Re: [lojban] Questions about jorne

coi uiban

Sorry for not getting back to you sooner!

[snippet A]

On Mon, Apr 19, 2010 at 12:30, Brian D. Eubanks <brian@buildsoftware.com> wrote:

The project is currently focused on these tasks:
1. use Lojban text to add statements to knowledge bases, or ask questions about the content

2. describe data contained within knowledge bases, using Lojban text

[snippet B]

SKOS is also an excellent way to implement a Semantic-Web-friendly thesaurus.

[snippet C]

...using technologies such as in the links given (perhaps starting with a Wordnet implementation?) would bring great benefit to the Lojban community.

So the way I see it, there are really three separate tasks here: [A] lojban knowledge authoring and extraction, [B] describing lojban using SKOS and [C] describing Wordnet using lojban, all of which I think are brilliant and awesome.

As for task [A], knowledge authoring and extraction, I think it has obvious applications with the other two, but could stand alone: lojban is already parseable (well, you know) so there's nothing stopping us from having an online lojban submission form for entering parseable knowledge (that is, any lojban text). In fact its frighteningly similar to the recent page http://lojban.org/cgi-bin/corpus, but optimally I think there should be parser validation upon submission. Searching it for completions of queries is step two. Of the applications mentioned this one, i think, is the easiest to implement: it's just a web-based parser [submission] and search [extraction].

When I say knowledge extraction, I'm thinking something like this:

>> mi klama ma //input query

mi klama lo zarci

mi pu klama lo jarbu

... //output

Task [B], describing Lojban with SKOS, I think is slightly harder if done properly, and very valuable, as it would extend the utility of [A]. Searching for { ma danlu } could give a list of all animals, and could also search for all knowledge related to animals.

Start with the current TeX thesaurus from the lojban website.
Nice lojban uris. The way things work in Lojbanistan, it seems, is that the keys to the castle are available as soon as there is proof of concept. Once there's a queryable web interface, of the lojban corpus or some translated datastores, I'm sure Robin or whoever will let www.lojban.com/valsi#gismu be used for this. I have a domain name on a shared host if there's need for development space for proof of concept.
Immediately it's a perfect thesaurus format and so one trivial-to-implement but crucial-to-have feature is importing the latest jbovlaste xml dump for adding new gloss words as prefLabels etc.
References to other RDF schemas, not just wordnet [http://www.w3.org/TR/wordnet-rdf/]. For example, animals can be likewise translated to http://ontologi.es/biol/ns and linked to dbpedia pages, yadda yadda. This extends the domain of part A, as you put it, exponentially: open data would be searchable in a brief and robust form. This is what lojban dreams of when it says it may someday be used to talk to machines.

Now we could have input like:

>> lo cribe cu xabju ma //where do bears live

...//returns habitat information from all dbpedia entries referenced in 'cribe'

As for task [C], describing Wordnet synsets using lojban, I think this is the hardest, not merely because of the sheer number of synsets, but because of the mandate for precision-- synsets should have unique lojban terms. Semantic granularity on WordNet is pretty small, and our mere seven thousand or so current valsi are insufficient. For the part [B] task (referring to other URIs that a single lojban word *could* describe), some degree of polysemy/vagueness is fine, but for the task of describing wordnet, I think the best implementation has a one-to-one correspondence to lojban words.

This obviously entails a huge number of ad-hoc lujvo/tanru (ideally) or fi'uvla (suboptimal) that need to be created for similar shades of related synsets, or just technical or cultural words that have no lojban equivalent or approximate. I envision this only as feasible in a wiki-like, folksonomy web frontend where multiple people can help assign new lujvo to unassigned synsets. A great byproduct of this would be critical examination of the shortcomings of current gismu, and possibly accelerated specification of vague/poorly-defined terms and newly coined lojban terms.

---intermission---

I have a little proposal to distinguish these three tasks as I have described them ([A],[B], and [C]). While they are all part of the same project, I think they each play functionally distinct roles; namely a human portal, an extendable description, and the holy grail of lojban dictionaries. My proposal is that these three tasks be called jorne, sejorne and tejorne respectively. jorne is still the name of the whole idea (since the other two work through or enhance it), and

1) the 'connect' {lo jorne} refers to the user front end and basic input/query portion

2) the 'things-connected' {lo se jorne} refers to the linked searchable/translatable ontologies/schema

3) the 'connection types' {lo te jorne} refers to the clear definition of synsets found in wordnet using lojban

Or, in laymans terms, a web front-end, a thesaurus and a dictionary.

