Almost every English word or Chinese character has multiple meanings. A reasonably good translation is to match all pairs of those meanings appropriately. Of course, this process is much easier than to construct a harmonic society where one is required to match all couples appropriately. That is extremely difficult because people’s lifestyles and behaviors change with time. To describe the degree of difficulty in making a good automatic translation by software, let me first introduce two terminologies to you:

(i) Randomly pick a word from a dictionary = close your eyes and open a big dictionary + move your index finger on the page and suddenly stop at a place + open your eyes and jot down the word closest to your finger.

(ii) Consciously construct a sentence by randomly picked words = make arbitrary arrangements of those words to form a set of words or a sentence that looks making a great sense to you.

Here are the results:

(a) If you only randomly pick one word, that word always makes sense because it has at least one meaning.

(b) If you randomly pick two words, the conscious constructions will likely lead to a meaningful result in most cases. For example, if two words happen to be “noun” + “adjective” or “verb” + “adverb”, the meaningful result is almost guaranteed.

(c) As the number of the randomly picked words increases, the chance of your success in consciously constructing a meaningful sentence will decrease, i.e., it will become more difficult.

My guess is that the degree of difficulty of making a good automatic translation is approximately equivalent to consciously constructing a good sentence by 10 to 15 randomly picked words. You may try this experiment and see what the success rate is.

Good luck to your experiment!

(2005.09.06)