Word clipping

[Bunger Henry]

I've been listening to a lengthy fantasy series that I converted to text. It has a ton of strangely spelled names that Neospeech Paul and Kate had some trouble with. In an effort to get better pronunciations, I tried tinkering with the basic pronunciation editor. Instead of respelling the words, I tried something a little different by entering in the phonemes like in the advanced editor.

For example, for the name Aemon, I entered <pron sym="ey m ax n"/>

I did this for dozens (no, make that hundreds) of unusual names. Then I did some test sentences, and what I found was that the editted names sounded clipped at the beginning and end of the word. Almost as if the program does not recognize it as a word. In fact, if I strung two or more of these names together, instead of sounding like a series of names, it sounded like just one long name. Obviously there is something missing.

I tried adding a semicolon after the last ">" in the pronunciation field, but that just gives an unacceptably long pause after the word.

Any ideas?

[D.Leikin]

Hi John,

Haven't you tried editing pronunciations with the NeoSpeech voice-specific dictionary editor

..\Program Files\NeoSpeech\Paul16\lib\UserDicEng.exe

Just in case, Paul's dictionary file is

..\Program Files\NeoSpeech\Paul16\data-common\userdict\userdict_eng.csv

and Kate's is

...\Program Files\NeoSpeech\Kate16\data-common\userdict\userdict_eng.csv

The problem is only one of these files seems to be used depending on which voice initializes first, however, a workaround is to keep the files in sync to avoid the need for editing them separately.

[Bunger Henry]

I haven't tried that because I was a little intimidated by the pronunciation symbols in that utility. It wasn't clear to me what each symbol stood for, and I didn't want to go through trial and error on each one. Is there a key somewhere?

[kremesch]

I find this chart to be extremely helpful with neospeech

http://www.worldwidewords.org/pronguide.htm

[D.Leikin]

In fact, if I strung two or more of these names together, instead of sounding like a series of names, it sounded like just one long name. Obviously there is something missing.

I tried adding a semicolon after the last ">" in the pronunciation field, but that just gives an unacceptably long pause after the word.

Any ideas?

John, I just thought that maybe adding some extra spaces could help workaround the issue? What I mean is using pronunciations of the form

'<s> <pron_sym> <s>'

instead of just

'<pron_sym> '

If it helps, it would be possible to reformat the dictionary your are currently using in just couple of clicks by search-replace command

changing '<' to '<s><'

and then

changing '/>' to '/><s>'