Specifying Stress Level with nltk.corpus.cmudict.dict

[AaronF]

Hi,

I am new to the nltk and Python but I have a very specific question about geting a list of phonemes from a corpora using nltk.corpus.cmudict.dict. I have a corpora that I have already tokenized but I would to turn it into phonemes using the CMU Dictionary,  but I would like to only have the resulting phonomes for each token correlate to stress level 1 from the CMU Dictionary. I have noticed that if I run the prondict[] on my corpora, it gives me a list of phonemes for each stress level for each token, instead of just for a specific stress level. Is this there an easy way to do this with the nltk?  

Thanks

[Joe]

This description seems kind of ambiguous.  Can you provide a couple of simple examples of the output that you are expecting/trying to obtain, versus the incorrect/unexpected output that you are currently seeing? That would probably make it easier for people to answer your question.

[AaronF]

For example, if in Python, I type prondict = nltk.corpus.cmudict.dict() and then type prondict['anointed'], the results are [['AH0', 'N', 'OY1', 'N', 'T', 'AH0', 'D'], ['AH0', 'N', 'OY1', 'N', 'T', 'IH0', 'D'], ['AH0', 'N', 'OY1', 'N', 'AH0', 'D'], ['AH0', 'N', 'OY1', 'N', 'IH0', 'D']], however, I only seek ['AH0', 'N', 'OY1', 'N', 'T', 'IH0', 'D'] as the desired result, and my corpora is very large so I would like to avoid making the distinction on a token by token basis if possible so I was curious if anyone else has done this before with the nltk.

[Joe]

Hi,

 I guess you should be able to get a specific pronunciation by just calling the index,

prondict['anointed'][1]

 although in general it would probably be simplest to just use the first pronunciations, prondict['anointed'][0].  The pronunciation order should (I think) be the same as the cmudict,

https://cmusphinx.svn.sourceforge.net/svnroot/cmusphinx/trunk/cmudict/

  It's still not entirely clear to me what you mean about the stress level though, as your desired result still contains two different stress markers.

Best

[AaronF]

Thank you for your reply, this is what I was looking for.