From dhawaljshah at gmail.com  Wed Sep  2 07:18:45 2009
From: dhawaljshah at gmail.com (Dhawal Shah)
Date: Wed, 2 Sep 2009 19:18:45 +0800
Subject: [Olympus developers 154]:  Testing the backend
Message-ID: <bdea07120909020418sa3e7b9bt7eaa3b842f8fbdc1@mail.gmail.com>

Hi,
I may seem a little ignorant - but in the RoomlineBackEnd folder, there is a
text file that states:
To test the Back End Server from your PC:

1.  Start runBE_server.bat.
2.  In another window, run the testBE.pl script.

All I see being output when I run testBE.pl is as follows:
USAGE: testABE.pl <integer > 0>

How exactly does doing this test the back end server?
Thanks.

Regards,
Dhawal Shah
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.srv.cs.cmu.edu/pipermail/olympus-developers/attachments/20090902/fa808979/attachment.html

From bfrisch at cs.cmu.edu  Wed Sep  2 12:16:34 2009
From: bfrisch at cs.cmu.edu (Benjamin Frisch)
Date: Wed, 2 Sep 2009 11:16:34 -0500
Subject: [Olympus developers 155]: Re: Testing the backend
In-Reply-To: <bdea07120909020418sa3e7b9bt7eaa3b842f8fbdc1@mail.gmail.com>
References: <bdea07120909020418sa3e7b9bt7eaa3b842f8fbdc1@mail.gmail.com>
Message-ID: <bd80926e0909020916l7d4adf9bgf373186dad80dab4@mail.gmail.com>

Hi Dhawal,
  As explained further on in that document, the second parameter to
testBE.pl is the array location of the frame in
TestFrames.pm<http://trac.speech.cs.cmu.edu/repos/olympus/example-systems/RoomLine/trunk/Agents/RoomLineBackend/TestFrames.pm>that
should be sent to the backend; however, TestFrames.pm is out of date
(and is probably best removed) with the new frame format change on the
Olympus trunk.  Your better option for right now is to run RoomLine and look
at the output, as each incoming backend frame is printed out along with how
it is parsed and processed.  I will see what I can do to either create some
documentation on writing a backend and/or a better test script.

Thanks,
Ben

On Wed, Sep 2, 2009 at 6:18 AM, Dhawal Shah <dhawaljshah at gmail.com> wrote:

> Hi,
> I may seem a little ignorant - but in the RoomlineBackEnd folder, there is
> a text file that states:
> To test the Back End Server from your PC:
>
> 1.  Start runBE_server.bat.
> 2.  In another window, run the testBE.pl script.
>
> All I see being output when I run testBE.pl is as follows:
> USAGE: testABE.pl <integer > 0>
>
> How exactly does doing this test the back end server?
> Thanks.
>
> Regards,
> Dhawal Shah
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.srv.cs.cmu.edu/pipermail/olympus-developers/attachments/20090902/cd9d33ba/attachment.html

From ligoriotiziana at gmail.com  Thu Sep  3 11:41:27 2009
From: ligoriotiziana at gmail.com (Tiziana Ligorio)
Date: Thu, 3 Sep 2009 11:41:27 -0400
Subject: [Olympus developers 156]:  explicit confirm help
Message-ID: <6b83216e0909030841p2007e63ey716e61ee69f2fbba@mail.gmail.com>

Hi,
I'm trying to set up our system so that certain concepts will be always
explicitly confirmed.
However, when the answer to an explicit confirm in "NO", the system keeps on
asking for a confirmation instead of trying to obtain the concept again.

This is my set up:

I have an expl.pol as follows:

EXPLORATION_MODE=epsilon-greedy
EXPLORATION_PARAMETER=0.1

                ACCEPT     EXPL_CONF     IMPL_CONF
INACTIVE            1          -             -
CONFIDENT           -          10            -
UNCONFIDENT         -          10            -
GROUNDED            1           -            -

and the grounding.policies file includes expl = expl.pol

then in the Task.cpp I specify the explicit policy for certain concepts, for
example:

STRING_USER_CONCEPT(area_code, "expl")

and the ExplicitConfirm.pm file in the NLG specifies the appropriate
prompts.


