Apertium has moved from SourceForge to GitHub.
If you have any questions, please come and talk to us on #apertium on irc.freenode.net or contact the GitHub migration team.

Task ideas for Google Code-in/Add transfer rule

From Apertium
< Task ideas for Google Code-in(Difference between revisions)
Jump to: navigation, search
(regression test)
 
Line 6: Line 6:
 
# Write a new rule or correct an existing rule.
 
# Write a new rule or correct an existing rule.
 
# Compile and test again.
 
# Compile and test again.
  +
# Add a regression test to the appropriate wiki page, so, for example, if you're doing Kyrgyz->English, add a test to [[English_and_Kyrgyz/Transfer_tests]].
 
# Submit a pull request on github and submit the url of the PR to your mentor.
 
# Submit a pull request on github and submit the url of the PR to your mentor.
   

Latest revision as of 22:39, 15 December 2019

  1. select a language pair, ideally such that the source language is a language you know (L₂) and the target language a language you use every day (L₁).
  2. Install Apertium locally; install the language pair; make sure that it works.
  3. Using a large enough corpus of the source language (e.g. plain text taken from Wikipedia), detect one structural transfer rule (.t1x, .t2x, .t3x) that is wrong or missing (local agreement, gender, number, etc. is inadequate, local word order in a phrase is inadequate, there is a word too much or a word missing, etc.).
    • In other words, run a bunch of text in L₂ through the pair and find a consistent error in the output text in L₁ that isn't grammatical.
  4. Write a new rule or correct an existing rule.
  5. Compile and test again.
  6. Add a regression test to the appropriate wiki page, so, for example, if you're doing Kyrgyz->English, add a test to English_and_Kyrgyz/Transfer_tests.
  7. Submit a pull request on github and submit the url of the PR to your mentor.
Personal tools