Shawalphabet YahooGroup Archive Browser

> From: Ph. D. [mailto:phild@...]

> As someone else pointed out, natural language processing
> in English is not trivial. I would suggest that the look-up
> table simply have a flag on those words with more than one
> Shavian transcription. The program which does the trans-
> scription could flag those words in the output. Then a
> human could select the correct spelling. Thus 95% (or more)
> of the text would be automatic, with a minimal amount of
> human intervention .

If you want a quick solution, this is how it should be done. Otherwise you
have to delve into NLP which really isn't very easy.

To be honest, you could even program a database to do it, if all you'd be
doing is swapping one word for another except where there are two possible
pronunciations for each written word. I work all day with Access - perhaps I
could make something with that. Obviously you'd need a list of Roman words
in one field with matching Shavian (or other phonemic) transcriptions in an
adjacent field.

Hugh B

--
No virus found in this outgoing message.
Checked by AVG Anti-Virus.
Version: 7.0.300 / Virus Database: 265.8.5 - Release Date: 03/02/2005

Paul, Phil, Star, and others,

Can a converter simply flag words that have two pronunciations?

Yes, but that is not the ideal solution when you are converting serveral
pages of text.

It is one thing to have alternative pronunciations in a dictionary.
It is another to have alternatives in a look up table.

When you are looking up one word such as at www.m-w.com,
You do not even need to spell it correctly because the program will list
possible alternatives in a list.

On a simple converter, every word must be spelled correctly in the original or
the program flags the misspelling and does not convert.

If you have a list of words with two pronunciations, you can check them on a
converter at www.foreignword.com/dictionary/truespel/transpel.htm
This program does not flag problem words.
It just displays one without regard to its part of speech and flags words
(boldface)
when they are not in the dictionary (usually because they have been
misspelled).

Check out the URL above to get an idea of how a Shavian converter would work
and how it could be used to produce an unlimited amount of text in
Shavian transcription.

In a reverse dictionary, you can sound-spel two pronunciations
and it will list the same tradsepl word.

Several people have offered very elaborate solutions to the problem.
No one has come up with a simple solution.
If you have any ideas, please post them.



In a message dated 2/4/2005 10:28:55 AM Central Standard Time, phild@...
writes:
As someone else pointed out, natural language processing
in English is not trivial. I would suggest that the look-up
table simply have a flag on those words with more than one
Shavian transcription. The program which does the trans-
scription could flag those words in the output. Then a
human could select the correct spelling. Thus 95% (or more)
of the text would be automatic, with a minimal amount of
human intervention .


Steve T. Bett, Ph.D.
Austin, Texas
mailto:sbett@... 512-302-3014
www.foolswisdom.com/~sbett
retired professor, volunteer literacy instructor
moderator of Saundspel -the phonology forum
http://groups.yahoo.com/group/saundspel

Thanks, Philip. And I think that's a wonderful idea. It would save us
all a bunch of trouble, and would help us all become more fluent
readers of shavian if we have more to read.

--Star


> This thread started out by suggesting that large amounts
> of literature could be transcribed by using a computer
> program and a look-up table. I believe Star's question
> was how to handle these homonyms in a look-up table.
> ("one" vs. "won" is not a problem; only those which
> can have a different stress.)
>
> As someone else pointed out, natural language processing
> in English is not trivial. I would suggest that the look-up
> table simply have a flag on those words with more than one
> Shavian transcription. The program which does the trans-
> scription could flag those words in the output. Then a
> human could select the correct spelling. Thus 95% (or more)
> of the text would be automatic, with a minimal amount of
> human intervention .
>
> --Ph. D.
>
>
>
>


====http://www.livejournal.com/users/wodentoad

Numfar! Do the Dance of Joy!



