Embed Size (px)
Citation preview
6 Computer
LANGUAGE TRANSLATION
In “Languages and the ComputingProfession” (The Profession, Mar.2004, pp. 104, 102-103), NevilleHolmes describes a method of auto-mated language translation using astandardized “completely unnatural”intermediate language and discussesvarious problems. This method maywork well for translating the literatureof various technical fields because theyhave well-defined vocabularies.
The problems Holmes discusses aremore serious in fields of human dis-course outside the technical areas.Because human languages do notmatch well with regard to vocabulary,phrases, puns, and so forth, any trans-lation that a human creates involvesmaking subjective choices in translat-ing words and other elements of thesource language. These choices dependon the particular translator’s biases.
Even if computers perform the trans-lations, a degree of subjectivity will bepresent in the translation software sinceit is unlikely that there could be a one-to-one mapping of the words, phrases,and so on in all human languages to theintermediate language.
In addition, for general literature, thecharacteristic of literality is problematic.Idioms, clichés, hackneyed phrases, andthe like cannot be excluded without pre-venting the richness of expression insource language documents from beingconveyed in the destination language—and these are the areas where transla-tor biases are the most evident.
Holmes’s discussion of work to bedone shows that he has thought aboutthese matters. However, he does notexplicitly discuss subjectivity. I aminterested in knowing if he expects thatsubjectivity can be eliminated from theprocess.Martin SachsWestport, [email protected]
Neville Holmes responds:The implication that there can be nosubjectivity in the actual machine
translation is well made. The machineprocesses data; the information, andthus the subjectivity, can only be in theminds of the people using or makingthe software.
To avoid, or at least lessen, the build-ing of bias into the software was whyI emphasized the importance of havingphilosophers (I had ethicists particu-larly in mind) and semanticists centralto the project. Indeed, it is anothergood reason for such a project to beunder the aegis of the United Nations.
On the other hand, the bias that anauthor or reader inevitably imposes ontext, even in technical fields, is won-derfully human, and the last thing Iwould want to do is eliminate it. Thatis why I suggested that departures fromliterality, perhaps the most obvioussource of bias, might be encoded punc-tuationally in the intermediate languageso that translation from the intermedi-ary could—when we’ve worked outhow—deal with it appropriately.
Furthermore, my suggestion of adding“parameters that allow selection [anddetection] of styles, periods, regionali-ties, and other variations” to translationprograms would, for instance, providefor a document in English with one spec-trum of biases to be translated into theintermediate language and then backinto English with a completely differentspectrum of biases.
What I am suggesting merges withinterpretation in the long term, butthere will be some texts that cannot beinterpreted, only mimicked. Oneexample is the kind of “Wockerjabby”doggerel that went the rounds quite afew years ago:
Eye halve a spell ling check err.Eat came whither peace see.
Eat plane lea marques form I revue.Mist ache sigh mite knot sea.
I’ve run this pome threw eat,Aim shore yawp least two no.
Its let err perfect inn it’s weigh.My chequer tolled miso.
COMPILER ENHANCEMENTS
The techniques that Peter Maurer out-lines in “Metamorphic Programming:Unconventional High Performance”(Mar. 2004, pp. 30-38) indeed have asuccessful history among softwareengineers emulating CPUs (or virtualmachines) and creating fast statemachines. The sources below provideadditional explanations of the tech-niques as they are employed in varioustasks:
• A. Ertl, “Threaded Code;” www.complang.tuwien.ac.at/forth/threaded-code.html.
• E. Gagnon and L. Hendron,“SableVM: A Research Frame-work for Efficient Execution ofJava Bytecode,” Proc. Java VirtualMachine Research and Tech-nology Symp., Usenix 2001; www.usenix.org/publications/library/proceedings/jvm01/gagnon/gagnon.pdf.
• E. Miranda, “Portable Fast DirectThreaded Code,” 29 Mar. 1991;compilers.iecc.com/comparch/article/91-03-121.
• B. Hoff, “High-Speed Finite StateMachines, Dr. Dobbs J., Nov.1997; www.grouse.com.au/ggrep/.
• GCC Manual, “Labels as Values;”gcc.gnu.org/onlinedocs/gcc/Labels-as-Values.html#Labels%20as%20Values.
As Maurer explains, there is perfor-mance to be gained by using proceduralcode. There may be two explanationsfor this. First, the label-as-value tech-nique treats the compiler as a macroassembler, better matching how theunderlying hardware works. Second, theperformance ratios may be larger whenusing the GNU Compiler Collection.
L E T T E R S@@
