The Dwarves of Discworld

[bradrn]

Thanks Pedant! How did you find all these?

Now some notes on what I found looking through these.

Guards! Guards!

• "Okay [all correctly beamed and propped]?”

Slightly off-topic, but I think that's quite a nice metaphor for a mining species.

(The other nice one from Pratchett is trollish time: since trolls are naturally nocturnal, they associate sunset with beginnings and sunrise with endings. But I digress.)

Feet of Clay

• H’druk g’har dWatch, Sh’rt’azs! H’h Angua tConstable...Angua g’har, b’hk barg’ra Sh’rt’azs Kad’k… “Welcome to the Watch, Corporal Smallbottom! This is Constable Angua...Angua, show Smallbottom how well you’re learning Dwarfish…”

This one's interesting... from the capitalisation, I think we get to see some grammatical markers! d- is either a definite marker or some sort of oblique argument, but I have no idea what t- is.

The Fifth Elephant

• Ha’ak “not a real dwarf”

I thought this one was never actually translated in the book?

The Fifth Elephant

• P’akga, a p’akaga-ad… “It is the thing, and the whole of the thing...”

Presumably P’akga means "It is the thing", and a p’akaga-ad means "and [it is] the whole of the thing". The former would mean that in Kad'k either:

1. There is some sort of copulative affix, like in Turkish. E.g. from Wikipedia:
   

   Öğretmen-i-m

   teacher-COPULA-1sg

   
   Interestingly enough, in Turkish the third person copula is unmarked, so:
   

   Öğretmen-Ø

   teacher-COPULA.3sg

   
   Which would make sense in Kad'k as well considering the number of phonemes in P’akga.
2. There is zero-derivation forming the verb "to be a <noun>" from a noun, with the third person unmarked.
3. 'Nouns' are actually verbs meaning "to be a <noun>", and the third person is unmarked. This would be fairly weird, but would fit with the fantasy-language theme.

As for a p’akaga-ad, I can see two ways to gloss this:

a

CONJ

p’akaga-ad

it.is.the.thing-INT

or

a

INT

p’akaga-ad

it.is.the.thing-CONJ

although I think that the first interpretation is more likely.

[bradrn]

The Fifth Elephant

• P’akga, a p’akaga-ad… “It is the thing, and the whole of the thing...”

Presumably P’akga means "It is the thing", and a p’akaga-ad means "and [it is] the whole of the thing". The former would mean that in Kad'k either:

1. There is some sort of copulative affix, like in Turkish. E.g. from Wikipedia:
   

   Öğretmen-i-m

   teacher-COPULA-1sg

   
   Interestingly enough, in Turkish the third person copula is unmarked, so:
   

   Öğretmen-Ø

   teacher-COPULA.3sg

   
   Which would make sense in Kad'k as well considering the number of phonemes in P’akga.
2. There is zero-derivation forming the verb "to be a <noun>" from a noun, with the third person unmarked.
3. 'Nouns' are actually verbs meaning "to be a <noun>", and the third person is unmarked. This would be fairly weird, but would fit with the fantasy-language theme.

As for a p’akaga-ad, I can see two ways to gloss this:

a

CONJ

p’akaga-ad

it.is.the.thing-INT

or

a

INT

p’akaga-ad

it.is.the.thing-CONJ

although I think that the first interpretation is more likely.

So I've been thinking some more about this last example, and it's turned out to be more interesting than I thought. The first clause uses p’akga, but the second uses p’akaga, so there's some sort of nonconcatenative morphology going on here. I suspect that the -a- is internal reduplication of the final vowel marking the intensive; cf. Drudak’ak “they do not get out in the fresh air enough”, presumably derived from *Drudk’ak “they do not go into the fresh air” via partial internal reduplication of the last vowel. In that case, either a or -ad means something different than I thought, since I previously assigned intensifier status to one of them; exactly what they mean is something I don't know yet, apart from suspecting that one may be a conjunctive.

[Pedant]

Thanks Pedant! How did you find all these?

Two to three hours' worth of sifting through the books...but by gods was it worth it. (Also the stories are never bad to go through; I confess to stopping and reading a few choice paragraphs now and again, which might have slowed me down a little.)

Feet of Clay

• H’druk g’har dWatch, Sh’rt’azs! H’h Angua tConstable...Angua g’har, b’hk barg’ra Sh’rt’azs Kad’k… “Welcome to the Watch, Corporal Smallbottom! This is Constable Angua...Angua, show Smallbottom how well you’re learning Dwarfish…”

This one's interesting... from the capitalisation, I think we get to see some grammatical markers! d- is either a definite marker or some sort of oblique argument, but I have no idea what t- is.

Personally I assumed it was some sort of sandhi variation on -d-, you know, for voiced and voiceless consonants respectively. Almost certainly it's the oblique construction form if they actually are interrelated, or referencing a subordinate clause (in other words h'druk g'har dWatch might read "we welcome you [indicating colleague], [we who are] the Watch," while h'h Angua tConstable could be read as "this is Angua, [who is] a constable").

The Fifth Elephant

• Ha’ak “not a real dwarf”

I thought this one was never actually translated in the book?

I'd have to check again, but I'm pretty sure it was said somewhere in TFE...

The Fifth Elephant

• P’akga, a p’akaga-ad… “It is the thing, and the whole of the thing...”

Presumably P’akga means "It is the thing", and a p’akaga-ad means "and [it is] the whole of the thing". The former would mean that in Kad'k either:

1. There is some sort of copulative affix, like in Turkish. E.g. from Wikipedia:
   

   Öğretmen-i-m

   teacher-COPULA-1sg

   
   Interestingly enough, in Turkish the third person copula is unmarked, so:
   

   Öğretmen-Ø

   teacher-COPULA.3sg

   
   Which would make sense in Kad'k as well considering the number of phonemes in P’akga.
2. There is zero-derivation forming the verb "to be a <noun>" from a noun, with the third person unmarked.
3. 'Nouns' are actually verbs meaning "to be a <noun>", and the third person is unmarked. This would be fairly weird, but would fit with the fantasy-language theme.

As for a p’akaga-ad, I can see two ways to gloss this:

a

CONJ

p’akaga-ad

it.is.the.thing-INT

or

a

INT

p’akaga-ad

it.is.the.thing-CONJ

although I think that the first interpretation is more likely.

I personally agree with your last posting on the subject--it does seem nonconcatenative rather than suffixing--but I wonder about the specifications in the other example you used. Vimes, in TFE, parsed out drudak'ak as "above ground...they negatively," and in spite of his relatively poor grasp of Dwarfish I'd imagine he was on to something (at the very least it's all we have to go on). I myself might try something like drudak'-ak as a base gloss, with a prefix, suffix, or infix indicating third person plural (or possibly some form of collective). Still, I like the idea of -ad being an intensifier, maybe referencing a holistic or "pure" interpretation.

Ooh, this is exciting!

[bradrn]

Thanks Pedant! How did you find all these?

Two to three hours' worth of sifting through the books...but by gods was it worth it. (Also the stories are never bad to go through; I confess to stopping and reading a few choice paragraphs now and again, which might have slowed me down a little.)

Sounds about right!

Feet of Clay

• H’druk g’har dWatch, Sh’rt’azs! H’h Angua tConstable...Angua g’har, b’hk barg’ra Sh’rt’azs Kad’k… “Welcome to the Watch, Corporal Smallbottom! This is Constable Angua...Angua, show Smallbottom how well you’re learning Dwarfish…”

This one's interesting... from the capitalisation, I think we get to see some grammatical markers! d- is either a definite marker or some sort of oblique argument, but I have no idea what t- is.