__________________________________
Do you Yahoo!?
All your favorites on one personal page � Try My Yahoo!
http://my.yahoo.com

--- In shawalphabet@yahoogroups.com, "Ph. D." <phild@a...> wrote:
> paul vandenbrink said:
> > Star Raven wrote:
> >>
> >> what about the homonyms? PERfect vs. perFECT ect?
> >
> > It should be PERF-ect (Verb) vs per-FECT (Adjective)
> > which not only has a different syllable boundary but also a
> > distinct Shaw Spelling.
> >
> > pxf-ekt ----- pD-fekt
> >
> > There are going to be homonyms in Shavian, perhaps even more
> > than T.O In Shavian, won = one
> >
> > Overall, it still will be miles ahead of T.O.
>
>
> This thread started out by suggesting that large amounts
> of literature could be transcribed by using a computer
> program and a look-up table. I believe Star's question
> was how to handle these homonyms in a look-up table.
> ("one" vs. "won" is not a problem; only those which
> can have a different stress.)
>
> As someone else pointed out, natural language processing
> in English is not trivial. I would suggest that the look-up
> table simply have a flag on those words with more than one
> Shavian transcription. The program which does the trans-
> scription could flag those words in the output. Then a
> human could select the correct spelling. Thus 95% (or more)
> of the text would be automatic, with a minimal amount of
> human intervention .
>
> --Ph. D.

Exactly. My program Ghoti Filleter's look-up table had several fields
for Roman spellings which had homonyms. When the program reached any
word with stored homonyms, it would ask the user to select the
appropriate one. The queries from the program were welcome rather than
onerous -- much more fun than verifying OCR'd text, which I've been
doing in my spare time recently, and which can lead to hair loss.

Lionel Ghoti

Would you happen to be THE Lionel Ghoti who disappeared so long ago? If so,
welcome back!

As for converters, Mac users can get one here:
http://homepage.mac.com/poorant79/software/index.html

It uses the Mac's built-in text-to-speech capability to convert plain text
to phonemes and then to any other phonemic alphabet you give it (Shavian,
Unifon, Deseret, IPA, etc.). The old version doesn't seem to work at all,
and I haven't tried the new one because it's not compatible with my version
of the OS (it requires at least Mac OS 10.3).

Regards,
Joe
/JO


On 2/3/05 6:07 PM, "lghoti" <ghoti@...> wrote:
>
> Hello -- Lionel Ghoti here. Aren't there any Shavian transliterators
> about yet? I did build one a few years ago, and offered it up to
> anyone on the group who felt like trying it out, but it was largely
> dismissed because it required the user to install the Borland database
> engine (largely due to my lack of programming experience). It didn't
> have a dictionary of 60,000 words: as I remember, it started with a
> dictionary of 0 words and I started feeding it with every word it
> didn't recognise. It's surprising how quickly you build up a decent
> list of spelling-pairs.
>
> If anyone's interested, I will try and dig out the program, but for
> the time being the Borland database engine would be required.
>
> Florid regards,
>
> Lionel Ghoti
>
>
> --- In shawalphabet@yahoogroups.com, stbetta@a... wrote:
>> Carl,
>>
>> Why not build a converter that will auto convert plain text into
> Shavian.
>> That would certainly be a quicker and more accurate way to achieve
> your goal.
>>
>> The first step in building a converter is to provide the Shavian
> equivalent
>> in ASCII of 60,000 words. The hard transcriptions should be posted
> on this
>> discussion group.
>>
>> --Steve
>
>
>
>
>
>
>
> Yahoo! Groups Links
>
>
>
>
>
>
>

Actually, I've built a perl script that runs on my server and converts
webpages, plain text, or a file no matter what OS you are using, as
well. It all uses the MySQL database and I've set it so it keeps a
record of each word consisting of phonetic, roman, ipa, mac, frequency,
modified date, person that modified, and other stuff.
Just go to my website look for the 'online text converter'. type in
the password which is 'www'. Any questions? email me.

ù¶ùÆù§
- Jake Hickenlooper
Programmer
Splinter Works Software
homepage.mac.com/poorant79/software

On Feb 4, 2005, at 1:13 PM, Joe wrote:

> Would you happen to be THE Lionel Ghoti who disappeared so long ago?
> If so, welcome back!
>
> As for converters, Mac users can get one here:
> http://homepage.mac.com/poorant79/software/index.html
>
> It uses the Mac's built-in text-to-speech capability to convert plain
> text to phonemes and then to any other phonemic alphabet you give it
> (Shavian, Unifon, Deseret, IPA, etc.). The old version doesn't seem
> to work at all, and I haven't tried the new one because it's not
> compatible with my version of the OS (it requires at least Mac OS
> 10.3).
>
> Regards,
> Joe
> /JO
>
>

Hi Steve

Where are the real programmers? <rhetorical>
Since you asked.

