toki pona lon sona pilin en sona ante

[janMato]

Here is the full text of the article on ROILA. jan Ante found a link earlier, but that link requires registration and this one doesn't, otherwise I think they're the same article.

http://www.bartneck.de/publications/201 ... tionROILA/

[janAlopo]

The first classes in ROILA have concluded. High school students, two-week course, about 100 students in four classes met three times in class to practice together and play with Lobo the Robot (takes orders and converses in ROILA), studied during the week using online course in Moodle (using first 7 lessons from http://hytext.com/roila). Student loved it. Follow-on study comparing efficacy of learning ROILA to control robot vs controlling English-savvy robot with English favored ROILA. School has asked for another course, new students, in January. toki pona was the inspiration for the whole effort!

[janMato]

pona a!

That gives me an idea-- what would it be like to order a robot around in toki pona? Unlike roila, the geographic words in toki pona are incomplete. forward/backward/up/down/to a fixed item in the context/towards the inside of a context ... no left or right. Giving commands to a robot would automatically become a puzzle because you'd have to pick a path that gets the robot to the destination. One would have to add unnecessary items to the environment to create a world where the are enough contextually based directions.

[janAlopo]

Robots need a pretty concrete instruction set. As an experiment, I pushed the toki-pona/ROILA idea a step further and am working on a conlang "lomijo" with three vowels (iou), six consonants (jklmnp), strict CV, CVCV, CVCVCV a priori word list (pared to eliminate things like mimimi, pupupu, etc.), no adjectives or adverbs, isolating grammar, with marker-flags for plurals, tenses, noun-from-verb, verb-from-noun, adj/adv from noun/verb, etc. The idea is make it machine-parsable, loose-pronunciation-tolerant, and at least semi-euphonic. The fact that it can typed on a QWERTY keyboard using just the right hand, from normal home-key positioning, is a happy serendipity. Also working on a written script orthography, one character for each of the 18 phonemes. Yes, I am in WAY over my head!

[janMato]

I'm half way through writing a toki pona parser from scratch. I'm finding the hardest part to be segmenting- some phrases like "pi" phrases segment easily, other phrases (like prep phrases), not so easy. Maybe the problem is just hard. For example in XML to segment an element that is a great-great-great grandchild of the root and the child of a root would take boat loads of angle brackets, but XML is always segmentable.

The other thing that caused me problems were segmenters that could be followed by a variety of things, like "li" can be followed by a predicate or a verb+DO. Segmenting those would be easier if there were a different particle for verb+DO and predicates.

What are you using to parse your conlang?

[janAlopo]

First I write the language, then I'll write the parser. I've written markup language parsers in flavors of BASIC for previous projects. Also some quick lookup from text translation aids in JavaScript. Any language with solid regexp routines and efficient hash (associative array) lookups will do. First step (after language design!) will be speech-to-text processing.

[jan-ante]

, i.e. precisely on "1 juni 2011 om 16.00 uur".
have a look http://www.drhu.eu/reports/2011-OmarMubin-ROILA.pdf

[janAlopo]

A stunning achievement! A PhD for a conlang! Congratulations to Dr. Omar Mubin.