See also Blair's list of localization issues to consider for Uplift.
- what is the goal?
- some initial thoughts: what kind of "natural" interface are we going for?
- What are the linguistic features of the commands we hope to deal with?
- parser design
- how to collect data - initial thoughts
- how are arguments marked in different languages? (contribute here)
- argument structure alternations (e.g. ditransitive alternation in English)
- arguments with no role markings
- an inventory of semantic roles for use in matching arguments to verb arguments
- parsing languages with case-marking
- parsing arguments with portmanteau'ed articles
- mapping argument structures between languages
- do the same oblique arguments (prepositional phrases) map cleanly into prepositions in other languages? Short answer: no.
- how do you parse for verb-final languages? => see Japanese for the time being
- the structure of imperatives in different languages
- what do imperatives look like in so-called "information-structure-heavy" or "topic-comment" languges?
- the position of the verb in target sentences
- multiple possible word orders: infinitives + subjunctives in imperative contexts (see German, Greek, etc.)
- pronouns, a.k.a. "magic words"
- how do deictic pronouns behave in imperative contexts?
- do we/should we distinguish between different kinds of pronouns?
- how do you learn new commands? => see friendlier command feed subscription
- what does learning/teaching natural languages teach us about how users can learn Ubiq?
- how to localize Ubiquity
- the presentation of suggestions