T.O. -> Shavian
You are not really talking about a simple filter or matchup
program here. Although, I suppose that is one brute force approach.
You need a fairly sophisticated transliteration Algorithm,
with a file of exceptions. It also needs some ability to
handle some situations through context.
It should produce clean output.

As a second step you need to run it through a Spell-Checker
which should flag any anomalies.
There are lots of good spell checkers for T.O.
I would think to convert one to SpellCheck Shavian English
would be a fairly easy project.

So you need two programs. Anyone else care to lay out the Algorithms
in some context free language like Pascal.

Regards, Paul V.

P.S. I actually wouldn't consider this first program to involve much
Natural Language Processing.


_______________attached_______________________________

--- In shawalphabet@yahoogroups.com, stbetta@a... wrote:
> Paul, Phil, Star, and others,
>
> Can a converter simply flag words that have two pronunciations?
>
> Yes, but that is not the ideal solution when you are converting
serveral
> pages of text.
>
> It is one thing to have alternative pronunciations in a dictionary.
> It is another to have alternatives in a look up table.
>
> When you are looking up one word such as at www.m-w.com,
> You do not even need to spell it correctly because the program will
list
> possible alternatives in a list.
>
> On a simple converter, every word must be spelled correctly in the
original or
> the program flags the misspelling and does not convert.
>
> If you have a list of words with two pronunciations, you can check
them on a
> converter at www.foreignword.com/dictionary/truespel/transpel.htm
> This program does not flag problem words.
> It just displays one without regard to its part of speech and flags
words
> (boldface)
> when they are not in the dictionary (usually because they have been
> misspelled).
>
> Check out the URL above to get an idea of how a Shavian converter
would work
> and how it could be used to produce an unlimited amount of text in
> Shavian transcription.
>
> In a reverse dictionary, you can sound-spel two pronunciations
> and it will list the same tradsepl word.
>
> Several people have offered very elaborate solutions to the problem.
> No one has come up with a simple solution.
> If you have any ideas, please post them.
>
>
>
> In a message dated 2/4/2005 10:28:55 AM Central Standard Time,
phild@a...
> writes:
> As someone else pointed out, natural language processing
> in English is not trivial. I would suggest that the look-up
> table simply have a flag on those words with more than one
> Shavian transcription. The program which does the trans-
> scription could flag those words in the output. Then a
> human could select the correct spelling. Thus 95% (or more)
> of the text would be automatic, with a minimal amount of
> human intervention .
>
>
> Steve T. Bett, Ph.D.
> Austin, Texas
> mailto:sbett@l... 512-302-3014
> www.foolswisdom.com/~sbett
> retired professor, volunteer literacy instructor
> moderator of Saundspel -the phonology forum
> http://groups.yahoo.com/group/saundspel

--- In shawalphabet@yahoogroups.com, Joe <wurdbendur@g...> wrote:
> Would you happen to be THE Lionel Ghoti who disappeared so long ago?
If so,
> welcome back!

I'm afraid so, and ta very much!

I have just uploaded my old Ghoti Filleter program (the
transliterator) to my web site. You can get it at:

www.saytheword.org.uk/shavian/fillet.html

...if you think that it's worth the hassle of installing the Borland
Database Engine too.

A couple of nights ago, I dug up the whole of my old Shavian site and
uploaded that too. It's at:

www.saytheword.org.uk/shavian

Perhaps the brute force answer is the easiest: word=word, but if it
matches one on the list or has no match, then bring up a dialogue box
to that effect. Going back the other way, Shaw to TO, I can see the
complicated algorhythm, but not from TO to shaw. There are fewer
spellings.

Also, a spell checker should of course be run before using such a
program, but on the other hand, a shavian spell checker defeats the
purpose of shavian, doesn't it? I mean, hasn't that been the subject of
so many of our previous debates?

