|
Pattern-based Translation Technology
Pattern-based translation uses a huge collection of translation patterns, each
of which is a pair of source context-free-grammar (CFG) rule and its corresponding
target CFG rule, and makes a translation by matching these translation patterns
to the input. The following examples show translation patterns.
(a) NOUN NOUN:1 -> NP:1 NP:1 <- NOUN NOUN:1
(b) have:VERB:1 to do with NOUN:2 -> VP:1 VP:1 <- NOUN:2 TO(with) KANKEI(relation) GA(subj) ARU(exist):1
The first pattern is a syntactic pattern which makes a compound noun from
adjacent two nouns, on the other hand, the second pattern is an idiomatic
pattern. Each number following a semicolon specifies correspondences between
rule terms.
Like this, pattern-based translation deals with syntactic translation
knowledge and idiomatic or colloquial translation knowledge by the
uniform framework.
Translation is done as follows: At first, an input sentence is analyzed by
a CFG parser using source CFG rules of translation patterns, and when a parse tree is given,
a target structure is created by synchronous derivation using
target CFG rules corresponding to used source CFG rules.
The advantage of this system is that it is very easy to add translation
knowledge to the system since most translation knowledge are expressed
as translation patterns, and this can lowerize the cost for MT system
customization.
|
