From joen_r at hotmail.com Mon Mar 4 09:10:31 2002
From: joen_r at hotmail.com (joen_r at hotmail.com)
Date: Mon, 04 Mar 2002 11:10:31 -0600
Subject: ASSET & BACKGROUND CHECKS!..WE CHECK PEOPLE OUT!!_
Message-ID: <00003bf12eda$000037f3$0000437c@>
FULL NATIONAL Asset & Background Searches
Just call us Toll Free at (888) 729-8976 and PROTECT YOURSELF
AmericaFind Inc.
24 hour OR LESS Turn Around Time.
Just call us Toll Free at (888) 729-8976 and PROTECT YOURSELF
What do you Really Know about your Employee?
What do you Really Know about your Lover?
What do you Really Know about your Baby Sitter?
What do you Really Know about your Business Associate?
You NEED to protect yourself! You NEED to know the TRUTH!
WE CHECK PEOPLE OUT FOR YOU!!!
Just call us Toll Free at (888) 729-8976 and PROTECT YOURSELF
In 24 hours OR LESS we can tell you everything to allow you to make an informed decision!!
*All Real Estate owned in past 15 years, *All Corporations & DBA's, *STATE AND FEDERAL Civil Judgments,
*Bankruptcies and Liens covering past 15 years.!! *All State and Federal Misdemeanor
and Felony Criminal Convictions.!! *Much more!
Do NOT go uninformed. Let AmericaFind Inc. Tell you the TRUTH!
REMEMBER..WE CHECK PEOPLE OUT FOR YOU!!
Just call us Toll Free at (888) 729-8976 and PROTECT YOURSELF.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
We Find Missing People for YOU.....OR it's FREE!!
As seen on OPRAH ....Satisfaction GUARANTEED!!
America Find your LOST LOVE from HIGH SCHOOL
America Find the Person who SKIPPED TOWN owing you MONEY
America Find the FRIEND you served with in COMBAT
America Find that DEADBEAT PARENT
LET AMERICAFIND that MISSING PERSON for you!
Results in 72 hours or Less!
CALL TOLL FREE!!! 877-269-3892
Satisfaction GUARANTEED in WRITING on our Web Site!!
MANY Thousands Founds...MANY Thousands still lost.
Let AMERICAFIND Help.
CALL TOLL FREE!!! 877-269-3892
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
AS A COURTESY, IF YOU WOULD LIKE
To be removed from further mailings send an email from the
address you want removed to mailto:crumbcake at myrealbox.com?subject=Remove
FULL NATIONAL Asset & Background Searches
Just call us Toll Free at (888) 729-8976 and PROTECT YOURSELF
AmericaFind Inc.
24 hour OR LESS Turn Around Time.
From stolcke at speech.sri.com Wed Mar 6 18:38:48 2002
From: stolcke at speech.sri.com (Andreas Stolcke)
Date: Wed, 06 Mar 2002 18:38:48 PST
Subject: srilm-htk interface compiler problems
In-Reply-To: Your message of Fri, 01 Mar 2002 10:41:04 +0800.
<007801c1c0ca$88abece0$b9176d8c@aiiaricer>
Message-ID: <200203070238.SAA27783@zap.speech.sri.com>
The HTK-SRILM interface was last tested at the Johns Hopkins workshop
in 1997, and I no longer have access to it. That fact that you want to
do this in Windows isn't going to make things easier either.
Also, the interface is not to HTK per se, but to the lattice rescoring
tools that Entropic developed specifically for the JHU workshop (first in
1995). I'm not even sure they are publically available.
So I'm afraid I can't help here.
If anyone has more information on the HTK lattice tools I'd like to
hear about it.
--Andreas
In message <007801c1c0ca$88abece0$b9176d8c at aiiaricer>you wrote:
> This is a multi-part message in MIME format.
