Instruction Manual

User’s Guide SYSTRAN 5.0 172
Appendix B. Preparing English Text for MT
This appendix provides guidelines for preparing English text for use with machine
translation (MT).
General Principles
The three most frequent problems found in analyzing English sentences are:
• Very long sentences, typically including more than one sub-clause.
• Sentences with incorrect punctuation usage, including sentences with too much
punctuation.
• Use of lexically or syntactically ambiguous words.
To improve translations, follow these writing recommendations:
• Write shorter sentences.
Restrict a sentence so it expresses one or two thoughts. Multiple sentences,
connected by the appropriate conjunctions, normally result in better machine
translation.
• Use fewer sub-clauses.
This suggestion is closely related to sentence length. You should avoid sub-
clauses when possible, except in obvious cases, such as conditional (If…
then…) phrases.
• Use punctuation wisely.
Machine translation can digest only so much punctuation. Therefore, use just
enough punctuation to clarify ideas. Too much punctuation, especially commas,
reaches a point of diminishing return quickly.
• Avoid using syntactically or lexically ambiguous words.
Use grammatical markers (such as determinants) to mark the word category. Use
adjectives to specify the noun (for example: “activation key” instead of “key”).
The following sections provide additional information about these rules.