Lojban

[David Ireland]

Hi,

I have been studying Lojban for human-computer interaction research and by coincidence I stumbled on Ben's presentations on the topic.  I was wondering if the OpenCog group still considers this an area of interest and if any insights have been discovered. 

Thanks. 

David

[Ben Goertzel]

We hadn't been doing much with Lojban for a while but actually right
now we are experimenting with training a seq2seq network for English
=> Lojban translation, along with one for English => Lojbanic-Atomese
... will let you know how it works!

> --
> You received this message because you are subscribed to the Google Groups "opencog" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to opencog+u...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/opencog/CAF5Y0%2BF1GUu_NQpasbBqjJsbGCMgwYkP1BXiCr-eKzNxLX-Ubg%40mail.gmail.com.



--
Ben Goertzel, PhD
http://goertzel.org

“The only people for me are the mad ones, the ones who are mad to
live, mad to talk, mad to be saved, desirous of everything at the same
time, the ones who never yawn or say a commonplace thing, but burn,
burn, burn like fabulous yellow roman candles exploding like spiders
across the stars.” -- Jack Kerouac

[h0p3]

Bump. I would be interested to hear anyone's conjectures, feelings, anecdotes, or poetry on this matter too. Anything, please.

On Monday, March 16, 2020 at 2:26:22 PM UTC-4, Ben Goertzel wrote:

We hadn't been doing much with Lojban for a while but actually right
now we are experimenting with training a seq2seq network for English
=> Lojban translation, along with one for English => Lojbanic-Atomese
... will let you know how it works!


On Mon, Mar 16, 2020 at 1:51 AM David Ireland <djire...@gmail.com> wrote:
>
>
> Hi,
>
> I have been studying Lojban for human-computer interaction research and by coincidence I stumbled on Ben's presentations on the topic.  I was wondering if the OpenCog group still considers this an area of interest and if any insights have been discovered.
>
> Thanks.
> David
>
> --
> You received this message because you are subscribed to the Google Groups "opencog" group.

> To unsubscribe from this group and stop receiving emails from it, send an email to ope...@googlegroups.com.

[Matt Chapman]

There once was a linguistic treat,

It wrapped up ambiguity neat,

But when AI coders called

The problems it solved

Were found to be quite obsolete.

If you prefer prose: Lojban once may have helped address some of the gaps in machine understanding of human speech, but current speech-to-text and text disambiguation techniques are so good, I doubt it's worth the trouble anymore. If it were, I suspect someone would have done it by now. Just my non-profession 2c. I kinda hope I'm wrong; Lojban a fun idea to play around with.

All the Best,

Matt

--
Please interpret brevity as me valuing your time, and not as any negative intention.

To unsubscribe from this group and stop receiving emails from it, send an email to opencog+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/opencog/a394aefc-ba6b-4b6e-9b36-43b6d9e7bd35o%40googlegroups.com.

[Ben Goertzel]

yeah kind of true...

I mean -- NNs don't yet do everything that Lojban does. However, we
have chosen to put more energy into symbolic extraction of grammatical
patterns from transformer NNs rather than on Lojban analytics...

https://arxiv.org/abs/2005.12533

We have also btw tried training a seq2seq NN on an English/Lojban
parallel corpus (of about 15K sentences), to translate English to
Lojban. But given the size of the corpus and the complexity of Lojan,
this was too hard for the standard-ish NN techniques we tried. We
may try again later with fancier neural-symbolic methods....

Playing w/ Lojban is still cool but Matt is correct that the odds of
it being a really critical part of a cognitive NLP pipeline are
seeming lower than they did before...

> To view this discussion on the web visit https://groups.google.com/d/msgid/opencog/CAPE4pjAiDyYtRXGzR8U44Pr4tJdf1LiAJYt-MAe_b9OS79kEJA%40mail.gmail.com.