>
> ------=_NextPart_000_0075_01C1C10D.96909D40
> Content-Type: text/plain;
> charset="big5"
> Content-Transfer-Encoding: quoted-printable
>
> hi:
> i'm a new user of SRILM
> in SRILM/HTK
> i want to compiler=20
> but have fllowing error messages:
>
> "gcc: D:hk3/HTKib/HTKLib..a : No such file or directory
> gcc:......
> gcc:..
> ....."
>
> i did have those files in
>
> $(HTK_LATTICE_DIR)/MyLangModel.h \
> LIBRARY =3D $(OBJDIR)/libHTKlattice.a
> $(HTK_LATTICE_DIR)/JLib.$(HTK_CPU_TYPE).a \
> $(HTK_LIB_DIR)/HTKLib.$(HTK_CPU_TYPE).a \
> $(HTK_LIB_DIR)/HTKLibHE.$(HTK_CPU_TYPE).a \
>
> so where can i get those files
> and
> i use HTK3 on Widoes 2000
> What environment i must set
>
>
> thank u very much!!!
>
>
> ------=_NextPart_000_0075_01C1C10D.96909D40
> Content-Type: text/html;
> charset="big5"
> Content-Transfer-Encoding: quoted-printable
>
>
>
>
>
>
>
>
> hi:
> i'm a new user of SRILM
> in SRILM/HTK
> i want to compiler
> but have fllowing error messages:
>
> "gcc: D:hk3/HTKib/HTKLib..a : No such file or directory
> gcc:......
> gcc:..
> ....."
>
> i did have those files in
>
> $(HTK_LATTICE_DIR)/MyLangModel.h \
> LIBRARY =3D $(OBJDIR)/libHTKlattice.a
> $(HTK_LATTICE_DIR)/JLib.$(HTK_CPU_TYPE).a=20
> \
$(HTK_LIB_DIR)/HTKLib.$(HTK_CPU_TYPE).a=20
> \
$(HTK_LIB_DIR)/HTKLibHE.$(HTK_CPU_TYPE).a \
>
> so where can i get those files
> and
> i use HTK3 on Widoes 2000
> What environment i must set
>
>
> thank u very much!!!
>
>
> ------=_NextPart_000_0075_01C1C10D.96909D40--
>
From ge204 at eng.cam.ac.uk Thu Mar 7 02:23:04 2002
From: ge204 at eng.cam.ac.uk (Gunnar Evermann)
Date: 07 Mar 2002 10:23:04 +0000
Subject: srilm-htk interface compiler problems
In-Reply-To: <200203070238.SAA27783@zap.speech.sri.com>
References: <200203070238.SAA27783@zap.speech.sri.com>
Message-ID:
Andreas Stolcke writes:
> The HTK-SRILM interface was last tested at the Johns Hopkins workshop
> in 1997, and I no longer have access to it. That fact that you want to
> do this in Windows isn't going to make things easier either.
> Also, the interface is not to HTK per se, but to the lattice rescoring
> tools that Entropic developed specifically for the JHU workshop (first in
> 1995). I'm not even sure they are publically available.
As far as I know they were never really publicly available, i.e. these
days they are only used at CUED and possibly at JHU.
> So I'm afraid I can't help here.
> If anyone has more information on the HTK lattice tools I'd like to
> hear about it.
I have implemented a replacement tool (HLRescore) for most of the
lattice toolkit functionality. This tool is going to be part of
HTK3.2, which will be our next major release. It can read normal ARPA
LMs and use them to rescore HTK lattices. If you need access to the
fancier features of the SRILM then somebody would need to re-implement
an interface, this should be relatively easy, though.
Andreas, I'd be happy to discuss the implementation of such an
interface and/or to contribute some code for that to SRILM. However, I
won't have time for this before the end of April -- I'm sure you'll
understand :-)
Gunnar
From stolcke at speech.sri.com Thu Mar 7 08:34:28 2002
From: stolcke at speech.sri.com (Andreas Stolcke)
Date: Thu, 07 Mar 2002 08:34:28 PST
Subject: srilm-htk interface compiler problems
In-Reply-To: Your message of 07 Mar 2002 10:23:04 +0000.
