[Corpora-List] Corpus methods for prosodic analysis

Tiberiu Boros tibi at racai.ro
Thu Apr 17 09:40:19 CEST 2014


Dear Matias,

Regarding the forced alignment proposed by Simon: there is also an alternative to perform a flat start initialization using HTK, in which case you do not require a pre-existing phonetic model, but if you are using a well resourced language you should use an available model (if possible). Also, from your mail I see that you want to check for a certain pattern in the files. You could try to do this automatically: maybe cluster some values for F0 and see if it hold on the entire corpus (anomaly detection).

Best wishes, Tiberiu

On 16.04.2014 17:59, Simon Smith wrote:
> Hi Matías
>
> I think I agree with Marcin that a prosodic analysis may not be quite
> what you need here. Also it would be nice to know what the actual
> morphological-sentential correlation you found is.
>
> If you do want to look into prosodically motivated segmentation,
> though, you could take a look at my PhD thesis
> http://etheses.bham.ac.uk/19/ . It describes a prosody-based utterance
> classification system. It makes use of the INTSINT and MOMEL prosodic
> analysis algorithms, by Daniel Hirst, and you might like to look those
> up as well https://groups.yahoo.com/neo/groups/momel-intsint/info .
>
> Simon
>
>> W dniu 2014-04-16 08:35, Matías Guzmán Naranjo pisze:
>>> Dear all,
>>>
>>> I found a strong correlation between what seems are sentence boundaries
>>> (in this case marked as punctuation marks), and a particular
>>> morphological alternation. Now I want to actually check for prosodic
>>> patterns in an oral corpus and see whether the correlation is actually
>>> there or if it's just because of idiosyncrasies of how the transcriber
>>> annotated the corpus.
>>>
>>> My question is this, are there any techniques for doing this
>>> automatically? Something like say, manually splitting the audio files in
>>> the relevant text units, but automatically doing an F0 analysis, or
>>> something similar. Also, is there any previous research that does this?
>> Wouldn't simply a search for the particular alternation and a sentence
>> boundary in another written corpus be enough to confirm this? You would
>> have a different annotator. Just asking -- the correlation may be owing
>> to the fact that this is a likely utterance ending.
>>
>> Of course, the sentence ending in a written corpus may or may not
>> correlate with prosody at the utterance ending. But this is another
>> correlation, of sentence boundary annotation and prosody, not of
>> morphological alternation and sentence boundary annotation, right? It
>> may be evidence that the annotator was correct or not, but I'm not sure
>> if that's what you're interested in.
>> HTH,
>> Marcin
> ___________________________
>
>
> Simon Smith, PhD
> Senior Lecturer
> Dept of English & Languages
> Coventry University
>
> +44 2476 887 643
>
> http://www.linkedin.com/pub/simon-smith/42/b77/173
>
> http://tinyurl.com/simoncov
>
> _______________________________________________
> UNSUBSCRIBE from this page: http://mailman.uib.no/options/corpora
> Corpora mailing list
> Corpora at uib.no
> http://mailman.uib.no/listinfo/corpora



More information about the Corpora mailing list