by Daniel Dacanay
Out of the 112 total entries with itwêwina correspondences, 39 were exact orthographic matches, with identical spellings in the LaCombe as in the SRO. For the 73 remaining entries, the most prevalent discrepancies, by far, were those pertaining to vowel-length marking and aspiration; 39 LaCombe entries had an incorrect or absent vowel length marking for , 16 had incorrect length-marking for an , 2 had missing length markings for an , and 33 were missing an . None of the entries in the sample marked with a circumflex, but given the non-contrasting nature of this marking, I elected not to consider this an error. In any case, all of these errors seem already to be accounted for in itwêwina's current spell-relax, so I don't imagine they shall cause any issues.
The second class major class of discrepancies were those of doubled consonants; 19 LaCombe entries had a doubled , 7 had a doubled , and 4 had a doubled . Given that double consonants are non-existent in the SRO, all such pairs can be converted to their singular counterparts. One potential point of interest here is a possible interaction between doubled consonants and /h/, as in various instances, a doubled consonant (CC) in the LaCombe orthography seems to correspond to an /h/ followed by that consonant (hC) in the SRO. For example:
KIKKÂNÂKWAN - kîhkânâkwan IYIKATETTEW - yîkatêhtêw MÂMÂTTAWIPAYIW - mamâhtâwipayiw
However, this is not always the case, and LaCombe's doubled consonants also correspond to simple single consonants:
MANIKKAHWEW - manikahwêw
Similarly, seems to be exempt from this relationship with
SENIBÂNASSABÂB - sênipânisapâp
The third major class of discrepancies were those of voiced consonants in the LaCombe, although this occurred much less frequently in this sample set than anticipated. Still, one can confidently reassign every LaCombe to an SRO
, to , and (although not present in this set), to :
SENIBÂN - sênipân MANISIGAN - manisikan
The fourth class of discrepancies concerns LaCombe's use of and , which, at least in this dataset, appear to cleanly correspond in the SRO to and respectively, with seen instead of 9 times, and seen instead of 7 times:
MÂMITJIMOW - mamihcimow KIJIK - kîsik
My operating theory on why LaCombe distinguishes between and is that represents a voiced allophone of /s/ (which may have sounded to LaCombe's French ear akin to a /ʒ/), which may explain why only seems to appear intervocalically, where can appear in consonant clusters, such as in the following example which contains both:
« KIJIKÂSTEW, (v. im.) there is moonlight
The to correspondence is slightly more unclear; although it appears that all s in the LaCombe correspond to in the SRO, I have yet to find any meaningful distinction between it and , which appears to have been LaCombe's preferred method of writing /t͡ʃ/. Indeed, we see both being used to write the same morpheme in some some instances:
« KIMOTCH, (ad.) in hiding, e.g. kimotch sipwettew, he leaves in secret, kimotch nama miweyittam, he is not happy on the inside « KIMOTJIYÂWESIW, ok, (a. a.) he is angry in his heart
It seems most likely, then, that the - variation in LaCombe's orthography is simply another case of idiosyncratically recorded allophony, and that both and can be collapsed into the SRO the same way that and can be collapsed into . Although I am still not fully satisfied with this conclusion, and would not rule out the possibility that the modern Cree /s/ and /t͡ʃ/ are in fact modern convergences of numerous, now-lost historical phonemes, for all practical purposes, I would recommend and both being converted wholesale , and and both being similarly converted to .
A fifth class of discrepancies would concern LaCombe's use of , which seems, at least from its 12 appearances in the dataset, to cleanly correspond to :
ÂWASUW - awasow
Finally, there are a number of miscellaneous rewrite rules which do not fit into the above categories. Firstly, there is one instance of in LaCombe corresponding to in the SRO:
KIKAYONISEW - kikayiwinisêw
There is one occurrence of in LaCombe's orthography corresponding to <î> in the SRO:
SESSEGAN - sîsîkan
There is one instance of corresponding to :
SENIBÂNASSABÂB - sênipânisapâp
There are two instances of corresponding to :
MÂMITTISIW - mamihcisiw MÂMITTISIWIN - mamihcisiwin
There are eight instances of in the LaCombe corresponding to <yî> in the SRO:
IYIKÂTENEW - yîkatênêw IYIKÂTENAMÂWEW - yîkatênamawêw IYIKITCHIKÂWIW - yîkicikâwîw etc...
There is one instance of an in the SRO being completely omitted by LaCombe:
ENSIS - êsisis
Similarly, there are three instances of epenthetic s being added to entries which lack them in the SRO (presumably a consequence of LaCombe's French ear perceiving nasal vowels where they were either non-present, non-phonemic, or both):
ENS - êsa ENSIS - êsisis ENSKEW - êskêw
Finally, there are also two instances of an in the SRO not being written in the LaCombe:
PWÂT - pwâta ENS - êsa
In summary, there are 21 hypothetical rewrite rules necessary to convert the dataset LaCombe entries to the SRO, of which 13 could be applied as blanket statements:
A > â O > ô I > î KK > k SS > s TT > t Ø > h G > k B > p TCH > c TJ > c J > s U > o
The remaining 8 only appear to apply in certain, partially idiosyncratic circumstances:
O > iwi Ø > a A > i E > î N > Ø Ø > is IYI > yî TT > c
In summary, then, it seems a significant number of the entries in both the LaCombe dictionary and itwêwina are already orthographically identical, and that of those which are not, most are the product of pervasive discrepancies between the LaCombe and the SRO which can be solved by simple, blanket rewrite rules. However, a substantial minority may also require finer, more idiosyncratic changes, possibly better suited to simple human analysis. That said, this basic strategy of using rewrite rules in FOMA (or something similar) to reform most of the LaCombe entries into the SRO still seems by far to be the most reasonable approach to this problem.