The system does correctly ask for an explicit confirmation of the concept,
but it does not properly respond to a "NO" answer, and it simply keeps on
repeating the explicit confirm prompt.
I guess I"m missing where it is that I should specify the system's behavior
to try and obtain the concept again, but I thought that a "NO" answer to an
explicit confirmation would simply put back on the stack the agent that
initially requested the concept.

Can anyone help?
Thank you,
Tiziana Ligorio
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.srv.cs.cmu.edu/pipermail/olympus-developers/attachments/20090903/eb5fa479/attachment.html

From antoine.raux at gmail.com  Thu Sep  3 15:28:54 2009
From: antoine.raux at gmail.com (Antoine Raux)
Date: Thu, 03 Sep 2009 12:28:54 -0700
Subject: [Olympus developers 157]: Re: explicit confirm help
In-Reply-To: <6b83216e0909030841p2007e63ey716e61ee69f2fbba@mail.gmail.com>
References: <6b83216e0909030841p2007e63ey716e61ee69f2fbba@mail.gmail.com>
Message-ID: <4AA018F6.2050001@polytechnique.org>

Hi Tiziana,

Well what *should* happen is that the concept should go to the INACTIVE 
state after a "NO". What happens next depends on your task tree but most 
of the time, it will request the concept again (because the request 
agent should still be in a non-completed state at the top of the stack). 
Could you send to the mailing list both your task tree definition file 
and a log file that displays the behavior you're mentioning?

antoine

Tiziana Ligorio wrote:
> Hi,
> I'm trying to set up our system so that certain concepts will be 
> always explicitly confirmed.
> However, when the answer to an explicit confirm in "NO", the system 
> keeps on asking for a confirmation instead of trying to obtain the 
> concept again.
>
> This is my set up:
>
> I have an expl.pol as follows:
>
> EXPLORATION_MODE=epsilon-greedy
> EXPLORATION_PARAMETER=0.1
>
>                 ACCEPT     EXPL_CONF     IMPL_CONF
> INACTIVE            1          -             -
> CONFIDENT           -          10            -
> UNCONFIDENT         -          10            -
> GROUNDED            1           -            -
>
> and the grounding.policies file includes expl = expl.pol
>
> then in the Task.cpp I specify the explicit policy for certain 
> concepts, for example:
>
> STRING_USER_CONCEPT(area_code, "expl")
>
> and the ExplicitConfirm.pm file in the NLG specifies the appropriate 
> prompts.
>
>
> The system does correctly ask for an explicit confirmation of the 
> concept, but it does not properly respond to a "NO" answer, and it 
> simply keeps on repeating the explicit confirm prompt.
> I guess I"m missing where it is that I should specify the system's 
> behavior to try and obtain the concept again, but I thought that a 
> "NO" answer to an explicit confirmation would simply put back on the 
> stack the agent that initially requested the concept.
>
> Can anyone help?
> Thank you,
> Tiziana Ligorio
>
>


From nkatsaounos at sch.gr  Thu Sep 17 10:48:33 2009
From: nkatsaounos at sch.gr (=?iso-8859-7?B?zd/q7/IgyuH08+Hv/e3v8g==?=)
Date: Thu, 17 Sep 2009 17:48:33 +0300
Subject: [Olympus developers 158]:  no_understanding handling
Message-ID: <F565028DAFBD47F49BC370F1B2D750B6@wcl.ee.upatras.gr>

Hello,

I have build-up a small dialog application based on RavenClaw/Olympus application.
My "problem": I noticed that whenvever I say something that cannot be recognized (eg say a word that is not included in my dictionary), system is still waiting for my input. Ravenclaw dialog "continues" only when I say an utternace that can be recognized (so dialog continues to the next step) or when a time interval is reached up (I think about 60secs- so dialog repeats its last state).
My "target": I want to make my dialog continue to the next step even when I say something that cannot be recognized. In other words I want to make my dialog continue to the next step as soon as the end-utterance is reached-up, even when recognition has no available results. 

