Dictionary File Encoding

[Andrew Durstewitz]

Hi All,

Hopefully there is something simple I'm missing here.

I've downloaded a dictionary file from the LMTools website and am attempting
to remove one of the names. However, every time I change the file Sphinx
crashes on me. If I Command+Z (OSX) my way back and save, it's fine.

I think this is an encoding issue of some sort. Any ideas?

Thanks!

[Nickolay V. Shmyrev]

Hopefully there is something simple I'm missing here

Please provide modified dictionary. You can use any file sharing service to
upload it. Don't forget to give us a link after that!

[Andrew Durstewitz]

Here's my modified .dic file; http://www.mediafire.com/?egouvr5e18p6171

[Nickolay V. Shmyrev]

Your dictionary is good. What do you mean by "crashes on me"

[Andrew Durstewitz]

We're using PocketSphinx on the iPhone, and once I start recording (with the
altered file) I get a EXC_BAD_ACCESS calling the following code:

• (void) startDecode {
  ps_start_utt(, NULL);
  }

However, if I revert to the file downloaded from LMTools it works fine. That's
where I'm all confused.

[Nickolay V. Shmyrev]

Hm, when you remove the word from the dictionary, did you leave it in a
language model? Are you familar what language model is?

[Andrew Durstewitz]

I'm just using a dictionary and a grammar file and I've double checked that
the word removed from the dictionary is not in the grammar file.

[Andrew Durstewitz]

So far, from what I can tell - Sphinx is expecting a Unicode (UTF-8, no BOM)
Classic Mac (CR) file. Does any of that sound right?

[Andrew Durstewitz]

Okay, just to close this issue out, here's what I found. Even in the case of
Dict & Gram file, everything that is in the Gram file MUST be in the Dict
file. Unfortunately, Pocketsphinx doesn't seem to be a clear error stating
this fact.

[Nickolay V. Shmyrev]

Nice that this issue is resolved. You probably want to upgrade to latest
pocketsphinx, it has better error messages.