Personally I assumed it was some sort of sandhi variation on -d-, you know, for voiced and voiceless consonants respectively. Almost certainly it's the oblique construction form if they actually are interrelated, or referencing a subordinate clause (in other words h'druk g'har dWatch might read "we welcome you [indicating colleague], [we who are] the Watch," while h'h Angua tConstable could be read as "this is Angua, [who is] a constable").

I didn't even think about sandhi - I think that's very plausible.

The Fifth Elephant

• Ha’ak “not a real dwarf”

I thought this one was never actually translated in the book?

I'd have to check again, but I'm pretty sure it was said somewhere in TFE...

Checking my own copy, I can't find it. The reason I'm not sure is that I spent a lot of time getting distracted with the book itself, which slowed me down a little...

I personally agree with your last posting on the subject--it does seem nonconcatenative rather than suffixing--but I wonder about the specifications in the other example you used. Vimes, in TFE, parsed out drudak'ak as "above ground...they negatively," and in spite of his relatively poor grasp of Dwarfish I'd imagine he was on to something (at the very least it's all we have to go on). I myself might try something like drudak'-ak as a base gloss, with a prefix, suffix, or infix indicating third person plural (or possibly some form of collective). Still, I like the idea of -ad being an intensifier, maybe referencing a holistic or "pure" interpretation.

Actually, I don't think this is incompatible with my own suggestion. Potentially, you could have drudak'ak as the reduplication of drudk'ak, which in turn is composed of drudk' "they go above ground" + -ak "NEG". This works well with Vimes's attempted parse: it seems reasonable for a learner of Kad'k to figure out drudk' and -ak, but then get stuck on the a thrown into the middle.

(An interesting consequence of this is that you could have a distinction between drudak'ak (where the negative is applied first) and druduk'ak (where the reduplication is applied first), which opens up some subtle semantic distinctions.)

Ooh, this is exciting!

It is!

[bradrn]

Some more observations/theories:

• Kruk “dwarf mining law; laws of ownership, marriage laws, inheritance, rules for dealing with disputes of all kinds”
• Grag “dwarfish religious leader”
• Y’grad (noun--possibly a ceremony)

I can't believe I missed this before, but I think that at least the first two words here are related. Exactly how I'm not certain, but here's some theories:

1. Originally they were compounds *gra+uk and *gra+ag. Then simplify the vowel sequences to give gruk and grag. The original *gra could speculatively have meant "mineshaft" (cf. Jewish halacha or Islamic sharia, both literally "path" but also metaphorically extended to "law"), with *uk meaning "law" and *ag meaning "person". This theory has the advantage that it can also explain Y'grad by doing *gra-ad (and presumably y'- is some sort of prefix).
2. Again, consider compounds, but this time *rk+uX and *rg+ξaY, where X and Y are anything, and ξ is some voiced consonant which caused sandhi changing *rk > *rg. Then apply the intensive from my previous post (assuming it can double as an augmentative) to give *rukuX and *ragξaY. Applying a prefix k-/g- (possibly plural/collective/something derivational?) and dropping the final word of the compound (as an abbreviation?) gives kruk and grag respectively. Similarly to above, we could have *rk "mineshaft", *uX "law", *ξaY "person", meaning that kruk was originally derived from "AUG mineshaft-laws", and grag was originally "AUG mineshaft-people". I have my problems with this approach - particularly the ad hoc involvement of the augmentative, and the unexplained fact that k-/g- randomly changes voicing - but there are in fact reasons to believe this theory is correct (see below).

As you may be able to see, the main problem is figuring out why one word starts with k- but the other starts with g-. Any ideas?

• Kzad-bhat “head-banger”
• Zadkrdga “one who smelts”

From this I would postulate that zad means "head" originally, then metaphorically extended to "person" and thence to an indefinite pronoun. Note that this fits amazingly well with my Theory 2 above, where something of the form ξaY (with voiced ξ) was postulated to mean "person". (Oddly enough, Theory 2 was postulated independently of this - I worked out it had to be something of the form ξaY, and then realised the connection when I wrote this paragraph.)

[Pedant]

Some more observations/theories:

• Kruk “dwarf mining law; laws of ownership, marriage laws, inheritance, rules for dealing with disputes of all kinds”
• Grag “dwarfish religious leader”
• Y’grad (noun--possibly a ceremony)

I can't believe I missed this before, but I think that at least the first two words here are related. Exactly how I'm not certain, but here's some theories:

1. Originally they were compounds *gra+uk and *gra+ag. Then simplify the vowel sequences to give gruk and grag. The original *gra could speculatively have meant "mineshaft" (cf. Jewish halacha or Islamic sharia, both literally "path" but also metaphorically extended to "law"), with *uk meaning "law" and *ag meaning "person". This theory has the advantage that it can also explain Y'grad by doing *gra-ad (and presumably y'- is some sort of prefix).
2. Again, consider compounds, but this time *rk+uX and *rg+ξaY, where X and Y are anything, and ξ is some voiced consonant which caused sandhi changing *rk > *rg. Then apply the intensive from my previous post (assuming it can double as an augmentative) to give *rukuX and *ragξaY. Applying a prefix k-/g- (possibly plural/collective/something derivational?) and dropping the final word of the compound (as an abbreviation?) gives kruk and grag respectively. Similarly to above, we could have *rk "mineshaft", *uX "law", *ξaY "person", meaning that kruk was originally derived from "AUG mineshaft-laws", and grag was originally "AUG mineshaft-people". I have my problems with this approach - particularly the ad hoc involvement of the augmentative, and the unexplained fact that k-/g- randomly changes voicing - but there are in fact reasons to believe this theory is correct (see below).

As you may be able to see, the main problem is figuring out why one word starts with k- but the other starts with g-. Any ideas?

Your idea on voicing the consonant through an infix of some kind is an excellent one, and it adds so much more flexibility to the forms. As to the voiced and unvoiced onsets, I might suggest leaving out the augmentative particle and just relying on metathesis. I’ve combed through the available corpus, and there aren’t actually any words starting with r followed by a consonant, or even a glottal stop, it’s always followed by a vowel word-initially. So the initial forms may have been something like *rk-uk and *rk-ξ-ga/*rk-ξ-a-ga (thus rgag) respectively (see below for the -ga- suffix, but the latter word includes the intensive), and with, say, Second Convocation Dwarfish the liquid and stop switched places, giving us the modern words. A second similar metathesis might have happened with *rkξga; there is only one word in the modern Dwarfish corpus with more than three consonants in a row, zadkrdga, and I am absolutely certain in this case that the [r] is supposed to stand in for a vowel here. So another swap would make the transformation *rkξga>*rgξga>*grξga>grag, and voila! Or else just *rkξaga (similar to the -‘haga in jar’ahk’haga), to *rgaga to *rgag (no repeating syllables rule) to grag. (And then of course one could try applying the intensives at some other stop along the way, say rukuk-ad/karuk-ad “the whole of the law” or ragag-ad/garag-ad “the entire grag.”)
So, what do you think? Plausible?

• Kzad-bhat “head-banger”
• Zadkrdga “one who smelts”

From this I would postulate that zad means "head" originally, then metaphorically extended to "person" and thence to an indefinite pronoun. Note that this fits amazingly well with my Theory 2 above, where something of the form ξaY (with voiced ξ) was postulated to mean "person". (Oddly enough, Theory 2 was postulated independently of this - I worked out it had to be something of the form ξaY, and then realised the connection when I wrote this paragraph.)

See, I personally wondered whether or not it could be the other way around. Both zadkrdga “smelter” and jar’ahk’haga “ideas taster” contain the suffix -ga, which I thought might be used to refer to some form of habitual active participle or just a general agentive. (As above, I think it might be applied to grag as well, far enough back.)

[Neon Fox]

H’druk g’har dWatch, Sh’rt’azs! ... “Welcome to the Watch, Corporal Smallbottom!

