next up previous contents index
Next: Sublanguage MT Up: Input Previous: SGML Markup and

Controlled Languages

 

The notion of controlled languages was introduced in Chapter gif where we described it as a form of language usage restricted by grammar and vocabulary rules. The original idea arose during the 1930s, when a number of influential linguists and scholars devoted considerable effort to establishing a `minimal' variety of English, a variety specifically designed to make English accessible to and usable by the largest possible number of people world wide. Basic English , as it was called, differed from previous attempts to construct universal languages in that it was a perfectly well-formed part of English, rather than some entirely artificial or hybrid construction such as Esperanto. One of the central ideas of the Basic English movement was that the number of general-purpose words needed for writing anything from a simple letter of receipt through to a major speech on the world economic situation could be a few hundred rather than the 75 000 upward available to skilled native speakers. This lexical economy was to be achieved in part by using `operator verbs' with the set of nouns and adjectives to stand in for the vast number of derived verbs which are frequently used. For example, whereas in ordinary English we might write The disc controller design was perfected over numerous revisions, Basic English  would say ... was made perfect ..., where make is one of the operator verbs and perfect one of the licensed Basic English adjectives.

The authors of Basic English  explicitly recognised that the dictionary would need to be extended with special terminology for scientific and technical writing. However, even if a text contained terminology specific to a certain subject field, the general language component of the text could perfectly well be accommodated within Basic English. The important point remains that, for writing in a particular subject field, no more is needed than the Basic English dictionary together with a (relatively small) technical vocabulary for that field.

The idea was later taken on by English-language based (predominantly North American) corporations marketing capital goods on a world-wide basis. Rather than try to translate engine manuals and the like into every possible language that might be required, it was assumed that if they were written with sufficient care and attention, they could be read fairly easy by service engineers and mechanics with limited English skills.

Although controlled languages were introduced partly to avoid or reduce human translation costs, two important additional benefits were discovered. First, the readability and clarity of a controlled language technical text often seems better than uncontrolled texts --- even for native English readers. Second, controlled languages produce better results with MT than uncontrolled languages.

The reasons for controlled languages' superior MT performance are easy to understand. First, the restricted vocabulary means that fewer words need to be added to the MT system dictionaries and more effort can be put into getting the entries which are required right. Second, the grammar component of the system can be tailored to handle all and only those constructions which are licensed by the controlled language specification, a specification which excludes the most difficult and ambiguous  constructions anyway.

 

 

 

A flavour of what is involved can be obtained by looking at the writing rules given above and the dictionary excerpt on page gif, which are based on those of PACE, the controlled English used by the UK Engineering company  Perkins Engines.gif As will be clear from the dictionary excerpt, the general principle is `one word, one meaning', for example, the only use of the verb advise is `to give advice'. Thus, a usage such as Please advise us of the availability of parts at your earliest convenience would not be allowed, since here it means `tell'. A useful development of such a dictionary for MT purposes would be to add information about how these words translate.

Using a restricted pool of words and terms also means that the system dictionaries can be tailored (by the MT supplier or responsible translator) to cover exactly that set of words and their translations. Being consistent about the use of terms will also help to improve the overall consistency and quality of the texts being translated. After all, one of the simplest and most direct benefits of MT for technical texts is that terms are always translated consistently because they are simply looked up in an electronic bilingual term dictionary .

 

 

In general, it can be seen that the rules are mainly advice on constructions that should be avoided, usually because they lead to ambiguity . The rules for controlled languages tend to be stylistic guidelines rather than hard and fast grammar specifications. In general, much of the success of controlled languages as corporate language tools stems from the emphasis placed on critical analysis of the text and precise presentation of ideas. This is particularly apparent in the first example on page gif, which illustrates the dramatic effect of using a controlled version of English.

It is not particularly difficult to train people to write controlled language text i.e. text which generally observes some set of fairly simple writing rules. For example, the Xerox  corporation currently offers its technical writers a one-day course in writing with MCE (Multinational Customised English, a Xerox proprietary language).  British Aerospace  teaches the rudiments of Simplified English  (a general purpose technical English for the international aerospace industry) in a few fairly short training sessions.

 

 


next up previous contents index
Next: Sublanguage MT Up: Input Previous: SGML Markup and



Arnold D J
Thu Dec 21 10:52:49 GMT 1995