysis, according to the ru-les I described and decided to use, and ran the tool once more. This time I had 18 re-petitions and my new vocabulary was under 50 %. During my translation I have had only one fuzzy match between 50 - 74 % that was of no use at all, all other matches from this category were very useful, as they could be applied with few changes. In regards to weather forecasts I was now aware of the programmes functional way and no further translations were needed. In my case it was implementation of theoretical knowledge in a practical way.
Suitability of weather forecasts as a text type for TM translation
At the heart of the machine-oriented translation concept is the idea of 'control-led language'. It can be defined as 'an explicitly defined restriction of a natural langu-age that specifies constraints on lexicon, grammar, and style' (Nyberg et al. in Somers 2003: 245). Such an idea implies some rules, although there is no fixed set of rules. Various scientists discuss different sets, all of them, however, differ insignificantly. Austermühl (2001: 164) speaks about 'Controlled English Rules', Arnold (2001: 149) is concerned with 'The Perkins Approved Clear English Writing Rules' and Nyberg et al. (in Somers 2003: 247) mention rules of 'Simplified English'. Generally a CL means restriction of the language and context in order to make a machine translation more efficient. At the beginning the translation costs were the most important factor but soon the specialists realized that 'controlled languages' can be best used with technical and specialized texts. More clear and precise translations were possible and it simpli-fied the work with translation memories. Technical translations were completed in a great many areas with their own variations of CL, sometimes called 'sublanguage' (Arnold 2001: 139). One of the fields, where 'sublanguage' is used, is meteorology.
Weather forecasts texts are usually very predictable because they are highly repetitive. They are organized in a very simple uniform way compared to other spe-cialized texts. Hutchins conducted a very extensive work on texts for the Canadian Meteorological Center. According to him (1992: 208), their range of vocabulary is limited, morphological variation is restricted (no pronouns, relative clauses, passive forms). Unlike other specialized texts, rules for texts like that can be applied at the same time (Hutchins 1992: 218). The translation of the text can precede the analysis of it (Hutchins 1992: 219), which is rather unlikely in texts with 'uncontrolled langu-age', for example, novels. The author of a weather forecast is successful in avoiding ambiguity and providing that its terminology can be reused in other similar texts. However, a weather forecast has some downsides, too. Meteorological reports cannot be applied in any other field; they are void of creativity and lack stylistic accuracies. The information output is hardly accessible for the reader, not least because of constraints in fluency. Like any other text, weather forecasts are still subject to pre- and post-editing. Of course, they are commonly used by meteorologists, who know how to work with them. They are able to link simplified sentences with lexica in order to make the text more suitable for other users. Pre-editing, however, involves the dan-ger to become time-consuming in numerous ways, from extra attention in the review of the text to the re-writing of sentences. Authors will spend much time creating the text in the first place. For translators weather forecasts constitute challenges on their own.
A TM translation programme is indeed a mighty tool in the hands of a trans-lator. In SDL Trados, for example, translation memory system applies a sophisticated method of both statistical and syntactical analysis of the sentences. TM systems are based on segments and weather forecasts suit perfectly their function. On the one hand, it seems that translators can easily work with their pattern structures. Error pro-bability of TM is very low - errors can only occur as result of transfer by humans. On the other hand, there are few difficulties in working with meteorological reports. One of them is excessive repetitiveness in the texts. Translators 'tend to think of their work in holistic terms' (Nyberg et al. in Somers 2003: 272) and need stylistic diversity. Since the aim of weather forecasts is basically different, a tendency to it can be stylis-tically unacceptable. Perhaps one of the main problems is the fact that a TM consists of different segments and cannot be presented as a complete text (Dennett, 1995: 29). A translator is confronted with a psychological aspect of translation (Dennett, 1995: 29). He must exceed the limits of a segment to think in categories of the whole con-text and see the place of the segment in it (Dennett, 1995: 29). In case of weather forecasts it can impose more difficulties on translator because linguistic and syntactical links between segments are subject to restrictions.
Overall, weather forecasts constitute a challenge in a pre- and post-editing stage; in the process of translation the benefits outweigh the disadvantages.
The re-writing of them is conducted by means of a 'controlled language'. Its main feature is functionality. Serious problems are more general in nature and affect com-mon issues of interactions between humans and computers. As with every text from specialized field of science, a translator needs to familiarize with the weather forecasts before he can actually start to work on it.
Guide for writing more TM-friendly weather forecasts:
The rules, which were used in this assignment, consist of methods, which seemed and proved to be useful in the course of text analysis and translation. These rules should help to provide more repetitions in the weather forecast texts and ensure that SDL Trados translation memories work faster and more efficiently:
Break up sentences. Form two or three short sentences. Long sentences tend to describe different weather aspects, in short sentences the information is restricted to one aspect.
Cloudy this evening and tonight with rain, heavy in places => Clouds this evening and tonight. Heavy rain in places.
Do not use synonyms. Decide for one possibility (most frequent) and replace all alternatives while editing the text.
showers => rain
dawn => morning
bright => sunny
Information about the day time or regions should also be given in one single way
Cloud thickening in the west through the afternoon bringing rain to the south and the west by teatime => Thick clouds in the west in the afternoon. Rain clouds in the south and west in the afternoon.
One or two light showers across the north of the country but otherwise the evening and night will be dry with skies clearing after dark => One or two light rains in the north. Otherwise the evening and the night will be dry. Skies will clear after dark.
Use easy and short sentences.
Saturday morning will be bright and sunny but showers will become widespread across the country during the afternoon, and many of these showers will be heavy =>
Sunny Saturday morning. Widespread rain across the country in the afternoon. Heavy rain in most cases.
Rather cloudy at first but some bright spells will develop during the day => Rather cloudy at first. Some sunny spells during the day.
Keep the same structure of the sentence.
[Substantive] - [verbs] - [direction/attribute]-[time] or [adjective]-[substantive]-[direction]
The weather will reach the west in the early night
Light winds from the northwest
Do not use pronouns but substantives or omit them completely.
Fog will form during the night, dense in places around dawn => Fog will form in the night. Dense fog in places by the morning.
Fog will form during the night, dense in places around dawn => Fog will form in the night. Dense in places by the morning.
(NOT: Fog will form in the night. It will be dense in places by the morning.)
Replace adjectives and adverbs with substantive forms.
Mainly dry => For the most part dry
Northwest breezes => breezes from the northwest
Temperatures will reach 15 to 18 degrees C in a moderate westerly breeze => Temperatures will reach 15 to 18 Celsius in a moderate breeze from the west
Do not use present continuous tense, replace it with substantive or verbal forms or omit it.
Continuing dry overnight with mostly clear skies and little or no wind => The weather dry overnight. Mostly clear skies. Little or no wind.
Temperatures recovering also to near or above normal => Temperatures will reach near or above normal
Some local sea breezes developing by afternoon => Some local sea breezes by afternoon
Do not use passive voice in your text edits - the sentences in weather forecasts always remain in simpler active voice.
In order to reach a higher repetition rate it could be helpful to reduce the vocabulary, so the whole edited text is not significantly larger (or perhaps even smaller) than the original text.
3) After few translations are finished and the TM saved various repetitive forms,
which can later be used for other texts, it is possible to introduce again more variety in the text, e.g. by using synonyms.