Is...is it just me or is the Dwarfish for 'Smallbottom' basically 'short-arse'?

[Pedant]

H’druk g’har dWatch, Sh’rt’azs! ... “Welcome to the Watch, Corporal Smallbottom!

Is...is it just me or is the Dwarfish for 'Smallbottom' basically 'short-arse'?

...no comment on that at all.

[Salmoneus]

H’druk g’har dWatch, Sh’rt’azs! ... “Welcome to the Watch, Corporal Smallbottom!

Is...is it just me or is the Dwarfish for 'Smallbottom' basically 'short-arse'?

Yes, that's the joke.

It's important to notice a few things about Dwarfish:

a) it's all a joke
b) it's intentionally as weird and indecipherable as possible
c) the Dwarfish in the early books is clearly different in style (in particular, it starts off with almost no <a> (and few vowels at all, but plenty of <e>) and ends up with much more <a> and no <e>) than that in the later novels (an in-world explanation could be that Pterry is using a different transcription scheme for the same language).

[Salmoneus]

But, if you really want to do this...



...putting together a list of 16 words that we can be relatively confidant are nouns, some patterns leap out.

First, we have what we might call the b'zugdoids:
B’zugda
K’zakra
H’ragna
G’rakha

These all have the pattern: C'CVCCa

Next, there are the semi-b'zugdoids:
D'rkza
tra’ka
Dezka
Shatta

These share the -Ca ending. However, the first lacks a vowel, or (if the 'r' is a vowel) a second consonant. The second lacks the first apostrophe; we can probably treat the second apostrophe as a consonant. The third also lacks the first apostrophe, and the second consonant. The fourth likewise lacks the second consonant and the apostrophe.

But several of these can be brought into the paradigm with some assumptions. Let's assume:
- roots are either CCC (trifid) or CC (bifid). Surface vowels in the above words are manifestations of qualities of the consonant.
- b'zugdoids have the pattern C'CVC-Ca (trifids) or CVC-CA (bifids). The penultimate consonant is, let's say, a noun class marker, which can also have a role in derivation. The final -a is purely derivative. The first apostrophe is a sandhi effect marking some phonation difference on the first consonant due to the presence of the second.
- d'rkza is a typical b'zugdoid, provided we assume a "weak vowel", deleted (or not projected) next to the rhotic
- tra'ka is a typical b'zugdoid, provided we assume a dissimilation rule whereby (presumably) glottalised (etc) consonant are de-glottalised before a following glottal stop (or whatever ' marks)
- dezka and shatta are then regular bifid b'zugdoids.


But again, why the different vowels? Well, the -u- can easily be explained as a surface manifestation of labialisation of the following -g. Most dwarfish -u- seems to be adjacent to either a velar or a labial. But what about 'dezka'? Well, that's from Guards! Guards!. So far as I can see, all but one -e- in the corpus is from that one book, which also has other weird features (like, all the weird internal capitals).

Here's my radical suggestion: In G!G!, Carrot is speaking in an evolved Ankh-Morpork accent (and/or being transliterated differently). In standard, low (i.e. high) or traditional dwarfish, AM -e- is instead -a-. That makes 'dezka' perfectly typical, just a different version of Low Dwarfish *dazka.

------

Armed with this information, we can find another possible noun in the corpus: p'akga, presumably "thing". This can be analysed as a bifid b'zugdoid noun, provided that <p'> can be an independent root consonant, or, in my view more likely, as a trifid, with <'> as the second consonant, and the same dissimilation rule as for 'tra'ka'.


------

Next, we can find three more nouns with a seemingly-related form:
Jar’ahk’haga
zadkrdga
Bura’zak-ka

These all have the -Ca suffix (the hyphen may be to make k-k distinct from kk (occuring elsewhere), or to make clear the stress pattern). The first two both have specifically -ga, and both (ideas-taster and smelter) indicate animate agent nouns, so I'm happy with that -g- as a noun class marker. The last has -ka and is a place (town hall), which fits with *dazka above - a dezka-knik is a "mine supervisor", so *dazka might be a mine ('knik' being perhaps a noun, or perhaps something more abstract, like a positional ('one who is below the mine')). That leaves tra'ka, which we don't know the meaning of - some useless thing of little concern. I'd suggest, semi-randomly, that this too is a place, perhaps "dungheap" or "toilet", or indeed just "vein of pyrite" or the like.


So I'd suggest -ga as -anim.agnt.-NOUN, and -ka as -place.-NOUN.

Notably, these three, more complicated forms are more complicated concepts, perhaps more recently derived, so this is less a template (as with b'zugdoids) and more an overt suffixation.



-----

Next, we have the nouns "grag" and "kruk". These seem to not have any suffixation, but they do have otherwise trifid root form, so these look like pure root nouns.

------

Now, what about the noun "b'tduz"? That looks weird by comparison. It's tempting to write it off as a G!G! oddity. But wait!

Using our glottalisation sandhi rule, we can break this down into a trifid root, b-t-d, and a suffix -uz.

And lo and behold, that suffix crops up elsewhere, too. In fact, in two places that we don't know for sure are nouns, but certainly could be.

Angua says: "Grr’dukk d’buz-h’drak" - "small, delightful mining tool of a feminine nature". Look at that: d'buz.

And then we have the name, "K'ez'rek d'b'duz" - "go around the other side of the mountain. And there we have: d'b'duz.

That second apostrophe is unexplained if we tread it as sandhi. Maybe there is a DDD > D'D'D rule... or maybe the -d- is one of those noun class markers, and the root for mountain is actually d-b-'... that also makes sense in terms of derivation. Perhaps we could see d-b as meaning something like "rock" (a mining tool could be described as, for instance, "rock-ripper"), and then d-b-' as, for example "rock-heap" or "big rock", a derived root for "mountain".

CC(C)-Cuz is then an alternative nominal derivation paralleling CCV(C)-Ca. Perhaps a *d'bu'ga is a mountaineer, and a *dubza would be 'not really a rock'?


Actually, I'm going to revise my last suggestion, and propose that d-b is something like "spike, tooth". A mountain could easily be some derived form of that, and it could also be the name of a mining tool.

then, -uz can be a feminine noun marker (also found as an abstract, as in b'tduz, the name of a game), and the phrase means "spike(feminine)-for_mine".

--------

That gives us another noun: h'drak. This looks like an ordinary trifid root, h-d-r, plus the locative-class suffix -k. But now there's a different template: instead of citation form masculine CCVC-Ca (or feminine CCC-Cuz), we instead have CCC-ak (or perhaps CCC-uk? would a lawn-maintaining tool be a *d'buz-b'zgak, or a *d'buz-b'zguk?



--------------


Anyway, I want to leave nouns and b'zugdoids aside for a moment and pick up on something you've already hinted at: a form we might call "alphids". There's two of them:

p'akaga (extracted from 'a'p'akaga-ad)
G'daraka

The second of these is described as a 'state', but I'm going to suggest that the alphid template, CCaCaCa is actually an adjective. In the latter case, "the state of being free" could just be the adjective "free".

In the latter, case, I'm going to suggest that p'akga, 'a'p'akaga-ad is literally "thing, completely-is.thing-and". Adjectives being in some ways stative verbs, and the adjectival/verbal marking being required in the second clause, but not the first, in order to be able to use the verbal/adjectival completive affix 'a'-.




Anyway, just a few ideas to be going on with.