Additionally, while the first two may only be of particular interest to lojbanists, I think the third may have even greater implications as an extension to Wordnet since it can do something Wordnet does not attempt to do: provide cardinality information. That is, lets say we want to use Wordnet to annotate a text unambiguously: that's totally possible. But if we wanted to then search that text (say, looking for query completions) theres no way to explicitly mark what lexeme falls in what argument number for a multi-argument phrase. Lojban, however, makes every bridi immediately accessible as a series of one or more triples for each sumti. Of course, this application is light-years off and might just be crazytalk. But that's why I would keep description of wordnet and description of lojban as two distinct objectives.

co'o mi'e korbi

btw, are you located in the united states by any chance? you might be interested in the North American Summer School in Logic, Language and Information.

Quoting Oren <get.oren@gmail.com>:

The jorne page on sourceforge [http://jorne.sourceforge.net/] doesn't
mention OWL or appear to have any source code... is there a newer
specification or codebase that I'm missing? The PEG parser?

As for the ideas proposed on the page, I still need to be sold. There seems
to be overlap with the W3C incubator projects for representing Wordnet in
RDF/OWL http://www.w3.org/TR/wordnet-rdf/, and quite frankly, lojban's
minimal and prescriptive vocabulary doesn't seem to offer much application
here.

Two separate overlapping W3C incubator projects seem to be more appropriate
for semantic querying, Common Web Language (semantic representation)
http://www.w3.org/2005/Incubator/cwl/XGR-cwl-20080331/ and Emotional Markup
Language http://www.w3.org/2005/Incubator/emotion/XGR-emotionml-20081120/.

Lojban, as a human language, can't offer what these robust proposals
describe-- that is, you can't really argue that lojban is any more
'readable' than these languages, nor believe that it would be briefer or
more thorough; but it may be fun to try and define the entire lojban
vocabulary using these technologies. Or maybe that's what you meant all
along?

<http://www.w3.org/TR/wordnet-rdf/>co'o mi'e korbi

On Thu, Apr 15, 2010 at 19:49, Brian Eubanks <brian@buildsoftware.com>wrote:

Hi Oren,

We corresponded last year about the Jorne (Lojban RDF) project I am trying
to get started.

The idea of using a Wordnet type approach is excellent. In fact, I would
love to see a LojWordNet in association with the Jorne OWL mapping.

Are you still interested in working on an OWL mapping for Lojban gismu? If
so, I would like you to join the Sourceforge Jorne project. The growing
amount of linked data makes this a great time to do this.

I am working with the PEG parser to import simple sentences into an RDF
triple store with the hope of converting between SPARQL and Lojban queries.
My Lojban is not even baby talk level yet, which is where I could use your
help too. I've been a lurker in the Lojban space but haven't spent time to
learn it.

Regards,
Brian Eubanks

Sent from my iPhone

On Apr 15, 2010, at 3:52 AM, Oren <get.oren@gmail.com> wrote:

I like the idea of categories (or... tags!), I think the wiki is the

place for it to happen, and I also think we shouldn't start from
scratch. The thesaurus on the wiki page already segregates all gismu
into hierarchical categories. We can make a page template that allows
people to add "lujvo requests" to a category. A sister project to
consider would be fleshing out that same ontology with the existing
specialized lujvo lists and the lujvo flat file.

I would also think that English/natlang glosses for the categories
should be optional while lojban section titles be mandatory and
default, for clarity.

Back to the original topic of finding a minimal wordlist for a
dictionary, I think the real forward-thinking approach would be to
find some sufficiently open project similar to EuroWordNet [a
multilingual WordNet], and then extracting a set number of unique
*syslinks* (word senses), so that when we sit down to define 'spring'
we don't have to remember jumping, metal coils and le printemps all by
our erring-human selves.

We could either use an arbitrary limit and go by frequency, and/or go
for all syslinks that contain an arbitrary number of constituent
languages. For example, only bother with 50% of all word senses that
appear in three or more languages.

co'o mi'e korbi

On Thu, Apr 15, 2010 at 15:13, Lindar <lindarthebard@yahoo.com> wrote:

My absolutely fantastic idea that donri/kribacr started and never
finished (or never even started, but definitely came up before I
thought of it [but it's still my idea]) is/was/will be to have groups
of people select topics, and then go through and come up with as many
words related to that topic as possible. I got this idea one day as I
was sticking masking tape to pretty much everything around my
apartment and writing the Lojban word for it in sharpie. I came across
the simple fact that jvs didn't have words for "pot", "kitchen",
"frying pan", etc., so I came up with words for them, and I think at
least "kitchen" (jupku'a) is up there. I tried this again with
computer terminology and it completely failed as nobody could agree
properly on things (like "window", on which I still harshly/
obnoxiously/rudely/insultingly disagree with xorxes).

