Possible issues with the Barwar Neo-Aramaic (assy1241)

[Cymelo92]

I think there are a few issues in the Barwar Neo-Aramaic number list, especially in the analysis/coding of numerals 11–19 and the tens (30–90). My comments below are suggestions, but I believe they reflect the morphological system more accurately than the current “maximal splitting” approach.

1. Surface vs. underlying representation is reversed for 11–19: For numerals 11–19, the annotation of surface vs. underlying representations appears to be wrong. For example, if twelve is trǝ́ssǝr, the segmentation/transcription should be:

t r + ǝ/a s s ǝ r

not

t r + a/ǝ s s ǝ r

(i.e., the surface form should be on the left of the slash).

2. Units in 11–19 can be treated as one morpheme with multiple surface realizations: For numerals 11–19, the unit part of the form seems to be transparently derivable from the base numerals. For example, if two is tre and twelve is trǝ́ssǝr, I would not split this into separate morphemes such as “two” = t r e vs. “two2” = t r. Instead, it seems more appropriate to treat this as a single morpheme (“two”) with two surface realizations, e.g.:

• t r e
• t r -/e

both labeled “two” and sharing the same morpheme and COGID. The same applies to most numerals 11–19 (though 13 and 17 may be less transparent).

3. Tens 30–90 are derivable from 3–9 + -i: The tens from 30 to 90 look like they can be derived straightforwardly from 3–9 by dropping the final vowel and adding the morpheme -i (etymologically the plural marker on nouns/adjectives). For example:

• three = tˤ lˤ a θ a
• thirty = tˤ lˤ a θ -/a + i

Under this analysis, the morphemes should be annotated as “three + ty” rather than as a single morpheme “thirty”.

[LinguList]

These points are important. The formula for the slash construction is surface/underlying. If we flip it, we run into trouble in the analysis, but our checks should also point out that there are such problems.

[LinguList]

I know it may seem as difficult to implement, but I am quite confident that there is a correct middle ground as to the depth of the analysis. Here, I would also argue that those who KNOW a language (as linguists or as speakers) may see MORE than those who analyze data blindly. The reason is that the kind of work requires a basic idea about numeral systems, and I'd even argue that humans who learn numbers in childhood will develop such ideas, as is reflected also in contamination that leads to certain pseudo-morphemes that help humans to make sense of partial structures in numerals.

[LinguList]

@Cymelo92, would you have the capacity to check the analysis and revise it?

[Cymelo92]

Sure, I could do that. I just want first to hear if that's ok with @chundrac , who is the creator of the list (if I'm not wrong).

[chundrac]

Ah OK, I didn't properly understand this convention, and will have to review and change all of my analyses in that case (I thought the slash delimited phonologically conditioned variants, but didn't realize that it represented SR/UR). (This will not affect analyses for our immediate project, which relies only on analyses at the morpheme level, but I will try to fix this ASAP.)

@Cymelo92, if you want to revise assy1241, that is fine with me.

[LinguList]

The SLASH is crucial for an annotation that allows us to group different allomorphs by string identity.

We call them inline alignments (introduced here, and earlier in a blog here).

The basic idea is that EDICTOR processes a string internally by addressing the right part after the slash or the main part if there is no slash. This allows us to group strings differently in the EDICTOR app and check them. We call the fictive root that we use to group forms (usually one form is taken as the precedent) the "underlying" form, but this is a technical device, not a theoretical thing.

Like German "eins" in "einundzwanzig" being represented as "ai n s" in "one" and "ai n -/s + U n + ts v a n ts i ç" in the latter.

This is like an alignment:

ai n s
ai n -

But it is crucial for the annotation and the logic of what is one morpheme with different allomorphs vs. what are different morphemes.

[Cymelo92]

Ah OK, I didn't properly understand this convention, and will have to review and change all of my analyses in that case (I thought the slash delimited phonologically conditioned variants, but didn't realize that it represented SR/UR). (This will not affect analyses for our immediate project, which relies only on analyses at the morpheme level, but I will try to fix this ASAP.)

@Cymelo92, if you want to revise assy1241, that is fine with me.

Sure, I'll do that

[Cymelo92]

The list had some additional issues:

• Wrong transcriptions (glottal stops transcribed as pharyngeal fricatives)
• The feminine forms of the numerals were missing

I corrected these issues too. In the future, if anybody plans to create a list for a Semitic language, let me know and I'll check it