[bradrn]

Your idea on voicing the consonant through an infix of some kind is an excellent one, and it adds so much more flexibility to the forms. As to the voiced and unvoiced onsets, I might suggest leaving out the augmentative particle and just relying on metathesis. I’ve combed through the available corpus, and there aren’t actually any words starting with r followed by a consonant, or even a glottal stop, it’s always followed by a vowel word-initially. So the initial forms may have been something like *rk-uk and *rk-ξ-ga/*rk-ξ-a-ga (thus rgag) respectively (see below for the -ga- suffix, but the latter word includes the intensive), and with, say, Second Convocation Dwarfish the liquid and stop switched places, giving us the modern words. A second similar metathesis might have happened with *rkξga; there is only one word in the modern Dwarfish corpus with more than three consonants in a row, zadkrdga, and I am absolutely certain in this case that the [r] is supposed to stand in for a vowel here. So another swap would make the transformation *rkξga>*rgξga>*grξga>grag, and voila! Or else just *rkξaga (similar to the -‘haga in jar’ahk’haga), to *rgaga to *rgag (no repeating syllables rule) to grag. (And then of course one could try applying the intensives at some other stop along the way, say rukuk-ad/karuk-ad “the whole of the law” or ragag-ad/garag-ad “the entire grag.”)
So, what do you think? Plausible?

I think you've misinterpreted my argument a bit: *ξ was not meant to be an infix, but rather just the first phoneme of a word *ξaY, meaning "person". Then *rkξaY was a compound word "mineshaft+person".

On the other hand, I do think your theory is better than mine, in that it can explain the voicing of the initial consonant. I'm still struggling to understand it, but if I'm interpreting it correctly:

1. Start with original *rk-uk and *rg-ξ-ga
2. Metathesis of word-initial *rC to get *kruk and *grξga
3. I'm still not quite sure how you turn the second word here into grag? Could you explain?

I particularly like the metathesis; that adds a lot of flexibility into the system. I can easily imagine Kad'k having a constraint against initial rC, and using metathesis when it comes up.

(As for r, I've just been interpreting it as a syllabic consonant (cf. Slovak vŕba [ˈvr̩ːba]). Your 'vowel placeholder' idea seems intriguing, but I don't quite understand how that would have worked.)

From this I would postulate that zad means "head" originally, then metaphorically extended to "person" and thence to an indefinite pronoun. Note that this fits amazingly well with my Theory 2 above, where something of the form ξaY (with voiced ξ) was postulated to mean "person". (Oddly enough, Theory 2 was postulated independently of this - I worked out it had to be something of the form ξaY, and then realised the connection when I wrote this paragraph.)

See, I personally wondered whether or not it could be the other way around. Both zadkrdga “smelter” and jar’ahk’haga “ideas taster” contain the suffix -ga, which I thought might be used to refer to some form of habitual active participle or just a general agentive. (As above, I think it might be applied to grag as well, far enough back.)

Can't we combine both our ideas? If -ga is a derivational agentive, then zad-krd-ga would be "person-???-AGV"; then zad-krd could be a compound word where zad is "person" (as I said above) and krd is (say) "forge". Something similar could be happening with jar’ahk’haga.

[bradrn]

First, we have what we might call the b'zugdoids:
B’zugda
K’zakra
H’ragna
G’rakha

These all have the pattern: C'CVCCa

Next, there are the semi-b'zugdoids:
D'rkza
tra’ka
Dezka
Shatta

These share the -Ca ending. However, the first lacks a vowel, or (if the 'r' is a vowel) a second consonant. The second lacks the first apostrophe; we can probably treat the second apostrophe as a consonant. The third also lacks the first apostrophe, and the second consonant. The fourth likewise lacks the second consonant and the apostrophe.

But several of these can be brought into the paradigm with some assumptions. Let's assume:
- roots are either CCC (trifid) or CC (bifid). Surface vowels in the above words are manifestations of qualities of the consonant.
- b'zugdoids have the pattern C'CVC-Ca (trifids) or CVC-CA (bifids). The penultimate consonant is, let's say, a noun class marker, which can also have a role in derivation. The final -a is purely derivative. The first apostrophe is a sandhi effect marking some phonation difference on the first consonant due to the presence of the second.
- d'rkza is a typical b'zugdoid, provided we assume a "weak vowel", deleted (or not projected) next to the rhotic
- tra'ka is a typical b'zugdoid, provided we assume a dissimilation rule whereby (presumably) glottalised (etc) consonant are de-glottalised before a following glottal stop (or whatever ' marks)
- dezka and shatta are then regular bifid b'zugdoids.

Thoughts:

• Excellent catch with the b'zugdoids! I'm thinking now: Tolkien's Khuzdul is well known for being triconsonantal, so perhaps Kad'k is too?
• If we consider r to be syllabic, then d'rkza could well be a typical bifid b'zugdoid.
• I don't understand your dissimilation rule - could you explain it in a bit more detail?

But again, why the different vowels? Well, the -u- can easily be explained as a surface manifestation of labialisation of the following -g. Most dwarfish -u- seems to be adjacent to either a velar or a labial. But what about 'dezka'? Well, that's from Guards! Guards!. So far as I can see, all but one -e- in the corpus is from that one book, which also has other weird features (like, all the weird internal capitals).

Here's my radical suggestion: In G!G!, Carrot is speaking in an evolved Ankh-Morpork accent (and/or being transliterated differently). In standard, low (i.e. high) or traditional dwarfish, AM -e- is instead -a-. That makes 'dezka' perfectly typical, just a different version of Low Dwarfish *dazka.

That's hardly a radical suggestion - in fact, it's nearly canon already! In The Fifth Elephant, we already have several references to "Ankh-Morpok Street Dwarfish"; Carrot was raised outside Ankh-Morpok, but with his freakish ability to 'blend in' it's easily conceivable that even in G!G! he's speaking Street Dwarfish.

Armed with this information, we can find another possible noun in the corpus: p'akga, presumably "thing". This can be analysed as a bifid b'zugdoid noun, provided that <p'> can be an independent root consonant, or, in my view more likely, as a trifid, with <'> as the second consonant, and the same dissimilation rule as for 'tra'ka'.

------

Next, we can find three more nouns with a seemingly-related form:
Jar’ahk’haga
zadkrdga
Bura’zak-ka

These all have the -Ca suffix (the hyphen may be to make k-k distinct from kk (occuring elsewhere), or to make clear the stress pattern). The first two both have specifically -ga, and both (ideas-taster and smelter) indicate animate agent nouns, so I'm happy with that -g- as a noun class marker. The last has -ka and is a place (town hall), which fits with *dazka above - a dezka-knik is a "mine supervisor", so *dazka might be a mine ('knik' being perhaps a noun, or perhaps something more abstract, like a positional ('one who is below the mine')). That leaves tra'ka, which we don't know the meaning of - some useless thing of little concern. I'd suggest, semi-randomly, that this too is a place, perhaps "dungheap" or "toilet", or indeed just "vein of pyrite" or the like.


So I'd suggest -ga as -anim.agnt.-NOUN, and -ka as -place.-NOUN.

Notably, these three, more complicated forms are more complicated concepts, perhaps more recently derived, so this is less a template (as with b'zugdoids) and more an overt suffixation.

I'm a bit confused by your interpretation of -ga. As far as I can understand, it is an animate noun class marker, but also an agentive derivational suffix? Or does it indicate that the noun already has an agentive meaning, and is a pure animate.agentive noun class marker?

Also, how does the agentive fit in with p'akga, which we're fairly certain means "thing"?

Next, we have the nouns "grag" and "kruk". These seem to not have any suffixation, but they do have otherwise trifid root form, so these look like pure root nouns.

What's a 'pure root noun'?

EDIT: I think I understand now - you're just saying that they don't have a noun class suffix. I'm tempted to say that this is just a null suffix -Ø, representing yet another noun class.

Now, what about the noun "b'tduz"? That looks weird by comparison. It's tempting to write it off as a G!G! oddity. But wait!

