Translation and Disambiguation

December 5th, 2009
As any professional translation agency or freelance translation service would attest to, disambiguation and human translation are intrinsically connected. Without word sense disambiguation, searching for translation equivalence would be extremely difficult, cultural gaffes would occur more often, accommodation would muddle the original more instead of clarifying it for the sake of the target audience, and the underlying message of the source text would ultimately be lost in a bilingual pile of gobbledygook. Then again, integrating disambiguation into the human translation process has its difficulties as well. For example, one problem with disambiguation is choosing what the senses are. Sometimes, senses can be obviously different, while other times, the different senses can be closely linked together (for example, a meaning may be a metonymic or metaphorical extension of another). During the latter case, dividing the words into senses becomes extremely hard. Word Sense Disambiguation Problems Professional translation companies go through the troubles and hardships of disambiguation almost every day, and as such have devised a system in order to specifically handle that translation front. Because different dictionaries will yield different divisions of words into senses, a translation service provider will usually stick to one dictionary and just use its set of senses as a standard of sorts. There are translators who also keep in mind that research results that use general distinctions in senses are much better than specific ones, so most researchers combat disambiguation problems by ignoring the fine-grained distinctions in their work. Inter-judge variance is another difficulty translators usually face when going about the disambiguation process. Truth be told, different people have different takes on any given list of senses and sentences; they will not always agree on which word belongs to which sense, which induces ambiguousness and unclearness. What's more, humans are much better on coarse-grained than fine-grained distinctions of a given translation, which is why disambiguation is mostly based on the latter approach on distinction. Word Sense Disambiguation Approaches There are two main schools of thought when going about word sense disambiguation: shallow approaches and deep approaches. Shallow approaches concentrates more on the context to which a word is used (the surrounding words) and doesn't try to understand the text. For instance, if the word "bass" is surrounded by words like "fish" or "sea", then "bass" means a type of fish. Conversely, if "bass" has the words like "song" or "music" beside it, then it's used in a musical sense. In this context, even a computer can be used to go about the shallow approach disambiguation. On the other hand, the deep approach is the method that's closer to human translation—in fact, the type of approach that expert translators use—such that a comprehensive body of world knowledge is used to discern what context a word belongs to. Knowledge here translates to understanding the fact that you can't fish for low frequency sounds and a certain type of fish has nothing to do with your song's sounds. In computer use, such approaches aren't very successful because computers at present are incapable of handling comprehension of that caliber.

You might also like: