User Guide

SYSTRAN Desktop 7 User Guide 213
Semantic Categorization
Semantic information is not language dependent, but the effect of the clues depends on the
source or target language rules.
These clues are self-explanatory, as shown below:
• (event)
• (location)
• (work place)
• (animal)
• (human), (non human)
• (device)
• (product name)
• (first name)
• (last name)
• (title)
Protected Sequences
Protected sequences are used to exclude entries or parts of entries from the IntuitiveCoding
analysis. This can be achieved by using quotes (“ “) to indicate a partly protected sequence,
or in the special grammatical category field in SDM.
Protected sequences should be used when:
• The entry does not follow regular language rules and the analysis will therefore be
incorrect. Foreign words, foreign multiword expressions, trademarks, and other unusual
terms can cause this to occur.
• The entry is part of a sentence that is normally not expected as a dictionary entry and will
therefore not be coded by IntuitiveCoding.
• The complete entry is very long and a part of the entry can be considered as a protected
sequence that does not need specific analysis.
• The entry is incorrectly analyzed by IntuitiveCoding. Protecting the entry in conjunction
with another clue will force the entry to respect the clue without reporting a coding error.
In that case, the morphology of the entry needs to be specified since protected
sequences are considered as unknown by the IntuitiveCoding engine.
Reserved characters within a protected sequence are naturally protected.
Protected sequences should be used carefully. Protecting too many entries
reduces coding quality. Entries may lose their inflection capability and their part
of speech.