Message-ID: <200203071634.IAA02406@huge>
>
> I have implemented a replacement tool (HLRescore) for most of the
> lattice toolkit functionality. This tool is going to be part of
> HTK3.2, which will be our next major release. It can read normal ARPA
> LMs and use them to rescore HTK lattices. If you need access to the
> fancier features of the SRILM then somebody would need to re-implement
> an interface, this should be relatively easy, though.
That's great!
> Andreas, I'd be happy to discuss the implementation of such an
> interface and/or to contribute some code for that to SRILM. However, I
> won't have time for this before the end of April -- I'm sure you'll
> understand :-)
And for the same reasons I wouldn't have any time either. After the
the RT workshop I would be glad to work with you on that.
Thanks for offering.
--Andreas
From ziem at excite.com Sat Mar 9 13:41:19 2002
From: ziem at excite.com (ziem at excite.com)
Date: Sun, 10 Mar 2002 02:41:19 +0500
Subject: ** OREGON INSTANT WINNER!! **
Message-ID: <015c28a46b4c$5353d0c1$8ba73de8@skegxh>
An HTML attachment was scrubbed...
URL:
From Conference.C at abo.fi Wed Mar 13 12:15:18 2002
From: Conference.C at abo.fi (Conference.C at abo.fi)
Date: Wed, 13 Mar 2002 12:15:18 -0800 (PST)
Subject: Do you make conference calls?
Message-ID: <1016045309.0074313399@ra.abo.fi>
An HTML attachment was scrubbed...
URL:
From hliu at inzigo.com Thu Mar 14 13:57:29 2002
From: hliu at inzigo.com (Hongqin Liu)
Date: Thu, 14 Mar 2002 16:57:29 -0500
Subject: class-SLM
Message-ID: <3C911CC9.C47D16B1@inzigo.com>
Hi,
I am trying to construct a class based trigram LM. The function
"ngram-class" only induces classes for a bigram model. I have my own
class definitions with the class-format. When I use these definition
with the "ngram" function (-classes option), the LM leads to a higher
perplexity and word error rate than those from a word based trigram. Is
there any other approach with which I can get a class-based LM with
lower perplexity the a word-based?
By the way, anyone tried a 4gram model with pfsg format?
Thanks!
Hongqin Liu
From stolcke at speech.sri.com Thu Mar 14 14:25:18 2002
From: stolcke at speech.sri.com (Andreas Stolcke)
Date: Thu, 14 Mar 2002 14:25:18 PST
Subject: class-SLM
In-Reply-To: Your message of Thu, 14 Mar 2002 16:57:29 -0500.
<3C911CC9.C47D16B1@inzigo.com>
Message-ID: <200203142225.OAA05042@huge>
Hongqin,
there is no guarantee that a class-based LM will have lower perplexity
than a word-based one. For small, task-oriented domains with little
training data (think ATIS), you can usually get a good improvement
with hand-defined word classes that reflect the properties of
the domain. For large-vocabulary, unconstrainted domains (such as
Switchboard or Broadcast News), a class-based LM by itself will usually
have higher perplexity. However, you can usually get a nice
perplexity reduction by interpolating the word and the class-based LMs.
Mostly, the class-based LM helps with the prediction of unseen word ngrams.
It is pure laziness that the make-ngram-pfsg script cannot handle
4-gram and higher-order LMs at this point. It shouldn't be hard to
do. If anybody wants to contribute a generalized version I'd be happy
to incorporate it.
--Andreas
In message <3C911CC9.C47D16B1 at inzigo.com>you wrote:
> Hi,
>
> I am trying to construct a class based trigram LM. The function
> "ngram-class" only induces classes for a bigram model. I have my own
> class definitions with the class-format. When I use these definition
> with the "ngram" function (-classes option), the LM leads to a higher
> perplexity and word error rate than those from a word based trigram. Is
> there any other approach with which I can get a class-based LM with
> lower perplexity the a word-based?
>
> By the way, anyone tried a 4gram model with pfsg format?
>
> Thanks!
