Hello, /development

[TireaAean]

Kaltxì.

Are there any other Linux users out there?

Also, who knows PHP and would like to help me out with fixing the dreaded navilator [Navilator.com]?

[Sight Unseen]

I use Linux, most of my knowledge is in the Debian side and not RHEL, but I have had to work on RHEL/Fedora a few times.
I understand a little PHP, learned just enough to make dynamic pages on a few sites, that's about it.

If you need help with Linux, I can probably help you. PHP? Not so much.

[akiwiguy]

I know PHP
But Swoka Ikran would also be a good guy to talk to about this.

[stdout]

Hi TireaAean!

I use Linux, but I don't know any PHP I'm afraid.

[Human No More]

I know some Linux, but really only know the basics of PHP, so probably not in a position to help even if I did know Na'vi.

[txen]

Time is my commodity in short supply, but what help do you need?

Been makig a living with Linux and PHP for the last ten years so yea I know a little.

[TireaAean]

awesome, awesome. thanks for the responses.

Basically, I am on a quest to improve Navilator. Plain and simple, it kinda sucks. I was wondering if we could come up with a way to make it similar to Google Translate. It would be a big project, but one on the backburner in priority.

Currently, the PHP source for it is basically a large hardcoded array of old words and their definitions. it is oblivious to grammar and different forms of words.

idk if any of you would have any interest in this project or not, but I figured it'd be worth a shot asking around.

[Clarke]

Google Translate relies on large texts available in multiple languages, from what I understand. Na'vi examples seem lacking, unfortunately.

[Human No More]

Yeah, google translate builds its algorithms from publicly available documents (like ones from the UN, I think) which are translated into various languages, I think it works like a genetic algorithm for improving. For a simple improvement, you might just want to change the word/definition to a database format rather than hardcoded, would at least make it more maintainable.

[Clarke]

You could try writing a formal grammar for Na'vi, but I'm not sure how easy that would be.

[TireaAean]

William Annis' Horen Lì'fyayä leNa'vi is as close as we will get to a formal grammar for Na'vi. no one cares about that. they want a translator machine to do the work for them. (I dont like that, and its just better to learn the language, but whatever) It just occured to me that if the navilator site is up, it might as well at least do its job well.

Unfortunately, it will be difficult to have everything automated and nothing hardcoded. so many exceptional conditions exist such that a machine will have trouble stripping affixes off a word unless the machine is highly sophisticated. Still worth a shot and could be fun.

I'll be thinking on this.

[Clarke]

Well, if you had some method of reliably lexing Na'vi, (i.e. parse it and assign meanings to some parts) that would help significantly. The lenition rules might lead to a combinatorial explosion, but that shouldn't produce too many problems here.

I'm sure it oculd be done, it'd just be a bit complicated.

[TireaAean]

the main issues are lenition, and lexical roots which APPEAR to contain valid infixes but don't.

eg there is a verb infix ay, the future tense. (ay is also a lenitioncausing prefix for making nouns plural)

there is a root verb tswayon, to fly. tswayon with ay infix becomes tswayayon. tswon does not exist.

ay infix actually comes first as if a prefix on verbs which begin with a vowel, eg omum, to know. ayomum, will know.

so yeah, there are little quirks like that which people pick up on but is hard to tell a machine without making the code very hackish.

wht do you mean by "lexing Na'vi, (i.e. parse it and assign meanings to some parts)"?

[Clarke]

Originally Posted by TireaAean

the main issues are lenition, and lexical roots which APPEAR to contain valid infixes but don't.

eg there is a verb infix ay, the future tense. (ay is also a lenitioncausing prefix for making nouns plural)

there is a root verb tswayon, to fly. tswayon with ay infix becomes tswayayon. tswon does not exist.

(Oh come on, Frommer, you didn't have the decency to make the language unambigious? )

How many of those problems could be solved by having the dictionary and a list of infixes on hand? (Perhaps you might also need the infix positions of all the verbs) After all, tswayon would appear in your dictionary as a verb, and so you could tell that tswon doesn't exist.

wht do you mean by "lexing Na'vi, (i.e. parse it and assign meanings to some parts)"?

The "naive" way of parsing text, usually used with HTML, is to use pattern matching techniques to grab the text in a certain place. (For instance, grabbing all the images by looking for all the URLs inside img tags.) A more complex but more useful method is to parse and understand the of the text, i.e. You know English well enough that you can see that there's a noun, specifically, missing from that last sentence, rather than just any word. One common way of doing that is explained here.

Once you've "understood" the structure of the text, translating it becomes a lot easier because the program can see where lenition, infixes, etc, might appear.

[TireaAean]

Yes. My attempts in my vrrtep project for stripping affixes are proving very messy. Im doing it in the most naive way possible, methinks.

Also, Pawl LIKES the ambiguity in the language. There is no natural language on Earth which does not have at least one way to create ambiguity. He designed Na'vi in such a way that it would at least seem natural, and not "sterile", having no exceptions to rules(even though there really are few), or natural change over time(eg contractions).

Thanks for the link. In Vrrtep, My problem is, I have a list of infixes, and I am using a containment operator to see if an affix is in a word. Probably not the best way to do it. I need to put more thought on this...


See, if I can get vrrtep to work really well, the navilator can just use that as backend for translations.