So..
How can I do this?
--Is there any configuration I can make inside a.cfg file (eg Apollo.cfg) in order to have the above described behaviour? 
--Is there any Ravenclaw commands I can use inside my dialog (DialogTask.cpp) in order to have the above described behaviour? (eg section for no_understanding issues).
--Do I have to modify some piece of code and where?

Thanks in advance

Nikos Katsaounos
University of Patra-Greece
Wireless Communication Laboratory



-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.srv.cs.cmu.edu/pipermail/olympus-developers/attachments/20090917/4ccd1126/attachment.html

From antoine.raux at gmail.com  Thu Sep 17 12:32:35 2009
From: antoine.raux at gmail.com (Antoine Raux)
Date: Thu, 17 Sep 2009 09:32:35 -0700
Subject: [Olympus developers 159]: Re: no_understanding handling
In-Reply-To: <13519_1253199324_n8HEtMcC008277_F565028DAFBD47F49BC370F1B2D750B6@wcl.ee.upatras.gr>
References: <13519_1253199324_n8HEtMcC008277_F565028DAFBD47F49BC370F1B2D750B6@wcl.ee.upatras.gr>
Message-ID: <4AB264A3.3050205@polytechnique.org>

Hi Nikos,

System behavior on non-understandings is part of the grounding 
mechanisms handled by RavenClaw. You shouldn't need to modify any code 
(except your task specification) but add some config files, etc.
Although I'm not sure if it's 100% up-to-date (or complete), 
http://wiki.speech.cs.cmu.edu/olympus/index.php/Grounding_in_RavenClaw 
should get you started (look at Turn Grounding). You can also check one 
of the example systems (e.g. MeetingLine) since they all have some 
behavior coded for non-understandings (smarter than just waiting).

Hope this helps,
antoine

????? ?????????? wrote:
> Hello,
>  
> I have build-up a small dialog application based on RavenClaw/Olympus 
> application.
> _My "problem"_: I noticed that whenvever I say something that cannot 
> be recognized (eg say a word that is not included in my dictionary), 
> system is still waiting for my input. Ravenclaw dialog "continues" 
> only when I say an utternace that can be recognized (so dialog 
> continues to the next step) or when a time interval is reached up (I 
> think about 60secs- so dialog repeats its last state).
> _My "target":_ I want to make my dialog continue to the next step even 
> when I say something that cannot be recognized. In other words I want 
> to make my dialog continue to the next step as soon as the 
> end-utterance is reached-up, even when recognition has no available 
> results.
>  
> So..
> How can I do this?
> --Is there any configuration I can make inside a.cfg file (eg 
> Apollo.cfg) in order to have the above described behaviour?
> --Is there any Ravenclaw commands I can use inside my dialog 
> (DialogTask.cpp) in order to have the above described behaviour? (eg 
> section for no_understanding issues).
> --Do I have to modify some piece of code and where?
>  
> Thanks in advance
>  
> Nikos Katsaounos
> University of Patra-Greece
> Wireless Communication Laboratory
>  
>  
>
>  
>


From nkatsaounos at wcl.ee.upatras.gr  Fri Sep 18 09:17:57 2009
From: nkatsaounos at wcl.ee.upatras.gr (=?iso-8859-7?B?zd/q7/IgyuH08+Hv/e3v8g==?=)
Date: Fri, 18 Sep 2009 16:17:57 +0300
Subject: [Olympus developers 160]: Re: no_understanding handling
References: <13519_1253199324_n8HEtMcC008277_F565028DAFBD47F49BC370F1B2D750B6@wcl.ee.upatras.gr>
	<4AB264A3.3050205@polytechnique.org>
Message-ID: <8162BF77498C4ADD865C211E8C61F9F9@wcl.ee.upatras.gr>

Dear Antoine,
thanks for your answer!