>
> Hongqin Liu
>
>
From stolcke at speech.sri.com Thu Mar 14 16:54:16 2002
From: stolcke at speech.sri.com (Andreas Stolcke)
Date: Thu, 14 Mar 2002 16:54:16 PST
Subject: New posting policy for srilm-user
Message-ID: <200203150054.QAA20651@huge>
Of late, we have seen quite a bit of junk mail being sent to
srilm-user at speech.sri.com. I have therefore changed the posting
policy for the list so that only subscribed members of the list
are allowed to post to it.
Therefore, please make sure you post from an email account that matches
the address you are subscribed under. You can unsubscribe and
re-subscribe yourself if necessary. (For detailed instructions on how
to do this, mail the line "help" in the body of a message to
majordomo at speech.sri.com.)
Sorry for any inconvenience. Regards,
--Andreas
From stolcke at speech.sri.com Tue Mar 19 15:15:25 2002
From: stolcke at speech.sri.com (Andreas Stolcke)
Date: Tue, 19 Mar 2002 15:15:25 PST
Subject: SRILM transcription format
In-Reply-To: Your message of Tue, 19 Mar 2002 14:56:24 -0800.
<000701c1cf99$5416a280$dd00a8c0@dejima.com>
Message-ID: <200203192315.PAA25057@huge>
Ben,
SRILM does not rely on any fancy transcription conventions.
It tokenizes the input using the strtok() function from the C library.
It doesn't know about XML or any other tagging schemes.
What this boils down to is:
Everything that is separated by whitespace (space, newline, tabs) is
considered a word. Case distinctions are preserved unless you use the
"-tolower" option in various tools. Punctuation is treated as just another
non-whitespace character. So you would have to strip punctuation if you
wanted to ignore it in your modeling, or surround punctuation marks with
whitespace if you wanted to model them as word tokens of their own.
--Andreas
In message <000701c1cf99$5416a280$dd00a8c0 at dejima.com>you wrote:
> Andreas,
>
> Hello, could you point me to a document describing in detail the
> transcription conventions for SRILM tools?
>
> For example, can words be capitalized? What punctuation is permitted
> (apostrophe? period? comma?)
>
> Thank you,
>
> ________________________________
> Ben Reaves benreaves at ieee.org
>
>
From hliu at inzigo.com Tue Mar 26 11:38:41 2002
From: hliu at inzigo.com (Hongqin Liu)
Date: Tue, 26 Mar 2002 14:38:41 -0500
Subject: Nuance grammar
Message-ID: <3CA0CE41.4D5DC4C9@inzigo.com>
Hi, folks,
This is a question beyound the SRI tk. But I would like to ask you guys
here since I found you guys are really smart and nice, and were very
helpful for me.
I have two grammars: one is the pfsg format (SLM), another one with any
other format from cfg. I used a top grammar includes them: #include
"dir/pfsg.gammar"
#include "dir2/any.grammar"
I found that Nuance worked well when any of the "#include " is included,
but gave a error message when both the two lines are included:
ERROR: RAPI::ProcessContext: Failed to set context.
ERROR: NodeArrayHandler::SetGrammarAndNodeArray: Grammar "?]??dl^A?dl^A?
^D:
?eTK??bK?\200^??dl^A?NTK?" not found!
ERROR: RecEngine::StartUtterance: Called with bad grammar ?]??dl^A?dl^A?
^D:
?eTK??bK?\200^??dl^A?NTK?
ERROR: PP_METHOD :: CheckFrameIncrement: illegal frame increment (0 ->
-1)
ERROR: PP_METHOD :: InitFrame: Invalid frame increment detected!
ERROR: RecEngine::MidRecognizeSentence: Cannot pp->InitFrame(-1)!
ERROR: PP_METHOD :: CheckFrameIncrement: illegal frame increment (-1 ->
-1)
ERROR: PP_METHOD :: InitFrame: Invalid frame increment detected!
ERROR: RecEngine::MidRecognizeSentence: Cannot pp->InitFrame(-1)!
Anyone has a solution?
Great thanks!
Hongqin