IndraTranslation

Indonesian-English Translation inen

inen is a transfer grammar from INDRA (in) to ERG (en) (see LogonTransfer).

Transfer Grammars

inen is in tm folder, along with other transfer grammars, which can be downloaded from the Github.

For INDRA, we cloned the tm folder to grammar folder.

~/grammar$ git clone http://github.com/sanghoun/tm.git

Machine Translation

1. update and compile the source grammar. In this case, INDRA.

~/ind$ ace -g ace/config.tdl -G ind.dat

2. update and compile the target grammar. In this case, ERG.

~logon/lingo/erg$ ace -g ace/config.tdl -G erg.dat

3. update and compile the transfer grammar. In this case, inen.

~/grammar/tm/inen$ ace -g ace/config.tdl -G inen.dat

4. translating by INPUT | PARSING | TRANSFER | GENERATION

$ echo "anjing menggonggong" | ace -g YOUR_INDRA_DIRECTORY/ind.dat | ace -g YOUR_INEN_DIRECTORY/inen.dat | ace -g YOUR_ERG_DIRECTORY/erg.dat -e

Transfer Rules

1. Edit in.vpm under ~/grammar/tm/inen to transfer for example:

• the underspecified tense in Indonesian into present tense in English,
• third person (underspecified for number) in Indonesian into third person plural in English,
• perfect aspect in Indonesian into past tense in English.

E.TENSE : TENSE
  tense >> pres

PNG.PERNUM : PERS NUM
  3rd >> 3 pl

E.ASPECT : TENSE
  perf >> past

2. In out.vpm

TENSE : TENSE
  * >> *

PERS : PERS
  * >> *

NUM : NUM
  * >> *

3. In test.mtr, map menggonggong to bark and anjing to dog

bark_mtr := monotonic_mtr &
[ INPUT.RELS < [ PRED "_menggonggong_v_rel" ] >,
  OUTPUT.RELS < +copy+ & [ PRED "_bark_v_1_rel" ] > ].

dog_mtr := monotonic_mtr &
[ INPUT.RELS < [ PRED "_anjing_n_rel" ] >,
  OUTPUT.RELS < +copy+ & [ PRED "_dog_n_1_rel" ] > ].

udef_mtr := monotonic_mtr &
[ INPUT.RELS < [ PRED "exist_q_rel" ] >,
  OUTPUT.RELS < [ PRED udef_q_rel ] > ].