Rather than having one person sit through some big gehorsenshitfesten
(parden my German) trying to pick out the most common concepts in the
universe, why don't we use the wiki idea and create "conversational
categories" under which we can place words (probably a lot of fu'ivla
and lujvo) relevant to the topic. This will generate a much larger and
relevant body of information, and it's a -much- less daunting task.
For example, I am a recording engineer, so I would be likely to start
a "recording technology" topic, and possibly contribute to the "music"
topic as I would be more likely than anybody else to need/use words
like "Hertz"/"kHz", "microphone", "nearfield monitors", "synthesizer",
"MIDI", "mixing console", "bass", "treble", and I would probably be
more qualified to determine what kind of terminology in Lojban is the
most suitable. I'd also be fairly interested in the "kitchen and
cooking" topic, and I think a great many a newbie would be very
interested in the "household objects" topic, which would probably
include a pointer to the "kitchen and cooking" topic and maybe even a
"bathroom and hygiene" topic. This way people find what interests them
and contribute to topics that they enjoy, which doesn't necessarily
give an accurate picture of common usage based on an average through
world cultures, but definitely gives a good sampling of words to use
in conversation for the types of conversation that people learning
Lojban would have. It works as a double edged sword (of handiness) in
that we have people that are going to enjoy working because they're
learning how to talk about things that interest them by contributing
(which means things are more likely to get added, being that it's fun
and not a chore) -AND- that we have quick 'topic reference'
dictionaries so you can just leave the list open and peak through to
make it easier to carry on conversations about what an arse your
government leader is without having to poke through a list for ten
minutes while the conversation has already passed because you wanted a
word for "idiot" and jvs only had "stupid" as a gloss word for
tolmencre. (Bad example, you get the picture.)

Perhaps we can quickly brainstorm a few major topics just to have
something up on a wiki?

household items
kitchen and cooking
bathroom and hygiene
sports and spectating
automotive and driving
computer ((hot topic, prone to arguments))
music
politics and law
school and education
work and the workplace
friends and family

The idea would be to have a big list of topics (and possibly
subtopics), and on the pages of each we have brief glosses with Lojban
words, with links to a page detailing the place structure, examples of
usage, actual usage example if available, and potentially a relevant
image (for those that learn by seeing and not reading).

Perhaps under "household items" is "garage", and on the page for that
it includes a little link for "see section: automotive and driving",
and perhaps even "garage" is also located under "automotive and
driving" or somesuch.

Neatonifty idea, right?

--
You received this message because you are subscribed to the Google Groups
"lojban" group.
To post to this group, send email to lojban@googlegroups.com.
To unsubscribe from this group, send email to

lojban+unsubscribe@googlegroups.com<lojban%2Bunsubscribe@googlegroups.com>

.
For more options, visit this group at
http://groups.google.com/group/lojban?hl=en.

--
You received this message because you are subscribed to the Google Groups
"lojban" group.
To post to this group, send email to lojban@googlegroups.com.
To unsubscribe from this group, send email to

lojban+unsubscribe@googlegroups.com<lojban%2Bunsubscribe@googlegroups.com>

.
For more options, visit this group at
http://groups.google.com/group/lojban?hl=en.

--
You received this message because you are subscribed to the Google Groups
"lojban" group.
To post to this group, send email to lojban@googlegroups.com.
To unsubscribe from this group, send email to

lojban+unsubscribe@googlegroups.com<lojban%2Bunsubscribe@googlegroups.com>

.
For more options, visit this group at
http://groups.google.com/group/lojban?hl=en.

--
You received this message because you are subscribed to the Google Groups "lojban" group.
To post to this group, send email to lojban@googlegroups.com.
To unsubscribe from this group, send email to lojban+unsubscribe@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/lojban?hl=en.

--
You received this message because you are subscribed to the Google Groups "lojban" group.
To post to this group, send email to lojban@googlegroups.com.
To unsubscribe from this group, send email to lojban+unsubscribe@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/lojban?hl=en.