I tried to "implement" your insrtuctions without success at the moment..
More specifically:
I tried to implement the "request_default" policy in my dialog. I attach the
Raveclaw.cfg file as well as the grounding.policies/request_default.pol
files. Inside my DialogTask.cpp file I write for the desired agent the
following code:
DEFINE_AGENCY( CMyGame,

IS_MAIN_TOPIC()

DEFINE_CONCEPTS(

...

)

DEFINE_SUBAGENTS(

SUBAGENT(RequestAction, CRequestAction, "request_default")

SUBAGENT(ExecuteBackendCall, CExecuteBackendCall, "")

)

..

When I ran my application,  I type init_session and RavenClaw module appears
the following error:

FATAL_ERROR at 15:42:23.841 (44)] Error in CGMRequestAgent::LoadPolicy().
Invalid state-space size (3 states expected, 0 found). <file
.\DMCore\Grounding\GroundingModels\GMRequestAgent.cpp, line 156>.

What is going wrong?

Well, I am not sure that I have understand the following section of
http://wiki.speech.cs.cmu.edu/olympus/index.php/Grounding_in_RavenClaw
you refer to

Policy files for agents have a similar structure. For a hand-written policy,
it first defines a small number of states using input features (mostly
features computed by Helios, starting with h4_...) in blocks like the
following. This example defines a state called UNDERSTANDING for any turn
that is not a non-understanding:
STATE[UNDERSTANDING]
 [h4_nonu] = 0
END
So..do I have to write something more in order to define the states refered
to request_default.pol(UNDERSTANDING, NONUNDERSTANDING, FAILED) ? If so,
where do I have to write such STATE definitions? Inside .pol file, inside
DialogTask.cpp or anywhere else? In which point/line?Thanks in advance!Nikos
Katsaounos ----- Original Message ----- 
From: "Antoine Raux" <antoine.raux at gmail.com>
To: "????? ??????????" <nkatsaounos at sch.gr>
Cc: <olympus-developers at cs.cmu.edu>
Sent: Thursday, September 17, 2009 7:32 PM
Subject: Re: [Olympus developers 158]: no_understanding handling


> Hi Nikos,
>
> System behavior on non-understandings is part of the grounding mechanisms
> handled by RavenClaw. You shouldn't need to modify any code (except your
> task specification) but add some config files, etc.
> Although I'm not sure if it's 100% up-to-date (or complete),
> http://wiki.speech.cs.cmu.edu/olympus/index.php/Grounding_in_RavenClaw
> should get you started (look at Turn Grounding). You can also check one of
> the example systems (e.g. MeetingLine) since they all have some behavior
> coded for non-understandings (smarter than just waiting).
>
> Hope this helps,
> antoine
>
> ????? ?????????? wrote:
>> Hello,
>>  I have build-up a small dialog application based on RavenClaw/Olympus
>> application.
>> _My "problem"_: I noticed that whenvever I say something that cannot be
>> recognized (eg say a word that is not included in my dictionary), system
>> is still waiting for my input. Ravenclaw dialog "continues" only when I
>> say an utternace that can be recognized (so dialog continues to the next
>> step) or when a time interval is reached up (I think about 60secs- so
>> dialog repeats its last state).
>> _My "target":_ I want to make my dialog continue to the next step even
>> when I say something that cannot be recognized. In other words I want to
>> make my dialog continue to the next step as soon as the end-utterance is
>> reached-up, even when recognition has no available results.
>>  So..
>> How can I do this?
>> --Is there any configuration I can make inside a.cfg file (eg Apollo.cfg)
>> in order to have the above described behaviour?
>> --Is there any Ravenclaw commands I can use inside my dialog
>> (DialogTask.cpp) in order to have the above described behaviour? (eg
>> section for no_understanding issues).
>> --Do I have to modify some piece of code and where?
>>  Thanks in advance
>>  Nikos Katsaounos
>> University of Patra-Greece
>> Wireless Communication Laboratory
>>
>>
>
>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: RavenClaw.cfg
Type: application/octet-stream
Size: 1657 bytes
Desc: not available
Url : http://mailman.srv.cs.cmu.edu/pipermail/olympus-developers/attachments/20090918/2f9b9a27/RavenClaw.obj
-------------- next part --------------
A non-text attachment was scrubbed...
Name: request_default.pol
Type: application/octet-stream
Size: 1110 bytes
Desc: not available
Url : http://mailman.srv.cs.cmu.edu/pipermail/olympus-developers/attachments/20090918/2f9b9a27/request_default.obj
-------------- next part --------------
A non-text attachment was scrubbed...
Name: grounding.policies
Type: application/octet-stream
Size: 107 bytes
Desc: not available
Url : http://mailman.srv.cs.cmu.edu/pipermail/olympus-developers/attachments/20090918/2f9b9a27/grounding.obj

