There are 4 messages in this issue.
Topics in this digest:
1a. Re: [LCS Members] Conlang Card Exchange
From: David Peterson
2a. Re: Is there a word for this?
From: Jeff Sheets
2b. Re: Is there a word for this?
From: Gary Shannon
3a. Re: NLP class (was RE: Is there a word for this?)
From: Jeff Sheets
Messages
________________________________________________________________________
1a. Re: [LCS Members] Conlang Card Exchange
Posted by: "David Peterson" [email protected]
Date: Sun Jan 27, 2013 1:49 pm ((PST))
There's also http://pics.conlang.org/
David Peterson
LCS President
[email protected]
www.conlang.org
On Jan 27, 2013, at 1:30 PM, Alex Fink <[email protected]> wrote:
> On 27 January 2013 16:25, Jeffrey Brown <[email protected]> wrote:
>> It would be nice if there were single place to which all the cards could be
>> uploaded. Could the LCS create a space where people could upload them?
>
> There are galleries for 2009 and 2010 at
> http://exchange.conlang.org/?page_id=53
> http://exchange.conlang.org/?page_id=68
> so presumably alongside them would be the natural place.
>
> Alex
Messages in this topic (4)
________________________________________________________________________
________________________________________________________________________
2a. Re: Is there a word for this?
Posted by: "Jeff Sheets" [email protected]
Date: Sun Jan 27, 2013 1:59 pm ((PST))
I have never seen Link Grammars before. I've generally approached natural
and constructed languages from the linguistics side of things. It's
definitely an interesting way of going about parsing, and it definitely has
a more computer science-y feel to it than I'm used to (in languages). Seems
to work quite well, but I'm inherently wary of anything which doesn't
explicitly state the rules of grammar separately from the lexicon. I'm
biased, I suppose, but I'd prefer the grammar stand separate for my own
conlanging. My reasoning is simple: linguists are fairly certain that
grammar and lexicon are separate in the brain. Also, if the grammar is
separate, the number of grammatical rules will be minimized, leaving only
context clues in the lexicon.
On Sat, Jan 26, 2013 at 3:37 PM, Gary Shannon <[email protected]> wrote:
> Interesting links. Thank you. I ordered a used copy of that book
> (Grammar as Science).
>
> Are you familiar with Link Grammars? I wrote a parser based on a link
> grammar several years ago. Here's a good introduction:
>
> http://www.cs.cmu.edu/afs/cs.cmu.edu/project/link/pub/www/papers/ps/LG-IWPT93.pdf
>
> --gary
>
> On Sat, Jan 26, 2013 at 12:24 PM, Jeff Sheets <[email protected]>
> wrote:
> > Interesting choices for the names of categories/tags. I'm a bit too used
> to
> > Phrase Structure Rules to have figured that say RB means adverb, or that
> JJ
> > means adjective. But how you name your tags isn't important so long as
> the
> > system is consistent.
> >
> > Some other links for you. The first is a link to a Coursera online class
> on
> > natural language processing, which I think most closely matches your
> > interest. Might give you an idea of how academics are approaching the
> > problem today. NLP is without doubt an extremely difficult task for
> > computers.
> >
> > https://www.coursera.org/course/nlangp
> >
> > This next link is a crash course to Phrase Structure Rules. It gives an
> > idea of how I approach syntax, and also shows just how ambiguous natural
> > languages can be, and thus how difficult it can be to translate them
> > automatically with computers.
> >
> >
> http://people.umass.edu/afarudi/Phrase%20Structure%20Rules-Kyle%20Johnson.pdf
> >
> > If you have some money you can afford to spend, less than $50, I highly
> > recommend the following textbook:
> >
> > http://www.amazon.com/Grammar-as-Science-Richard-Larson/dp/026251303X
> >
> > It's a thorough read on how syntax is approached on its own, including
> how
> > to deal with features of words and phrases in a grammar (such as verbs of
> > motion as a feature, locative, dative, accusative in noun phrases, etc.)
> > and how to deal with movement of constituents in a sentence. Don't let
> its
> > cartoony diagrams fool you, either. It gets into the really complex
> > syntactic structures just as much as it does simplistic ones. Available
> in
> > paperback and a kindle version, and used versions at quite a low price.
> >
> > Now, I wish there was a nice, concise, definitive, and standardized list
> of
> > English phrase structure rules... but as far as I know, there is no such
> > compilation. Primarily because linguists don't necessarily agree on how
> to
> > deal with the really complex issues of English syntax.
> >
> >
> > On Fri, Jan 25, 2013 at 2:34 PM, Gary Shannon <[email protected]> wrote:
> >
> >> This is very interesting. Thanks for posting that link. I'll have to
> >> spend some more time looking into that.
> >>
> >> As for my formal grammar, I'm using a parenthetical notation that
> >> allows me to tag/parse a sentence, and then extract both the
> >> production rules and the lexicon directly from a collection of tagged
> >> sentences. Something like this:
> >>
> >> Sentence: Bravely the wounded soldier struggled on.
