feature vectors for some phonemes not getting properly merged about dev

[drammock]

221 reduced down to 28 by moving glyphID assignment to occur after markMarginal.

5 of these are possibly glyph errors in nmn from the GM datasource: ˀnʘ, ˀnǁ, ˀnǂ, ˀnǀ, ˀnǃ probably should all be ŋ instead of n. A couple are bare click symbols with missing dorsal place glyphs, and most of the rest are breathy pre-aspiration phonemes.

[drammock]

This is bringing to light a number of IPA issues that I thought we had already resolved. For example, I thought we agreed to eliminate all cases of dʰ in favor of dʱ (and similarly for all other cases of (pre-)aspiration on base segments that are voiced). Consequently I tried to just replace dʰ with dʱ in the features table, but in fact that causes lots of cases of dʰ (which I thought had been eliminated in the source data) to show up in the "not getting assigned a feature vector" list. Before I start changing every single case of ʰ on voiced segments to ʱ in the source data, I wanted to check in with you @bambooforest in case I'm forgetting some past decision about keeping phonemes like dʰ despite their phonetically questionable status.

[bambooforest]

It looks like this was simply (and on my part) not resolved:

https://code.google.com/p/phoible/issues/detail?id=84

So perhaps the best way to deal with these and the other issues for
future's sake is to keep track of them and resolve them in these git issues
by clearly stating what and why.

So any voiced segment that is aspirated, regardless pre or post, should use
the <ʱ> U+02B1 and voiceless should be regular IPA aspirated symbol <ʰ>
U+02B0. Correct?

We have some problematic cases:

http://phoible.org/parameters/7DA7B5896CFFC7827CBFB7C325E61FAB#6/-21.289/27.401

http://phoible.org/parameters/7D6B77168A83A83505CC8AF54D22B9DC#6/-21.289/27.401

I will go through the other google code issues ASAP.

On Wed, Apr 15, 2015 at 11:49 AM, Daniel McCloy [email protected]
wrote:

This is bringing to light a number of IPA issues that I thought we had
already resolved. For example, I thought we agreed to eliminate all cases
of dʰ in favor of dʱ (and similarly for all other cases of
(pre-)aspiration on base segments that are voiced). Consequently I tried to
just replace dʰ with dʱ in the features table, but in fact that causes
lots of cases of dʰ (which I thought had been eliminated in the source
data) to show up in the "not getting assigned a feature vector" list.
Before I start changing every single case of ʰ on voiced segments to ʱ in
the source data, I wanted to check in with you @bambooforest
https://github.com/bambooforest in case I'm forgetting some past
decision about keeping phonemes like dʰ despite their phonetically
questionable status.

—
Reply to this email directly or view it on GitHub
#72 (comment).

[drammock]

The aspiration issue is now #73.