Using our glottalisation sandhi rule, we can break this down into a trifid root, b-t-d, and a suffix -uz.

And lo and behold, that suffix crops up elsewhere, too. In fact, in two places that we don't know for sure are nouns, but certainly could be.

Angua says: "Grr’dukk d’buz-h’drak" - "small, delightful mining tool of a feminine nature". Look at that: d'buz.

And then we have the name, "K'ez'rek d'b'duz" - "go around the other side of the mountain. And there we have: d'b'duz.

That second apostrophe is unexplained if we tread it as sandhi. Maybe there is a DDD > D'D'D rule... or maybe the -d- is one of those noun class markers, and the root for mountain is actually d-b-'... that also makes sense in terms of derivation. Perhaps we could see d-b as meaning something like "rock" (a mining tool could be described as, for instance, "rock-ripper"), and then d-b-' as, for example "rock-heap" or "big rock", a derived root for "mountain".

CC(C)-Cuz is then an alternative nominal derivation paralleling CCV(C)-Ca. Perhaps a *d'bu'ga is a mountaineer, and a *dubza would be 'not really a rock'?


Actually, I'm going to revise my last suggestion, and propose that d-b is something like "spike, tooth". A mountain could easily be some derived form of that, and it could also be the name of a mining tool.

then, -uz can be a feminine noun marker (also found as an abstract, as in b'tduz, the name of a game), and the phrase means "spike(feminine)-for_mine".

Are you sure about the feminine? Recall that dwarfs don't really have a concept of gender. And while such things don't usually make it into language, I'd be surprised in this case if Kad'k has any sort of pervasive gender. (Although the "feminine nature" thing is convincing.)

That gives us another noun: h'drak. This looks like an ordinary trifid root, h-d-r, plus the locative-class suffix -k. But now there's a different template: instead of citation form masculine CCVC-Ca (or feminine CCC-Cuz), we instead have CCC-ak (or perhaps CCC-uk? would a lawn-maintaining tool be a *d'buz-b'zgak, or a *d'buz-b'zguk?

Looks plausible, especially if this is a bi-/triconsonantal system as I suggested above. Do you have any ideas on what the CCC-ak pattern could mean?

Anyway, I want to leave nouns and b'zugdoids aside for a moment and pick up on something you've already hinted at: a form we might call "alphids". There's two of them:

p'akaga (extracted from 'a'p'akaga-ad)
G'daraka

The second of these is described as a 'state', but I'm going to suggest that the alphid template, CCaCaCa is actually an adjective. In the latter case, "the state of being free" could just be the adjective "free".

In the latter, case, I'm going to suggest that p'akga, 'a'p'akaga-ad is literally "thing, completely-is.thing-and". Adjectives being in some ways stative verbs, and the adjectival/verbal marking being required in the second clause, but not the first, in order to be able to use the verbal/adjectival completive affix 'a'-.

You forgot one: drudak'ak, "they do not go into the fresh air enough". The first u could again be explained via labialisation, and potentially -'ak is another noun class. (A different type of agentive? plural, perhaps?)

Also, the spacing was originally p'akga, a p'akaga-ad, which seems to me more like a is the conjunction and -ad is the completive.

---------

Anyway, I'll attempt to summarize your theory here:

• Nouns are underlyingly bi- or triconsonantal (bifids/trifids respectively)
• The surface form of bifids are CVC-Xa (where Xa is noun class); trifids are C'CVC-Xa
• There is some sort of glottalic dissimilation rule - I don't understand the details
• The V in the surface form is either a (usually) or u (from labialisation)
• Two attested noun class markers: places -ka, animate agentives -ga
• Grag and kruk are 'pure root nouns', whatever those are
• Possible feminine noun class -uz?
• Another pattern CCC-ak, of uncertain meaning
• 'Alphids' of the form CCaCa-Xa derive adjectives

Looks much more plausible than my theory!

Now, seeing that we have a cohesive theory for the noun (as far as I can see), does anyone want to have a go at figuring out verbs or syntax?

[Pedant]

I think you've misinterpreted my argument a bit: *ξ was not meant to be an infix, but rather just the first phoneme of a word *ξaY, meaning "person". Then *rkξaY was a compound word "mineshaft+person".

On the other hand, I do think your theory is better than mine, in that it can explain the voicing of the initial consonant. I'm still struggling to understand it, but if I'm interpreting it correctly:

1. Start with original *rk-uk and *rg-ξ-ga
2. Metathesis of word-initial *rC to get *kruk and *grξga
3. I'm still not quite sure how you turn the second word here into grag? Could you explain?

I particularly like the metathesis; that adds a lot of flexibility into the system. I can easily imagine Kad'k having a constraint against initial rC, and using metathesis when it comes up.

(As for r, I've just been interpreting it as a syllabic consonant (cf. Slovak vŕba [ˈvr̩ːba]). Your 'vowel placeholder' idea seems intriguing, but I don't quite understand how that would have worked.)

You got most of it down pat. For the last bit: grag could be derived from two separate proto-forms, *rk-ξ-ga or *rk-a-ξ-ga (I should have mentioned an additional metathesis on the original), the latter using your intensive infix. Either one I might take to mean “mineshaft-AGENT” or possibly “mineshaft-DERIVATION-AGENT” for the ξ. The ξ voices the stop in *rk-, that’s no trouble, and it can be applied equally to both, and there didn’t seem to be any trouble with this. Next is where the two diverge:

• If we’re assuming the base form is *rgξga, then one first does the word-initial metathesis to get *grξga, then (because of an assumed no-more-than-three-consonants-in-a-row sound rule, assuming that the original r was not vocalic and that makes a distinction in sound changes) the ξ is substituted with a. We now have the form *graga.
• If we’re assuming the base form is *rgξaga (from *rk-a-ξ-ga to *rk-ξ-a-ga to *rg-ξ-aga), then my guess is that the ξ night be lost completely as opposed to becoming a vowel sound. We have the form *graga here as well.
• Either way, note a second assumed rule—no identical syllables appearing next to one another in the same word--and apply the rule accordingly, lopping the word-final -a off, leaving us with grag.

Also: looking back, I probably should have used the phrase “syllabic consonant” instead.

Can't we combine both our ideas? If -ga is a derivational agentive, then zad-krd-ga would be "person-???-AGV"; then zad-krd could be a compound word where zad is "person" (as I said above) and krd is (say) "forge". Something similar could be happening with jar’ahk’haga.

That...wouldn’t be impossible necessarily, just a matter of working out a suitable etymology for zad-krd including your meaning. Now, if zad is a root that means “head” then perhaps it could be extended to other things as well. There’s no specific mention of where dwarves think that thinking comes from, so let’s just assume they’re smarter than the local humans and understand it to be the head. The head is the instrument through which one categorizes thoughts and sensory perceptions--in other words, refining. *krd could then be said to perhaps mean “ore, impure lump of metal” or even might be a suffixed material concept (although that’s a little less than certain). So the proposed etymology would be something like “refine-ore-AGENT” or “refine-MATERIAL-AGENT,” whichever works better. Or there may be another one that makes even better sense, and if so I should be glad to hear it because my kzad is killing me...

[bradrn]

You got most of it down pat. For the last bit: grag could be derived from two separate proto-forms, *rk-ξ-ga or *rk-a-ξ-ga (I should have mentioned an additional metathesis on the original), the latter using your intensive infix. Either one I might take to mean “mineshaft-AGENT” or possibly “mineshaft-DERIVATION-AGENT” for the ξ. The ξ voices the stop in *rk-, that’s no trouble, and it can be applied equally to both, and there didn’t seem to be any trouble with this. Next is where the two diverge:

• If we’re assuming the base form is *rgξga, then one first does the word-initial metathesis to get *grξga, then (because of an assumed no-more-than-three-consonants-in-a-row sound rule, assuming that the original r was not vocalic and that makes a distinction in sound changes) the ξ is substituted with a. We now have the form *graga.
• If we’re assuming the base form is *rgξaga (from *rk-a-ξ-ga to *rk-ξ-a-ga to *rg-ξ-aga), then my guess is that the ξ night be lost completely as opposed to becoming a vowel sound. We have the form *graga here as well.
• Either way, note a second assumed rule—no identical syllables appearing next to one another in the same word--and apply the rule accordingly, lopping the word-final -a off, leaving us with grag.

