Home Forums Software Recognizers IAIS 04 ("Speech/non-speech") recognizer errors

IAIS 04 ("Speech/non-speech") recognizer errors

This topic contains 7 replies, has 4 voices, and was last updated by  Eric 3 years, 6 months ago.

Viewing 8 posts - 1 through 8 (of 8 total)
Author Posts
Author Posts
July 15, 2013 at 18:38 #8730

Christopher Cox

Hello,

I wanted to see if anyone had run into any difficulties recently using the IAIS 04 (“Speech/non-speech”) recognizer under ELAN 4.6.1 (OSX 10.7). Using the AVATecHExtensions file available on the website as of July 14, I’m able to run a test WAV file through IAIS 03 (“Fine audio segmentation”), producing the tier “Fine_Segmentation” described in the user guide.

However, when I then attempt to submit that “Fine_segmentation” tier and the same recording as input to IAIS 04 (“Speech/non-speech”), after some uploading and processing time, I receive an error message (“An error occurred inside the recognizer: Recognizer failed”). Looking over the report that the recognizer generates, it seems that (a) one or more of the parameters being supplied to IAIS 04 are invalid, and (b) it might be that the “Speech Sensitivity” parameter isn’t being provided (part of the error log includes a message that this parameter is “mandatory and must be set”, though it might be that this is also part of the default recognizer output). Could it be that a parameter required by IAIS 04 isn’t currently being provided by the output from IAIS 03?

Although I haven’t come across any sections in the user guide that describe this kind of issue in particular, it might be that I’m missing a step somewhere along the way, or that some specific user intervention is required between receiving the output from IAIS 03 and submitting it to IAIS 04. If anyone has any thoughts about this, I would be very grateful to hear them.

All the best,

– Chris

August 9, 2013 at 10:57 #8778

przlen

Dear Chris,

Your observations are correct, it seems that the Sensitivity Parameter was for some reason not included in the configuration file that we provided for download. We will try to fix this asap and upload the fixed version.

Best,
Przemek

October 9, 2013 at 14:53 #8891

przlen

Dear Chris,

Just so you know, the issue with the wrong configuration file has been fixed.

At the moment we are moving our recognizers to a new server, so they are actually unavailable at the moment. But please check in a day or two, everything should be back online.

Best,
Przemek

March 12, 2014 at 11:40 #9284

franziska.haeger

Hello,

for the last two days I tried using the speech / non-speech recognizer via ELAn as well as the Website. However when starting the recognizer I always get the following error mesage.

RESULT: DONE.
RESULT: FAILED.
ERROR: Could not create a temp file for processing! (Permission denied)
DEBUG: Created a temporary file at /tmp/avatech6423079511004160473.tmp

I started a project for the anonymous user called fTP2014. Could you take a look. Is there anything I am doing wrong?

Thanks for helping me out.
Franziska

March 18, 2014 at 16:31 #9293

Eric

Hi Franziska, the /tmp/avatech*.tmp file looks okay on our server, so the error message about creating a temp file seems to be about another file. I have forwarded your report to the authors of this particular recognizer at Fraunhofer IAIS… Are other recognizers working okay for you? Or is this recognizer the only one which you are trying to use? Regards, Eric

March 19, 2014 at 09:37 #9298

franziska.haeger

Hi Eric,

thank you for looking into this. So far the other recognizers seemed to work just fine. So this is the only one I had problems with.

Regards
Franzi

April 10, 2014 at 12:59 #9366

franziska.haeger

Hi,

is there any news on my issue from the people at Fraunhofer?

Greetings
Franziska

May 28, 2014 at 11:55 #9536

Eric

Hi, sorry about the delayed response… I found out that this recognizer has to write temporary files not only in the temp directory but also in the speech model directory. It was acceptable to let it write there – file names change for each run, so there is no risk of conflicts between multiple instances.

With the extra write permissions granted, the recognizer now seems to be working 🙂 Note that you first have to use the fine segmentation recognizer on the same audio file to get a basic segmentation XML tier, which you then use as input tier for the speech / non-speech recognizer on the same recording.

Again, sorry about the delay. Regards, Eric

Viewing 8 posts - 1 through 8 (of 8 total)

You must be logged in to reply to this topic.