--Star

--- paul vandenbrink <pvandenbrink@...> wrote:

>
> Hi Steve
>
> Where are the real programmers? <rhetorical>
> Since you asked.
>
> T.O. -> Shavian
> You are not really talking about a simple filter or matchup
> program here. Although, I suppose that is one brute force approach.
> You need a fairly sophisticated transliteration Algorithm,
> with a file of exceptions. It also needs some ability to
> handle some situations through context.
> It should produce clean output.
>
> As a second step you need to run it through a Spell-Checker
> which should flag any anomalies.
> There are lots of good spell checkers for T.O.
> I would think to convert one to SpellCheck Shavian English
> would be a fairly easy project.
>
> So you need two programs. Anyone else care to lay out the Algorithms
> in some context free language like Pascal.
>
> Regards, Paul V.
>
> P.S. I actually wouldn't consider this first program to involve much
> Natural Language Processing.
>
>
> _______________attached_______________________________
>
> --- In shawalphabet@yahoogroups.com, stbetta@a... wrote:
> > Paul, Phil, Star, and others,
> >
> > Can a converter simply flag words that have two pronunciations?
> >
> > Yes, but that is not the ideal solution when you are converting
> serveral
> > pages of text.
> >
> > It is one thing to have alternative pronunciations in a dictionary.
> > It is another to have alternatives in a look up table.
> >
> > When you are looking up one word such as at www.m-w.com,
> > You do not even need to spell it correctly because the program will
>
> list
> > possible alternatives in a list.
> >
> > On a simple converter, every word must be spelled correctly in the
> original or
> > the program flags the misspelling and does not convert.
> >
> > If you have a list of words with two pronunciations, you can check
> them on a
> > converter at www.foreignword.com/dictionary/truespel/transpel.htm
> > This program does not flag problem words.
> > It just displays one without regard to its part of speech and flags
>
> words
> > (boldface)
> > when they are not in the dictionary (usually because they have been
>
> > misspelled).
> >
> > Check out the URL above to get an idea of how a Shavian converter
> would work
> > and how it could be used to produce an unlimited amount of text in
> > Shavian transcription.
> >
> > In a reverse dictionary, you can sound-spel two pronunciations
> > and it will list the same tradsepl word.
> >
> > Several people have offered very elaborate solutions to the
> problem.
> > No one has come up with a simple solution.
> > If you have any ideas, please post them.
> >
> >
> >
> > In a message dated 2/4/2005 10:28:55 AM Central Standard Time,
> phild@a...
> > writes:
> > As someone else pointed out, natural language processing
> > in English is not trivial. I would suggest that the look-up
> > table simply have a flag on those words with more than one
> > Shavian transcription. The program which does the trans-
> > scription could flag those words in the output. Then a
> > human could select the correct spelling. Thus 95% (or more)
> > of the text would be automatic, with a minimal amount of
> > human intervention .
> >
> >
> > Steve T. Bett, Ph.D.
> > Austin, Texas
> > mailto:sbett@l... 512-302-3014
> > www.foolswisdom.com/~sbett
> > retired professor, volunteer literacy instructor
> > moderator of Saundspel -the phonology forum
> > http://groups.yahoo.com/group/saundspel
>
>
>
>


====http://www.livejournal.com/users/wodentoad

Numfar! Do the Dance of Joy!



__________________________________
Do you Yahoo!?
The all-new My Yahoo! - Get yours free!
http://my.yahoo.com

We had given up the search and declared you legally dead. There was a
memorial service and everything! And here it is you were here the whole
time, unless we've stepped into a soap opera and you are the evil twin
born three years later who just woke up from the coma.

Welcome back! We missed you!
--Star

--- lghoti <ghoti@...> wrote:

>
> --- In shawalphabet@yahoogroups.com, Joe <wurdbendur@g...> wrote:
> > Would you happen to be THE Lionel Ghoti who disappeared so long
> ago?
> If so,
> > welcome back!
>
> I'm afraid so, and ta very much!
>
> I have just uploaded my old Ghoti Filleter program (the
> transliterator) to my web site. You can get it at:
>
> www.saytheword.org.uk/shavian/fillet.html
>
> ...if you think that it's worth the hassle of installing the Borland
> Database Engine too.
>
> A couple of nights ago, I dug up the whole of my old Shavian site and
> uploaded that too. It's at:
>
> www.saytheword.org.uk/shavian
>
>
>
>


====http://www.livejournal.com/users/wodentoad

Numfar! Do the Dance of Joy!



__________________________________
Do you Yahoo!?
Meet the all-new My Yahoo! - Try it today!
http://my.yahoo.com