Also: looking back, I probably should have used the phrase “syllabic consonant” instead.

I prefer using *rgξga - there's no need to randomly insert an intensive where it isn't present semantically. (And I realise I was the one who was originally guilty of this.) I also don't think you need a 'no identical syllables' rule - possibly there is a formal register of Kad'k which uses *graga, and attested grag is a near-universal colloquial abbreviation. I also think that a 'no-three-consonants' rule is implausible, considering the rest of the Kad'k corpus (containing zadkrdga, Jar'ahk'haga, d'hrarak etc.). One way to save this is to posit a 'no initial syllabic r' rule (as I did above), and assume compulsory metathesis when this does occur.

The other problem with going down this route is that it doesn't really fit with Salmoneus's observations. If we consider the fact that ordinary bifid b'zugdoids have a lemma of the form CVC-noun_class, then hypothesised *rk should really be something like *rak-ka, which doesn't play well with the sort of compounding we're doing. On the other hand, if we assume that *rk dates back to before Kad'k developed its semi-triconsonantal structure, then a variation of the process above could conceivably explain how that structure developed.

Can't we combine both our ideas? If -ga is a derivational agentive, then zad-krd-ga would be "person-???-AGV"; then zad-krd could be a compound word where zad is "person" (as I said above) and krd is (say) "forge". Something similar could be happening with jar’ahk’haga.

That...wouldn’t be impossible necessarily, just a matter of working out a suitable etymology for zad-krd including your meaning. Now, if zad is a root that means “head” then perhaps it could be extended to other things as well. There’s no specific mention of where dwarves think that thinking comes from, so let’s just assume they’re smarter than the local humans and understand it to be the head. The head is the instrument through which one categorizes thoughts and sensory perceptions--in other words, refining. *krd could then be said to perhaps mean “ore, impure lump of metal” or even might be a suffixed material concept (although that’s a little less than certain). So the proposed etymology would be something like “refine-ore-AGENT” or “refine-MATERIAL-AGENT,” whichever works better. Or there may be another one that makes even better sense, and if so I should be glad to hear it because my kzad is killing me...

I was suggesting something different to this, actually. I was imagining that zad "head" could be extended to mean "person", in which case zad-krd would be a compound word, with krd meaning "forge" (so an English calque of zadkrd could be "personforge"). Then add an agentive to get zadkrdga "one who person-forges" i.e. a smith.

On the other hand, I really like the think→refine metaphor! I might steal it sometime for my own use if I ever need it... (with your permission, of course.)

[Pedant]

I prefer using *rgξga - there's no need to randomly insert an intensive where it isn't present semantically. (And I realise I was the one who was originally guilty of this.) I also don't think you need a 'no identical syllables' rule - possibly there is a formal register of Kad'k which uses *graga, and attested grag is a near-universal colloquial abbreviation. I also think that a 'no-three-consonants' rule is implausible, considering the rest of the Kad'k corpus (containing zadkrdga, Jar'ahk'haga, d'hrarak etc.). One way to save this is to posit a 'no initial syllabic r' rule (as I did above), and assume compulsory metathesis when this does occur.

The other problem with going down this route is that it doesn't really fit with Salmoneus's observations. If we consider the fact that ordinary bifid b'zugdoids have a lemma of the form CVC-noun_class, then hypothesised *rk should really be something like *rak-ka, which doesn't play well with the sort of compounding we're doing. On the other hand, if we assume that *rk dates back to before Kad'k developed its semi-triconsonantal structure, then a variation of the process above could conceivably explain how that structure developed.

*rgξga it is then!
About the ≤3 Rule...I’d assumed, for example, that in a word like zadkrdga one could get away with the (frankly) terrifying cluster by having it be an example of strung-together words rather than morphemes (-krd- having a syllabic r), with jar’ahk’haga having a similar situation and d’hrarak’s schwa-plus-glottal-stop breaking up the worst consonant cluster in the lot and pushing it down to only two actual consonants in a cluster. But then who knows?
I’d be willing to drop the ‘no identical syllables rule’ as well; you’re quite right, swapping around one for the other is more than sufficient. (Maybe a sound change somewhere between Second Convocation and Shmaltzberger Dwarfish instead, then, removing word-final a before a -ξC- combination?)
I’m not honestly too sure about how to work the b’zugdoids into Kad’k, or whether they’re necessary at all. The language to my mind is more reminiscent of Georgian--heavy in consonant clusters, spoken in the mountains, and with relatively few vowels per word and a high degree of affixing derivational morphology (at last from what we’ve found in the few glosses available). While there does seem to be a consistent syllable structure, we know too little about the actual words--and have sufficient evidence compiled from the kruk/grag/y’grad and p’akga/p’akaga models to suggest that the derivation process is more concatenative than not. I’m equally uncertain about some of the suffixing, in particular the -uz suffix (which needs a bit of extra development, perhaps). Still, we’re all working on this together now; might as well see what we can safely come up with. I don’t mind being outvoted on the issue if it makes enough logical sense.

On the other hand, I really like the think→refine metaphor! I might steal it sometime for my own use if I ever need it... (with your permission, of course.)

Permission freely granted! Conlangers need to stick together, Pterrydactls/Pratchetteers equally so.

[bradrn]

I prefer using *rgξga - there's no need to randomly insert an intensive where it isn't present semantically. (And I realise I was the one who was originally guilty of this.) I also don't think you need a 'no identical syllables' rule - possibly there is a formal register of Kad'k which uses *graga, and attested grag is a near-universal colloquial abbreviation. I also think that a 'no-three-consonants' rule is implausible, considering the rest of the Kad'k corpus (containing zadkrdga, Jar'ahk'haga, d'hrarak etc.). One way to save this is to posit a 'no initial syllabic r' rule (as I did above), and assume compulsory metathesis when this does occur.

The other problem with going down this route is that it doesn't really fit with Salmoneus's observations. If we consider the fact that ordinary bifid b'zugdoids have a lemma of the form CVC-noun_class, then hypothesised *rk should really be something like *rak-ka, which doesn't play well with the sort of compounding we're doing. On the other hand, if we assume that *rk dates back to before Kad'k developed its semi-triconsonantal structure, then a variation of the process above could conceivably explain how that structure developed.

*rgξga it is then!
About the ≤3 Rule...I’d assumed, for example, that in a word like zadkrdga one could get away with the (frankly) terrifying cluster by having it be an example of strung-together words rather than morphemes (-krd- having a syllabic r), with jar’ahk’haga having a similar situation and d’hrarak’s schwa-plus-glottal-stop breaking up the worst consonant cluster in the lot and pushing it down to only two actual consonants in a cluster. But then who knows?

Well, that's certainly a way of saving the ≤3 Rule. I think it's simpler to simply get rid of that rule, but as you said, who knows?

I’m not honestly too sure about how to work the b’zugdoids into Kad’k, or whether they’re necessary at all.

I don't think they're necessary as such — it's just that the evidence strongly suggests their presence, and we might as well take that into account.

The language to my mind is more reminiscent of Georgian--heavy in consonant clusters, spoken in the mountains, and with relatively few vowels per word and a high degree of affixing derivational morphology (at last from what we’ve found in the few glosses available).

