Ideas for Google Summer of Code/Adopt a language pair

This project will involve writing linguistic data, including morphological rules and transfer rules — which are specified in a declarative language. A good intro would be to look through Apertium New Language Pair HOWTO, see also Contributing to an existing pair. If the pair has OK dictionaries but a bad tagger (disambiguator), a GsoC project might include writing a good Constraint Grammar for the pair.

Coding challenge

The coding challenge for this task is to:

1. Install Apertium (see Minimal installation from SVN)
2. Go through the HOWTO
3. Go through the MT course here (или здесь)
4. Write a translator that translates as much of this story (or from here ) as possible — Minimum one sentence.
   • If there is no translation, translate it into the languages of your language pair first.
5. Upload your work to github (or equivalent).

If you don't complete it all, don't worry! We take many things into account when assessing your application. However, the URL to any work you do for the coding challenge work should be included in your application.

Frequently asked questions

• Can I do a pair with language $x$ and language $y$ ?

— Yes, there are no restrictions. But you should take the following into consideration: (a) Are there existing machine translation (MT) systems for this pair? (b) If there are existing systems, how good are they? -- Could you do better in three months? (c) How closely related is the pair? (d) How many resources already exist for the pair? (e) Are there any mentors who can evaluate your work?

As an example, we're very unlikely to accept applicants to eng-hin, which (a) has support from existing major systems (b) where the existing systems are much better than you'd get in 3 months and (c) which is not closely related. Too many students have tried this pair in the past, and not gotten anywhere. Fortunately, many students who speak Hindi and English also speak a third language – we're much more likely to support that.

• Do I need to have GNU/Linux installed, or can I use another operating system ?

— In theory you can use any operating system. In practice unless you are using GNU/Linux or Mac/OS you are going to have a hard time as the mentors cannot offer you support with alternative operating systems. You may want to check out Virtualbox if you are using Windows.

• What programming languages do I need to know ?

— For making a language pair, you don't need to know any specific programming language. Knowing a scripting language will be really helpful, but most of the work is done in Apertium's own linguistic formalisms, which are based on XML. To get an idea of what these formalisms look like, you should do the new language pair HOWTO.

• Do I have to know both language $x$ and language $y$ ?

— You don't have to be a native speaker of the languages, but you should be a fluent speaker of one, and have a good knowledge of the other.

• Do I have to be able to speak English well?

— No, it isn't necessary to speak English well. If you and your mentor can communicate, then we are happy. However, you should have a basic reading level of English and be able to ask questions when you don't understand.

• What are the criteria for a language pair to be considered "finished" for GSOC purposes ?

— This is something to discuss with your mentor, but a general idea might be that the pair is testvoc clean, and has a coverage of around 80% or more on a range of free corpora.

• If I take on a language pair, do I have to do both directions, e.g. $x$ → $y$ and $y$ → $x$?

— This will depend on your language pair, you will need to discuss it with your mentor. As a rough guide, if the pair is for dissemination, e.g. translating between related languages, then both directions will be made. If it is for assimilation, then perhaps only one.

Previous GSOC projects

Here is a list of pairs which were adopted in past years. In brackets following the pair is the current status in the repository.

• 2012
  • Indonesian and Malaysian (trunk)
  • Slovenian and Serbo-Croatian (trunk)
  • Maltese and Arabic (staging)
  • Kazakh and Tatar (trunk)
  • Quechua cuzqueño y castellano (nursery)
  • Turkish and Tatar (incubator)

• 2011
  • Serbo-Croatian and Macedonian (trunk)
  • Turkish and Azerbaijani
  • Turkish and Kyrgyz (staging)
  • Maltese and Hebrew (staging)
  • Slovenian and Spanish (nursery)
• 2010
  • Macedonian and Bulgarian (trunk)
  • French and Portuguese (staging)
  • North Sámi and Finnish (nursery)
  • Afrikaans and Dutch (GCI)
• 2009
  • Swedish and Danish (trunk)
  • Norwegian Nynorsk and Norwegian Bokmål (trunk)

Some pairs with suggestions

Pair	Current state	What needs to be done	Who ? (mentors)
eo-en (trunk)	The Esperanto->English pair was never really done properly. Therefore strange synonyms are selected, like 'novaĵo' (news item) it translated into 'departure' and 'loĝi' (to live) into 'populate'. Moreover the coverage is low because Esperanto has a grammar where you can effortlessly transform any word into any other word class (like making a verb into a noun) and do compound words	It would require a lot of cleanup (coding of scripts), introduce lex-tools scripts, introduce robust compounding, including do some pondering/experimenting on how to handle compounding. The result is expected to be applied on creating eo->es, eo->fr and/or eo->ca directions as well. Experimenting with a better English tagger disambiguation would also be nice if there is time left.	Jacob Nordfalk

See also

• List of language pairs
• an example work plan for a language pair: http://wiki.apertium.org/wiki/Maltese_and_Arabic/Work_plan