From nkatsaounos at sch.gr  Mon Sep 21 10:49:13 2009
From: nkatsaounos at sch.gr (=?iso-8859-7?B?zd/q7/IgyuH08+Hv/e3v8g==?=)
Date: Mon, 21 Sep 2009 17:49:13 +0300
Subject: [Olympus developers 161]: Re: no_understanding handling
References: <13519_1253199324_n8HEtMcC008277_F565028DAFBD47F49BC370F1B2D750B6@wcl.ee.upatras.gr>
	<4AB264A3.3050205@polytechnique.org>
Message-ID: <F7FA385A8D0F42D3A05A72C831DEFDE0@wcl.ee.upatras.gr>

Dear Antoinne,

I also attach some messages that appear during loading of  grounding models, 
that I hadn't notice before..
I confirmed that the .pol files I use are stored into the correct path . I 
assume that something is wrong inside the .pol file..

Nikos Katsaounos

[GMS at 17:44:41.184 (47)] Loading grounding models ...
[GMS at 17:44:41.185 (48)] Loaded grounding model from expl_impl.pol .
[GMS at 17:44:41.185 (48)] Could not read grounding model from  expl_impl.pol
 .
[GMS at 17:44:41.185 (48)] Loaded grounding model from expl.pol .
[GMS at 17:44:41.185 (48)] Could not read grounding model from  expl.pol
 .
[GMS at 17:44:41.185 (48)] Loaded grounding model from request_default.pol .
[GMS at 17:44:41.185 (48)] Could not read grounding model from 
request_default.pol
 .
[GMS at 17:44:41.186 (49)] Loaded grounding model from request_lr.pol .
[GMS at 17:44:41.186 (49)] Could not read grounding model from  request_lr.pol 
.
[GMS at 17:44:41.186 (49)] 4 grounding model(s) loaded.
[CTH at 17:44:41.186 (49)] Auxiliary core dialog management agents created 
successfully.
[CTH at 17:44:41.186 (49)] Core initialization completed successfully.
[GMS at 17:44:41.186 (49)] Registering grounding action: NO_ACTION [NO_ACTION].
[GMS at 17:44:41.186 (49)] Registering grounding action: ACCEPT [ACCEPT].
[GMS at 17:44:41.186 (49)] Registering grounding action: EXPL_CONF [EXPL_CONF].
[GMS at 17:44:41.186 (49)] Registering grounding action: IMPL_CONF [IMPL_CONF].
[GMS at 17:44:41.186 (49)] Registering grounding action: ASK_REPEAT 
[ASK_REPEAT].
[GMS at 17:44:41.186 (49)] Registering grounding action: NOTIFY_AND_ASK_REPEAT 
[ASK_REPEAT].
[GMS at 17:44:41.186 (49)] Registering grounding action: ASK_REPHRASE 
[ASK_REPHRASE].
[GMS at 17:44:41.186 (49)] Registering grounding action: 
NOTIFY_AND_ASK_REPHRASE [ASK_REPHRASE].
[GMS at 17:44:41.186 (49)] Registering grounding action: REPEAT_PROMPT 