I’ve actually been thinking about it more like Semitic. Remember too that the Discworld is built off fantasy tropes, and the prototypical dwarfish language is Tolkien’s Khuzdul, which happens to be triconsonantal...

While there does seem to be a consistent syllable structure, we know too little about the actual words--and have sufficient evidence compiled from the kruk/grag/y’grad and p’akga/p’akaga models to suggest that the derivation process is more concatenative than not. I’m equally uncertain about some of the suffixing, in particular the -uz suffix (which needs a bit of extra development, perhaps). Still, we’re all working on this together now; might as well see what we can safely come up with. I don’t mind being outvoted on the issue if it makes enough logical sense.

Do you mean ‘non-concatenative’ rather than ‘concatenative’? And I know that there’s very little evidence, but I thought that the whole point of this exercise was to see how much we could deduce from what little information we have!

[bradrn]

I've found another sample of Kad’k which Pedant missed, from Maskerade, when the interpreter starts babbling in multiple languages:

Oh dear [...] Ai! Meu Deus! Dio Mio! O Goden! D’zuk f’t! Aagorahaa!

(Emphasis not from the original.)
If nothing else, this gives some more evidence to Salmoneus’s b’zugdoid theory; d’zuk seems to fit the pattern. As for a translation, I’m not familiar enough with the other languages here to give anything definitive; it would be nice to hear from anyone else who knows any of the above languages (which I think correspond to English, Portuguese, Italian, Dutch, and Trollish respectively).

[Pedant]

I've found another sample of Kad’k which Pedant missed, from Maskerade, when the interpreter starts babbling in multiple languages:

Oh dear [...] Ai! Meu Deus! Dio Mio! O Goden! D’zuk f’t! Aagorahaa!

(Emphasis not from the original.)
If nothing else, this gives some more evidence to Salmoneus’s b’zugdoid theory; d’zuk seems to fit the pattern. As for a translation, I’m not familiar enough with the other languages here to give anything definitive; it would be nice to hear from anyone else who knows any of the above languages (which I think correspond to English, Portuguese, Italian, Dutch, and Trollish respectively).

I must confess I’ve not yet read Maskerade; still missing most of the books from the Witches series. So thanks! If it’s quite alright with you, I think I’ll add it to the Big List...
Alright, alright, this does seem to follow something of a pattern. Perhaps not as cut-and-dry a pattern as Khuzdul, but a pattern nonetheless. The question is, what do the b’zugdoids actually mean?
As to a translation, that’s a little tricky. All the others (apart from Trollish and English) just mean “Oh God” or “My God,” so that should be fine. But we don’t really know much about Dwarvish spirituality beyond a journey through the underworld, Tak, and Agi Hammerthief. Perhaps a bit of worldbuilding might be necessary...

Ooh, as an added bit: remember this sentence?

• P’dan? K’raa! D’kraga “ha’ak”!

My reckoning is that d’kraga fulfills two of our previous constructions. The -ga agentive is easy enough, but then if d’- can be considered the same as d- and t- (in dWatch and tConstable (maybe from the Überwald dialect as opposed to Carrot’s Copperhead speak), then extrapolating a meaning of “come, arrive” for kra- we get a gloss like “This foreigner isn’t a real dwarf!” (Of course, kra- might have a more...colourful meaning, just a matter of checking for similar-sounding roots...)

[bradrn]

I've found another sample of Kad’k which Pedant missed, from Maskerade, when the interpreter starts babbling in multiple languages:

Oh dear [...] Ai! Meu Deus! Dio Mio! O Goden! D’zuk f’t! Aagorahaa!

(Emphasis not from the original.)
If nothing else, this gives some more evidence to Salmoneus’s b’zugdoid theory; d’zuk seems to fit the pattern. As for a translation, I’m not familiar enough with the other languages here to give anything definitive; it would be nice to hear from anyone else who knows any of the above languages (which I think correspond to English, Portuguese, Italian, Dutch, and Trollish respectively).

I must confess I’ve not yet read Maskerade; still missing most of the books from the Witches series. So thanks! If it’s quite alright with you, I think I’ll add it to the Big List...

That would be great — thank you!

Alright, alright, this does seem to follow something of a pattern. Perhaps not as cut-and-dry a pattern as Khuzdul, but a pattern nonetheless. The question is, what do the b’zugdoids actually mean?

I don’t think they actually mean anything in particular — I’ve been interpreting them as just noun conjugation pattern. Possibly the most common patterns C'CVC-Ca/CVC-CA are the unmarked ‘default’ conjugations for tri-/bifids, and the other patterns give different nuances of meaning? Hebrew already does something very similar with verbs, e.g. from the old board:

Hebrew has seven binyanim (singular binyan, literally "building"). Each one of them carries a certain meaning, and when a certain root is placed within that binyan (which has its own conjugations/derivational morphology), the binyan's meaning sort of becomes an overtone to the root...

I'll try to make that a bit clearer through examples further below.

These seven binyanim are commonly called pa'al, nif'al, pi'el, pu'al, hitpa'el, hif'il, and huf'al. The names come from the masculine singular 3rd person past form of the root *P-?-L "act, do" within each binyan.

Binyan Pa'al is the base binyan, a sort of catch-all with a wide array of semantic meanings. Binyan Nif'al represents verbs that are active (think active~stative distinction), inchoative, passive, reciprocal, and it also contains about 5 reflexive verbs. Hitpa'el consists of the same basic categories as Nif'al, except that the passive is extremely rare, and the active and reflexives are extremely common. Pi'el contains mostly agentives (and a few causatives, mostly along the lines of "cause to become") Hif'il consists mostly of causatives. Pu'al is the passive counterpart of pi'el, and huf'al of hif'il.

These categories are much stronger for more commonly-used verbs (or recent creations), while older verbs have a greater tendency toward drifting meaning-wise, especially when a binyan may completely collapse and disappear.

Now, let's give a few examples of roots across multiple binyanim (the form given is the citation form, the masculine singular 3rd person past, within the specified binyan).

K-T-V "write"
katav "write" (Binyan pa'al, as it is the root meaning)
nichtav "be written (Binyan nif'al, as it is passive)
hitkatev "correspond" (Binyan hitpa'el, more of a reciprocal)
hichtiv "dictate" (Binyan hif'il, "cause to write")
huchtav "be dictated" (Binyan huf'al, the passive form of hif'il)

H-L-CH "go, walk"
halach "go" (Pa'al, root meaning)
holich "lead, transport, conduct" (Hif'il, "cause to go"; the form is irregular because one aitch dropped through haplology)

CH-SH-V "think, consider"
chashav "think" (Pa'al, root meaning)
nechshav "be considered" (Nif'al, passive; /X/ has a tendency to screw up preceding vowels)
chishev "calculate, esteem" (Pi'el, agentive)
chushav "be calculated, esteemed" (Pu'al, passive pi'el)
hitchashev "consider, take into consideration" (Hitpa'el, active verb)
hechshiv "esteem, ascribe importance" (Hif'il, causative)

?-L-H "go up, rise, grow, excel, immigrate to Israel" (ever heard of making aliyah? Aliyah is the Pa'al Gerund of ?-L-H)
'alah "rise, go up, immigrate to Israel" (Pa'al, root meaning)
he'alah "raise, lift, promote, cause to immigrate to Israel" (Hif'il)
hu'alah "be raised, be promoted" (Huf'al, passive hif'il)
hit'alah "rise, be raised, exalt oneself, boast" (Hitpa'el, active)

?-L-M "vanish" (no pa'al form)
ne'elam "vanish, disappear" (Nif'al, active)
hit'alam "ignore, overlook" (Hitpa'el, active "making itself vanish")
he'elim "hide, conceal" (Hif'al, causative "cause to vanish")

Get it? You can also see how the meaning also drifts as different binyanim are applied to each root.