> >>
> >> Tagged/parsed:
> >>
> >> SNT(RB(Bravely) SNT(ND(DT(the) NJ(JJ(wounded) NN(soldier)))
> >> VBP(VB(struggled) RBP(on))))
> >>
> >> Words removed:
> >>
> >> SNT(RB SNT(ND(DT NJ(JJ NN)) VBP(VB RBP)))
> >>
> >> Rules extracted:
> >>
> >> ND(DT NJ)
> >> NJ(JJ NN)
> >> SNT(ND VBP)
> >> SNT(RB SNT)
> >> VBP(VB RBP)
> >>
> >> Lexicon extracted:
> >>
> >> DT(the)
> >> JJ(wounded)
> >> NN(soldier)
> >> RB(bravely)
> >> RBP(on)
> >> VB(struggled)
> >>
> >> Sorted by word:
> >>
> >> bravely RB
> >> on RBP
> >> soldier NN
> >> struggled VB
> >> the DT
> >> wounded JJ
> >>
> >> And, of course, when the same word shows up with different parts of
> >> speech, all those alternatives would appear in the lexicon. My tags
> >> are borrowed from the Brown Corpus tag set, with several modifications
> >> to fit my specific application. (
> >> http://www.comp.leeds.ac.uk/ccalas/tagsets/brown.html )
> >>
> >> --gary
> >>
> >>
> >> On Fri, Jan 25, 2013 at 11:14 AM, Jeff Sheets <[email protected]>
> >> wrote:
> >> > I'm surprised nobody has mentioned "constituent" yet.
> >> >
> >> > http://en.wikipedia.org/wiki/Constituent_(linguistics)
> >> >
> >> > The set of all constituents then, is all phrases and single words in
> the
> >> > language. As the context becomes more known, the set of potential
> >> > constituents is reduced to a subset. Note, however:
> >> >
> >> > The box ___ down.
> >> >
> >> > does not offer a constituent in the technical meaning of the word,
> but I
> >> > still think that that is as close as you are likely to get. About all
> >> that
> >> > fits there is the subset of verbs that function and fit with the
> adverb
> >> > down, though I'm not putting too much thought into that. However, if
> you
> >> > start with:
> >> >
> >> > The box ___.
> >> >
> >> > You know that any of the following will fit:
> >> >
> >> > came in the mail today.
> >> > fell down.
> >> > is really rather large and ungainly to transport across the distance
> of
> >> 16
> >> > miles by foot both uphill an downhill.
> >> >
> >> > The context allows a much broader set of constituents. However, below
> >> > constituents are just the parts of speech. The reason why verbs like
> >> > "spoke" don't fit in the first sentence is that they lack some
> features.
> >> > Some verbs will be transitive, and thus require a direct object. Some
> >> verbs
> >> > are ditransitive and require both a direct and indirect object. In
> this
> >> > case, the feature is more that the verbs must describe movement.
> >> >
> >> > The box slides down.
> >> > The box fell down.
> >> > The box ran down.
> >> > The box jumped down.
> >> > The box teleported down.
> >> > * The box spoke down.
> >> > * The box thought down.
> >> > * The box befriended down.
> >> > x The box ascended down.
> >> >
> >> > That last sentence feels grammatical to me, though obviously it makes
> no
> >> > sense, but the three marked with * are very much syntactically
> incorrect
> >> > for me. The key thing is that slides, fell, ran, jumped, teleported,
> and
> >> > ascended are all verbs which have the feature of describing motion.
> >> >
> >> > One question I have is, how are you defining/describing the grammar
> and
> >> > lexicon of your language? Are you using a formal grammar notation like
> >> the
> >> > following?
> >> >
> >> > S -> NP VP
> >> > NP -> (Det) N
> >> > NP -> NP PP
> >> > NP -> Adj NP
> >> > PP -> Prep NP
> >> > VP -> V
> >> > VP -> VP Adv
> >> > etc.
> >> >
> >> > You may want to identify that adverbs like "down" must modify a verb
> with
> >> > the feature of "motion", and then for every motion verb, add that
> feature
> >> > to a list of features. Other features you should probably have is the
> >> > transitivity of the verb.
> >> >
> >> >
> >> > On Tue, Jan 22, 2013 at 8:57 PM, Ralph DeCarli <
> [email protected]
> >> >wrote:
> >> >
> >> >> On Mon, 21 Jan 2013 20:20:30 -0600
> >> >> George Corley <[email protected]> wrote:
> >> >>
> >> >> >
> >> >> > Do you consider the instrument or other prepositional elements
> >> >> > inherently part of the verb?
> >> >> >
> >> >> In this specific case I consider the fork to be a data element of
> >> >> the 'eating' predicate, if that makes any sense. I tend to think of
> >> >> the language in data modeling terms.