[REPEAT_PROMPT].
[GMS at 17:44:41.186 (49)] Registering grounding action: 
NOTIFY_AND_REPEAT_PROMPT [REPEAT_PROMPT].
[GMS at 17:44:41.186 (49)] Registering grounding action: EXPLAIN_MORE 
[EXPLAIN_MORE].
[GMS at 17:44:41.186 (49)] Registering grounding action: 
NOTIFY_AND_EXPLAIN_MORE [EXPLAIN_MORE].
[GMS at 17:44:41.186 (49)] Registering grounding action: FULL_HELP [FULL_HELP].
[GMS at 17:44:41.186 (49)] Registering grounding action: NOTIFY_AND_FULL_HELP 
[FULL_HELP].
[GMS at 17:44:41.186 (49)] Registering grounding action: WHAT_CAN_I_SAY 
[WHAT_CAN_I_SAY].
[GMS at 17:44:41.186 (49)] Registering grounding action: TERSE_WHAT_CAN_I_SAY 
[WHAT_CAN_I_SAY].
[GMS at 17:44:41.186 (49)] Registering grounding action: 
NOTIFY_AND_WHAT_CAN_I_SAY [WHAT_CAN_I_SAY].
[GMS at 17:44:41.186 (49)] Registering grounding action: 
NOTIFY_AND_TERSE_WHAT_CAN_I_SAY [WHAT_CAN_I_SAY].
[GMS at 17:44:41.186 (49)] Registering grounding action: 
TIPS_AND_WHAT_CAN_I_SAY [WHAT_CAN_I_SAY].
[GMS at 17:44:41.186 (49)] Registering grounding action: 
NOTIFY_NONUNDERSTANDING [NOTIFY_NONUNDERSTANDING].
[GMS at 17:44:41.187 (50)] Registering grounding action: YIELD_TURN 
[YIELD_TURN].
[GMS at 17:44:41.187 (50)] Registering grounding action: FAIL_REQUEST 
[FAIL_REQUEST].
[GMS at 17:44:41.187 (50)] Registering grounding action: ASK_STARTOVER 
[ASK_STARTOVER].
[GMS at 17:44:41.187 (50)] Registering grounding action: 
SPEAK_LESS_LOUD_AND_REPROMPT [SPEAKLESSLOUD_AND_REPROMPT].
[GMS at 17:44:41.187 (50)] Registering grounding action: 
ASK_SHORT_ANSWER_AND_REPROMPT [ASK_SHORT_ANSWER_AND_REPROMPT].
[GMS at 17:44:41.187 (50)] Registering grounding action: 
ASK_SHORT_ANSWER_AND_WHAT_CAN_I_SAY [ASK_SHORT_ANSWER_AND_WHAT_CAN_I_SAY].
[GMS at 17:44:41.187 (50)] Registering grounding action: INTERACTION_TIPS 
[INTERACTION_TIPS].
[GMS at 17:44:41.187 (50)] Registering grounding action: GIVE_UP [GIVE_UP].
[GMS at 17:44:41.187 (50)] Registering grounding action: MOVE_ON [MOVE_ON].
[DTT at 17:44:41.187 (50)] Starting Dialog Tree Creation Phase ..
----- Original Message ----- 
From: "Antoine Raux" <antoine.raux at gmail.com>
To: "????? ??????????" <nkatsaounos at sch.gr>
Cc: <olympus-developers at cs.cmu.edu>
Sent: Thursday, September 17, 2009 7:32 PM
Subject: Re: [Olympus developers 158]: no_understanding handling


