Refactor the Halkomelem mapping and connect it to sal-apa for path to ipa and arpabet

[joanise]

hur_apa_to_hur_orthog.json has a number of lines that could be factored out with "or"s in the context before and after, e.g.:

    {"in": "y̓", "out": "’y", "context_before": "e", "context_after": "i", "prevent_feeding": true},
    {"in": "y̓", "out": "’y", "context_before": "e", "context_after": "a", "prevent_feeding": true},
    {"in": "y̓", "out": "’y", "context_before": "e", "context_after": "ə", "prevent_feeding": true},
    {"in": "y̓", "out": "’y", "context_before": "i", "context_after": "a", "prevent_feeding": true},
    {"in": "y̓", "out": "’y", "context_before": "i", "context_after": "ə", "prevent_feeding": true},
    {"in": "y̓", "out": "’y", "context_before": "a", "context_after": "ə", "prevent_feeding": true},

Note that this would have to be done carefully not to change which contexts are recognized. data/hur.psv clearly shows that "between any pair of vowels" would not be correct.

Also, this mapping is configured rule_ordering: apply-longest-first, but the ordering between rules of the same input length is critical here for correct results, e.g., all the "in": "y̓", cases quoted above are clearly meant to be applied before {"in": "y̓", "out": "y’"},, so as-written would seem more appropriate to me, which would require analyzing the file to make sure everything else is ordered correctly. When we sort, do we use a stable sort? I.e., with apply-longest-first, is as-written a second-order criterion, either by design or by accident?

Next, from hur_orthog_to_hur_apa.json, it would be easy to write a mapping that feeds into sal-apa to connect hur to IPA and ARPABET for cheap. The requirement would be to make sure every "out": in hur_orthog_to_hur_apa.json is either already handled as in "in": in sal_apa_to_ipa.csv or gets mapped to something that does. In the process, we could make the latter more general.

PS: I'm not sure it was correct to use the code hur for the orthog here, since this mapping is explicitly the Island dialect, but hur covers Island, Upriver/Stó꞉lō and Downriver/Musqueam dialects. We will have to split this code if support for the other hur dialects is requested.

[goodzack]

The rule for glottalized resonants and vowels is this:

• for a glottalized resonant R in the set l, m, n, y, w represented in APA as R̓,
• in the environment V₁RV₂,
• the practical orthography glottal stop character ’ is attracted to vowel that is greater in the ordering ə < e < a < i

so əm̓i comes out as um’i but im̓ə comes out as i’mu.

I'm not sure apply-longest-first is best, but it's worked in all of my other transduction scripts in the past. For example, tl’ should come out as ƛ̓, but if you have the rule for l’ apply first, you'll get 'tl̕`. It just happens to be the case that the longer character combinations specify the correct character first... and thus I don't have to figure out what the ordering should be 😛

[goodzack]

I also think it would be more accurate to call the codes hur for APA and hur-orthog for the Island orthography. Part of the impetus of the work here is to convert Island materials into APA for sharing with Downriver speakers and vice versa.

[joanise]

@goodzack Would hur-island work too? Intuitively, "orthog" says to me it's someone's orthography without specifying where or who. But hur-island would lead me to think "Island orthography" in which case I would assume hur to mean "default" or "standard" orthography by contrast.