> >> >>
> >> >> A given prepositional phrase could modify the subject, the object or
> >> >> the predicate, but it can't modify the entire sentence. I think this
> >> >> actually stems from my general fear of 'global variables'.
> >> >>
> >> >> In other words, I'm really still more of a programmer and a "data
> >> >> bigot" than a linguist, so my conlang (or con-patois, more
> >> >> accurately) is going to reflect my learned habits.
> >> >>
> >> >> Ralph
> >> >> --
> >> >>
> >> >> Have you heard of the new post-neo-modern art style?
> >> >> They haven't decided what it looks like yet.
> >> >>
> >>
>
Messages in this topic (23)
________________________________________________________________________
2b. Re: Is there a word for this?
Posted by: "Gary Shannon" [email protected]
Date: Sun Jan 27, 2013 2:35 pm ((PST))
I agree for conlang purposes. Link grammars are kind of fun to play
with, though. They do have some problems, especially using
conjunctions. A sentence like "He stole the tarts and ran away." can't
be parsed with their link grammar because "ran" doesn't have a subject
that can be linked without crossing lines. So they have to make a
special "cheat" pass to resolve those kinds of problems. That makes it
less than elegant as far as I'm concerned.
--gary
On Sun, Jan 27, 2013 at 1:59 PM, Jeff Sheets <[email protected]> wrote:
> I have never seen Link Grammars before. I've generally approached natural
> and constructed languages from the linguistics side of things. It's
> definitely an interesting way of going about parsing, and it definitely has
> a more computer science-y feel to it than I'm used to (in languages). Seems
> to work quite well, but I'm inherently wary of anything which doesn't
> explicitly state the rules of grammar separately from the lexicon. I'm
> biased, I suppose, but I'd prefer the grammar stand separate for my own
> conlanging. My reasoning is simple: linguists are fairly certain that
> grammar and lexicon are separate in the brain. Also, if the grammar is
> separate, the number of grammatical rules will be minimized, leaving only
> context clues in the lexicon.
Messages in this topic (23)
________________________________________________________________________
________________________________________________________________________
3a. Re: NLP class (was RE: Is there a word for this?)
Posted by: "Jeff Sheets" [email protected]
Date: Sun Jan 27, 2013 2:24 pm ((PST))
I don't see any specific reason not to take both. The Columbia U course
(the one I suggested) has a set start date; my main reason for suggesting
it. The Stanford course has yet to have its start date announced. Quite
likely one would be able to participate in both. They both recommend having
a decent grasp of certain mathematics fields, namely probability, calculus,
algorithms and discrete math (though they don't explicitly state that last
one). The Stanford course also suggests programming experience, and it has
programming exercises to complete. Neither course requires a textbook (par
for the Coursera course), but the Stanford one recommends some books. My
inherent cynicism flares at this point, in that the books they recommend
are written by the instructors. I have no idea if my cynicism is justified
though.
But again, assuming the two don't overlap in time, which they probably
won't, why not take both? You may end up covering the same material twice
once in a while, but multiple viewpoints on the same topic have often
helped me to better understand it.
If I were to be nailed down on answering "which one seems better and why?",
I'd have to say the Columbia U course. First, the syllabus seems to
indicate that it'll be covering more content. Second, it is also self
contained according to the description given, which is always a plus for
me. Just one of the Stanford course's recommended books is ~$115 on Amazon.
The other is ~$51.
That said, what's better for me, is not necessarily better for you. And...
I have by now posted more in the past couple days than I have in the last
year entirely, I think. I tend to just lurk. :)
On Sat, Jan 26, 2013 at 4:22 PM, Nikolay Ivankov <[email protected]>wrote:
> On Sat, Jan 26, 2013 at 10:07 PM, Mathieu Roy <[email protected]
> >wrote:
>
> > Jeff: <<Some other links for you. The first is a link to a Coursera
> online
> > class on natural language processing, which I think most closely matches
> > your interest. Might give you an idea of how academics are approaching
> the
> > problem today. NLP is without doubt an extremely difficult task for
> > computers.
> >
> > https://www.coursera.org/course/nlangp>>
> >
> > I was hesitating between this course and a similar one
> > (https://www.coursera.org/course/nlp). In your opinion, which one seems
> > better and why?
> >
> > -Mathieu
> >
>
> Seconding this question.
>
Messages in this topic (3)
------------------------------------------------------------------------
Yahoo! Groups Links
<*> To visit your group on the web, go to:
http://groups.yahoo.com/group/conlang/
<*> Your email settings:
Digest Email | Traditional
<*> To change settings online go to:
http://groups.yahoo.com/group/conlang/join
(Yahoo! ID required)
<*> To change settings via email:
[email protected]
[email protected]
<*> To unsubscribe from this group, send an email to:
[email protected]
<*> Your use of Yahoo! Groups is subject to:
http://docs.yahoo.com/info/terms/
------------------------------------------------------------------------