> Hi Nikos,
>
> System behavior on non-understandings is part of the grounding mechanisms 
> handled by RavenClaw. You shouldn't need to modify any code (except your 
> task specification) but add some config files, etc.
> Although I'm not sure if it's 100% up-to-date (or complete), 
> http://wiki.speech.cs.cmu.edu/olympus/index.php/Grounding_in_RavenClaw 
> should get you started (look at Turn Grounding). You can also check one of 
> the example systems (e.g. MeetingLine) since they all have some behavior 
> coded for non-understandings (smarter than just waiting).
>
> Hope this helps,
> antoine
>
> ????? ?????????? wrote:
>> Hello,
>>  I have build-up a small dialog application based on RavenClaw/Olympus 
>> application.
>> _My "problem"_: I noticed that whenvever I say something that cannot be 
>> recognized (eg say a word that is not included in my dictionary), system 
>> is still waiting for my input. Ravenclaw dialog "continues" only when I 
>> say an utternace that can be recognized (so dialog continues to the next 
>> step) or when a time interval is reached up (I think about 60secs- so 
>> dialog repeats its last state).
>> _My "target":_ I want to make my dialog continue to the next step even 
>> when I say something that cannot be recognized. In other words I want to 
>> make my dialog continue to the next step as soon as the end-utterance is 
>> reached-up, even when recognition has no available results.
>>  So..
>> How can I do this?
>> --Is there any configuration I can make inside a.cfg file (eg Apollo.cfg) 
>> in order to have the above described behaviour?
>> --Is there any Ravenclaw commands I can use inside my dialog 
>> (DialogTask.cpp) in order to have the above described behaviour? (eg 
>> section for no_understanding issues).
>> --Do I have to modify some piece of code and where?
>>  Thanks in advance
>>  Nikos Katsaounos
>> University of Patra-Greece
>> Wireless Communication Laboratory
>>
>>
>
> 


From nkatsaounos at sch.gr  Wed Sep 23 10:36:55 2009
From: nkatsaounos at sch.gr (=?iso-8859-7?B?zd/q7/IgyuH08+Hv/e3v8g==?=)
Date: Wed, 23 Sep 2009 17:36:55 +0300
Subject: [Olympus developers 162]:  Kalliope performance in embedded system
Message-ID: <90215D5E157B4C6A9DBF38C4DA9392FD@wcl.ee.upatras.gr>

Dear Brian,

I am running an Olympus-based dialog system in my desktop pc (Intel Core 2, 2.4GHz, 2GB ram). Everything is OK!
My problems start when my partners try to run it to the real embedded system of the project (VIA PICO-ITX board  --> 1GHz VIA C7, 1GB RAM).  TTS engine (KalliopeFlite) is too slow..That means that play-back of the prompts sounds too "sparsely"..(it sounds like a robot saying words syllable-by-syllable..)
I noticed that KalliopeFlite occupies a significant amount of cpu (in my desktop pc reaches up to 25-29%) even when a short-length prompt is played back.  I also noticed that approx. the same cpu amount is occupied even when KalliopeSAPI is used instead of KalliopeFlite. So, I suspect that cpu effort has to do with kalliope core-code and not the specific tts I may use..

What's your opinion? Am I "missing" something?
Has kalliope and/or flite indeed had such performance problems?
If so, is there any way to make kalliope faster? 
What are the minimum requirements (cpu / ram) in order to have satisfactory tts performance?  Are cpu 1GHZ  / ram 1 GB "enough" for having an acceptable performance?

Mention here that I use the "official" Olympus version (v. 2.2). 