As to a translation, that’s a little tricky. All the others (apart from Trollish and English) just mean “Oh God” or “My God,” so that should be fine. But we don’t really know much about Dwarvish spirituality beyond a journey through the underworld, Tak, and Agi Hammerthief. Perhaps a bit of worldbuilding might be necessary...

I don't remember anything about a journey through the underworld.

And as for worldbuilding, I think the Discworld canon is expansive enough to find something that fits. Considering the English, maybe he was saying ‘Oh dear’ in all the other languages, and he wasn’t translating exactly.

Ooh, as an added bit: remember this sentence?

• P’dan? K’raa! D’kraga “ha’ak”!

My reckoning is that d’kraga fulfills two of our previous constructions. The -ga agentive is easy enough, but then if d’- can be considered the same as d- and t- (in dWatch and tConstable (maybe from the Überwald dialect as opposed to Carrot’s Copperhead speak), then extrapolating a meaning of “come, arrive” for kra- we get a gloss like “This foreigner isn’t a real dwarf!” (Of course, kra- might have a more...colourful meaning, just a matter of checking for similar-sounding roots...)

I'm not quite sure how you get that meaning? Could you maybe go through each word and explain? But I do think you're right about the composition of d’kraga.

[Pedant]

Alright, alright, this does seem to follow something of a pattern. Perhaps not as cut-and-dry a pattern as Khuzdul, but a pattern nonetheless. The question is, what do the b’zugdoids actually mean?

I don’t think they actually mean anything in particular — I’ve been interpreting them as just noun conjugation pattern. Possibly the most common patterns C'CVC-Ca/CVC-CA are the unmarked ‘default’ conjugations for tri-/bifids, and the other patterns give different nuances of meaning? Hebrew already does something very similar with verbs, e.g. from the old board:

Hebrew has seven binyanim (singular binyan, literally "building"). Each one of them carries a certain meaning, and when a certain root is placed within that binyan (which has its own conjugations/derivational morphology), the binyan's meaning sort of becomes an overtone to the root...

I'll try to make that a bit clearer through examples further below.

These seven binyanim are commonly called pa'al, nif'al, pi'el, pu'al, hitpa'el, hif'il, and huf'al. The names come from the masculine singular 3rd person past form of the root *P-?-L "act, do" within each binyan.

Binyan Pa'al is the base binyan, a sort of catch-all with a wide array of semantic meanings. Binyan Nif'al represents verbs that are active (think active~stative distinction), inchoative, passive, reciprocal, and it also contains about 5 reflexive verbs. Hitpa'el consists of the same basic categories as Nif'al, except that the passive is extremely rare, and the active and reflexives are extremely common. Pi'el contains mostly agentives (and a few causatives, mostly along the lines of "cause to become") Hif'il consists mostly of causatives. Pu'al is the passive counterpart of pi'el, and huf'al of hif'il.

These categories are much stronger for more commonly-used verbs (or recent creations), while older verbs have a greater tendency toward drifting meaning-wise, especially when a binyan may completely collapse and disappear.

Now, let's give a few examples of roots across multiple binyanim (the form given is the citation form, the masculine singular 3rd person past, within the specified binyan).

K-T-V "write"
katav "write" (Binyan pa'al, as it is the root meaning)
nichtav "be written (Binyan nif'al, as it is passive)
hitkatev "correspond" (Binyan hitpa'el, more of a reciprocal)
hichtiv "dictate" (Binyan hif'il, "cause to write")
huchtav "be dictated" (Binyan huf'al, the passive form of hif'il)

H-L-CH "go, walk"
halach "go" (Pa'al, root meaning)
holich "lead, transport, conduct" (Hif'il, "cause to go"; the form is irregular because one aitch dropped through haplology)

CH-SH-V "think, consider"
chashav "think" (Pa'al, root meaning)
nechshav "be considered" (Nif'al, passive; /X/ has a tendency to screw up preceding vowels)
chishev "calculate, esteem" (Pi'el, agentive)
chushav "be calculated, esteemed" (Pu'al, passive pi'el)
hitchashev "consider, take into consideration" (Hitpa'el, active verb)
hechshiv "esteem, ascribe importance" (Hif'il, causative)

?-L-H "go up, rise, grow, excel, immigrate to Israel" (ever heard of making aliyah? Aliyah is the Pa'al Gerund of ?-L-H)
'alah "rise, go up, immigrate to Israel" (Pa'al, root meaning)
he'alah "raise, lift, promote, cause to immigrate to Israel" (Hif'il)
hu'alah "be raised, be promoted" (Huf'al, passive hif'il)
hit'alah "rise, be raised, exalt oneself, boast" (Hitpa'el, active)

?-L-M "vanish" (no pa'al form)
ne'elam "vanish, disappear" (Nif'al, active)
hit'alam "ignore, overlook" (Hitpa'el, active "making itself vanish")
he'elim "hide, conceal" (Hif'al, causative "cause to vanish")

Get it? You can also see how the meaning also drifts as different binyanim are applied to each root.

I do indeed--I once tried to design a Semitic language for a friend (I don’t think he ever used it), so I am familiar with the concept. Ah, well, like I said, if it can be reasonably incorporated then let’s do so.

As to a translation, that’s a little tricky. All the others (apart from Trollish and English) just mean “Oh God” or “My God,” so that should be fine. But we don’t really know much about Dwarvish spirituality beyond a journey through the underworld, Tak, and Agi Hammerthief. Perhaps a bit of worldbuilding might be necessary...

I don't remember anything about a journey through the underworld.

And as for worldbuilding, I think the Discworld canon is expansive enough to find something that fits. Considering the English, maybe he was saying ‘Oh dear’ in all the other languages, and he wasn’t translating exactly.

I believe there’s a quote in Men at Arms that says something to that effect:

“Oh no,” said Carrot, “[that war-axe] is a burial weapon.”
“I should think it is!”
“I mean, it’s made to be buried with a dwarf. Every dwarf is buried with a weapon. You know? To take with him to...wherever he’s going.”
“But it’s fine workmanship! And it’s got an edge like--aargh,” Vimes sucked his finger, “like a razor.”
Carrot looked shocked. “Of course. It’d be no use facing them with an inferior weapon.
“What are you talking about?”
“Anything bad he encounters on his journey after death...it’s an ancient tradition.”
“I thought dwarfs didn’t believe in devils and demons and stuff like that.”
“That’s true, but...we’re not sure if they know.”

Ooh, as an added bit: remember this sentence?

My reckoning is that d’kraga fulfills two of our previous constructions. The -ga agentive is easy enough, but then if d’- can be considered the same as d- and t- (in dWatch and tConstable (maybe from the Überwald dialect as opposed to Carrot’s Copperhead speak), then extrapolating a meaning of “come, arrive” for kra- we get a gloss like “This foreigner isn’t a real dwarf!” (Of course, kra- might have a more...colourful meaning, just a matter of checking for similar-sounding roots...)

I'm not quite sure how you get that meaning? Could you maybe go through each word and explain? But I do think you're right about the composition of d’kraga.

Okay, so the first two words (B’dan? K’raa!) I have no real translation for. But the second two...well, we already have a decent definition for ha’ak, but d’kraga I parsed out above as “it is this [X-er] who is” or something to that effect. The -ga suffix we all seem to be in agreement on as some sort of agentive suffix, so that just leaves the main root, *kra. I took it to mean something like “come, arrive,” which in the agentive might mean “one who arrives, one who is defined by their arrival”--in other words, “foreigner.” This makes the whole sentence “It is the one who is defined by their arrival who is not a real dwarf!” But again, I could be wrong. I don’t have a proper definition for the root, after all...