Labs/Ubiquity/Meetings/2009-05-13 Weekly Meeting

Details

Time

5:00pm Pacific time

Location

• IRC channel: #ubiquity
• Dial in:
  • +1 800 707 2533 (pin 369) Conf# 202 (US Toll Free/Skype)
  • +1 650 903 0800 x92 Conf# 202 (US/International)
  • +1 416 848 3114 x92 Conf# 202 (Canada)

Agenda

1.  ?

Attendees

1. mitcho
2. satyr
3. marsf
4. cers

Notes

• So this meeting turned into a regular i18n meeting... so some status updates:
  • mitcho's been working on some features useful for romance languages
  • also accepted some patches from marsf for the Parser 1 Japanese commands...
    • ...but also a reminder that Parser 2 is the future.
  • also rewrote whitespace handling last week on satyr's suggestion
  • New parsers: new Catalan parser
    • mitcho: felipc said he will try adding the new parser features into the Portuguese parser later this week.
    • cers is contacting Swedish and Norwegian community members
    • Contributors can look at this tutorial: http://mitcho.com/blog/how-to/adding-your-language-to-ubiquity-parser-2/ They're also more than welcome to find me (mitcho) on IRC if they have any questions or run into some difficulty.

• Q&A (mostly on IRC)
  • marsf: how does normalizeArgument work with languages with no spaces?
    • mitcho: normalizeArgument was created mostly to handle some weird complexities of romance languages and how they represent male and female arguments... in languages like Japanese or Chinese (which don't have spaces) I don't think there's a need for a custom normalizeArgument. The default normalizeArgument (defined in parser.js, which simply returns []) should be fine for the majority of languages.
  • satyr: will Parser 2 support commands which have tons of non-natural language arguments? (ex: http://gist.github.com/60561)
    • mitcho: no... if a command wants to accept a specific syntax of arguments, it can take a single noun_arb_text or custom nountype argument and try parsing it itself.
  • satyr: will Parser 2 support two arguments in the same role?
    • mitcho: yes, Parser 2 does currently parse multiple arguments into the same role, and detect the nountypes for all of them, but such parses are currently scored down... in general, it's something we should probably discourage.
    • cers: In the future many of these many-argument verbs may be better served with command chaining of multiple verbs.
  • satyr: will Parser 2 deal with old commands?
    • mitcho: no, Parser 2 will not automagically convert old commands into the new format. Jono and I came up with a way to write hybrid Parser 1- and Parser 2-compatible commands, and we'll blog about that format soon. For now, the "translate" command may be the best example.
    • cers: perhaps we should warn the user when they have old commands loaded in
  • marsf: should we add an order specification to the verbs' .arguments property?
    • mitcho: I don't think that's necessary. I think in most of the languages we're looking at, the order of arguments don't have to be used as a way to identify arguments. There might be "more natural" orders than others, but we (the parser) don't have to choose one and not the other--and, even if there was a langauge where we had to look at word order, that should happen in the parser itself, not be specified in verbs. The idea with semantic roles is that the verbs just need to specify which kinds of arguments it takes and the parser figures out what to give it for whatever language the user inputs.