Thanks in advance

Nikos Katsaounos
WCL / University of Patras - Greece
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.srv.cs.cmu.edu/pipermail/olympus-developers/attachments/20090923/2994b480/attachment.html

From blangner at cs.cmu.edu  Wed Sep 23 10:55:03 2009
From: blangner at cs.cmu.edu (Brian Langner)
Date: Wed, 23 Sep 2009 10:55:03 -0400
Subject: [Olympus developers 163]: Re: Kalliope performance in embedded
	system
In-Reply-To: <90215D5E157B4C6A9DBF38C4DA9392FD@wcl.ee.upatras.gr>
References: <90215D5E157B4C6A9DBF38C4DA9392FD@wcl.ee.upatras.gr>
Message-ID: <4ABA36C7.4060609@cs.cmu.edu>

Hi Nikos,

In my experience we haven't noticed any performance problems with 
Kalliope/tts modules, even on low-power systems such as the embedded 
system you describe.  I recall several years ago, in fact, that we 
successfully used Olympus v1 code on Pentium III machines with about 1GB 
of memory.

flite itself is designed to be used in minimal environments, so you're 
correct that it's more likely to be something in the Kalliope code.

In terms of minimum requirements, I'm not sure what the current codebase 
needs to be effective.  I know the production system we use is a modern 
machine, similar in capabilities to your desktop.  Perhaps one of the 
other developers can provide a more precise description of the hardware 
requirements for Olympus.

Brian


????? ?????????? wrote:
> Dear Brian,
> 
> I am running an Olympus-based dialog system in my desktop pc (Intel Core 2, 2.4GHz, 2GB ram). Everything is OK!
> My problems start when my partners try to run it to the real embedded system of the project (VIA PICO-ITX board  --> 1GHz VIA C7, 1GB RAM).  TTS engine (KalliopeFlite) is too slow..That means that play-back of the prompts sounds too "sparsely"..(it sounds like a robot saying words syllable-by-syllable..)
> I noticed that KalliopeFlite occupies a significant amount of cpu (in my desktop pc reaches up to 25-29%) even when a short-length prompt is played back.  I also noticed that approx. the same cpu amount is occupied even when KalliopeSAPI is used instead of KalliopeFlite. So, I suspect that cpu effort has to do with kalliope core-code and not the specific tts I may use..
> 
> What's your opinion? Am I "missing" something?
> Has kalliope and/or flite indeed had such performance problems?
> If so, is there any way to make kalliope faster? 
> What are the minimum requirements (cpu / ram) in order to have satisfactory tts performance?  Are cpu 1GHZ  / ram 1 GB "enough" for having an acceptable performance?
> 
> Mention here that I use the "official" Olympus version (v. 2.2). 
> 
> Thanks in advance
> 
> Nikos Katsaounos
> WCL / University of Patras - Greece


From swaticute30 at gmail.com  Wed Sep 23 20:32:15 2009
From: swaticute30 at gmail.com (Archita)
Date: Thu, 24 Sep 2009 06:02:15 +0530
Subject: [Olympus developers 164]:  Check out my profile..
Message-ID: <55a3eba30909231732j3527a10fu9cd7443608bc03f@mail.gmail.com>

           *SiliconIndia* <http://siliconindia.com/>- *Empowering
Professionals*
<http://www.siliconindia.com/register.php?id=hY7TG9Oh>   Archita
Arora
HR Management
  Hi ,

I would like to invite you to my professional network on SiliconIndia. It
allows us to invite trusted contacts and stay connected with ease.

Thanks,
     Archita

Let us connect and stay in touch.* Click here to
join<http://www.siliconindia.com/register.php?id=hY7TG9Oh>
*
  <http://siliconindia.com/profiles/var+yahoocoin-b3NfTgY9>
Kick Start Your CAT '09 Preparation - *Take a Free CAT Online
Test*<http://siliconindia.com/education/cat_test/index.php>


If you do not wish to receive future mailings from SiliconIndia, please opt
out. <http://www.siliconindia.com/network/optoutemails.php>
Info Connect WTI Pvt. Ltd. 124, Surya Chambers, Airport Main Road, Bangalore
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.srv.cs.cmu.edu/pipermail/olympus-developers/attachments/20090924/06ef908c/attachment.html
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: image/jpeg
Size: 3094 bytes
Desc: not available
Url : http://mailman.srv.cs.cmu.edu/pipermail/olympus-developers/attachments/20090924/06ef908c/attachment.jpe