Conlang Random Thread

[Ares Land]

So I should probably go through the first dictionary and change everything for the thousandth time. But that dictionary now has 1183 entries, so I really don't wanna do it. *_* Have you ever had this problem, and what did you do?

First, hey, both versions look great!
I can't say I've had this problem but that's my lexicons live in ugly ass spreadsheets...

What's the original format for these dictionaries? Basically, I'd extract the raw data and convert it into tabular format with a good deal of string substitution..

[Ahzoh]

Been making some affixes from which verbs are derived

Code: Select all

ṣarar/ṣār "push"
ṣaras     "throw, say"
ṣaran     "poke, nudge"
ṣamur     "persuade, convince"
kaṣar     "pull"

Code: Select all

danin/dān "be heavy, press down"
danas     "crush"
kadin     "build, raise up"

Code: Select all

matut/māt "go by foot to"
matas     "run to"
manut     "journey to" > manāta "traveller, adventurer"

Code: Select all

ūś        "burn (tr.), heat up"
aśas      "scorch, destroy by fire"
aśan      "smoulder (intr.)"
amuś      "cook"

Code: Select all

maśuś/māś "drag, haul"
maśas     "grind up, shred"
manuś     "write"

[Ares Land]

A question on terminology.

So, Simbri has two tenses that are closely related in meaning:

Tense I covers: events you have personally witnessed ('I lived there for a while', 'he was an annoying kid'), past events you haven't witnessed, but you're reasonably sure of that occured within your life time ('Gorbachev was ousted by Elstin'), predictable future events (a total solar eclipse will occur next year, the elections will be held in November) , logical consequences ('If you sneeze wearing a face mask, you'll regret it').

Tense II covers: events you haven't personally witnessed
('he lived in Chicago for a while'), indirect speech ('he told me to get lost') relatively recent events you have some doubts about ('Kennedy was assassinated by the CIA'), anything historical or mythological ('In the beginning God created the heavens and the Earth', 'Julius Caesar was assassinated on the Ides of March'). Remote future predictions also belong there ('the sun will be eaten by a giant wolf', 'the Earth will be destroyed when the sun exhausts its hydrogen fuel'), general truths ('humans are mammals').

Basically, tense I prototypically covers events you've been a witness of, but the meaning extends to events you're sure of, and that you could have witnessed.

Tense II is protoypically a quotative: 'They say that...' which casts some doubt on recent events, but must be used for older events or mythology. (You can't have witnessed Caesar's assassination, or the creation of the universe).

The future meanings came about because of 1) a cyclical conception of time (so in a sense, the far future is the same as the remote past) 2) straightforward etymology (the tense I marker simply meant 'so' or 'then' in the protolanguage, which were then repurposed as a past tense, but the original meanings survived, tense II was used to mythology, then extended to eschatological stories and general truths)


I have doubts on how to label these: I've played with 'past' and 'indirect past' or 'past' and 'remote past', or even 'past' and 'reportative'. One issue is that none of these labels really fit. (They're mostly past tenses, but not always, tense II is usually a reportative, but sometimes these belong in tense I)

In fact, I'm tempted to label these 'preterit' and 'aorist'. It would suggest a perfective (which they are) past (which they are, most of the time), but the terms are vague enough not to bring any preconceptions. There are solid precedents, too, for using 'aorist' for verb forms that don't lend themselves to straightforward labels. But maybe these terms would just be confusing.

What do you think?

[Creyeditor]

Why not call them moods or evidentials and treat the tense information as implicatures?

[bradrn]

I agree with Creyeditor: these seem more like modals or evidentials. I’d call them realis (‘Tense I’) and irrealis (‘Tense II’).

[Ares Land]

Hmm, I think I'll go for 'past' and 'mediative' (of all the labels I found for this kind of verb form, this seems to be the best suited).

I think describing these as tenses is more clear, if not always entirely correct; Simbri has an optative and a subjunctive that are orthogonal to these...

[Creyeditor]

If you want to stay with calling them past tenses, I would suggest "direct (past) tense" and "indirect (past) tense". This just combines evidential labels with past tense. Evidentiality being connected to tense is a thing in Turkish, IIRC.

Yet a different term you could use for Tense 1 is "experiental past", although this would be more of a prototype-based label. You could contrast this with quotative (Tense II).

[Qwynegold]

Ah, I got more replies to this than I had expected.

The last time I was in this situation, I decided to bite the bullet and just throw it out. I dont want to discourage you though, because you must have put more work into yours than I put into mine, from the looks of it .... I had about the same number of words, perhaps even more, ... but the definitions were very simple and many were redundant (e.g. seven words for moon, six words for cat, etc.)

If there's only one thing about the dictionary that needs to be changed, e.g. the etymologies, ... then maybe you dont have to throw it all away. But from your post I cant see what it is about it you want to change.

Yikes! It's not really necessary to throw out the entire dictionary, it's mostly just going through every single entry and organize it differently. There's basically four major issues:

1. I need to decide once and for all how explanations of words should be added. In a specific document SIL seems to suggest that the type of explanations I use most commonly (e.g. "only used about things that hang down from something else, not when a part of something is hanging down from the rest of the entity", "Used when addressing a group of mixed gender. Does not take the acc. suffix, as ~ is inherently acc.") should be added in the \de field, which means that the explanation appears together with the lexeme's translation(s) and not on some separate line.
2. I should do a better job of separating different senses of the same word. For example, for one word I've given the translations "step forward; come forward; report in; sign in", but the first two of these translations should be treated as a different sense than the second two translations.
3. I should be more coherent when dealing with subentries. For example, in the picture I showed, medordìnarìa (know) is a subentry of ordìnarìa (get used to), because they're really just different forms of the same word. But then what should I write in the part of speech-field of the subentry? Or should I just skip that field entirely?
4. The way Lexique Pro works is that in the \ge field you fill in all the English translations, separating words with semicolons, e.g. "ge\ used; familiar". These are the words that appear in the alphabetically ordered English word list, so you should keep each translation in \ge as short and simple as possible. Then, in the \de field you write how you want the translation to be displayed when you're viewing that single lexeme, e.g. "\de get used to; become familiar". Often times I don't need the \de field to have any additions, in which case I can just leave it empty. But when I started the new, better dictionary for the other conlang, I started writing the entries the way SIL wants you to, which is e.g. "\ge hard; stiff | \de hard, stiff; a strong force". So you separate synonyms in the \de field with commas, while explanations are offset with a semicolon. But in the \ge field synonyms still have to be separeted with semicolons. So this means that when I want the \ge and \de fields to be essentially identical, I still have to write everything twice; once with semicolons and once with commas. And this is very annoying.

Nope, never had that problem. (I’ve never had a dictionary with even 100 entries, let alone 1000!) One option I can think of: if you know how to program, and if your dictionary software lets you, you could write something to convert each item in the dictionary to the new format. (This is particularly so if you’re using software from SIL, which tends to use the programmer-friendly MDF format.)

What's the original format for these dictionaries? Basically, I'd extract the raw data and convert it into tabular format with a good deal of string substitution..

It's Lexique Pro by SIL. There is a database file which I can simply open with Notepad. Something could probably be done for point 4 above. But I don't know how to program. T_T

So I should probably go through the first dictionary and change everything for the thousandth time. But that dictionary now has 1183 entries, so I really don't wanna do it. *_* Have you ever had this problem, and what did you do?

First, hey, both versions look great!
I can't say I've had this problem but that's my lexicons live in ugly ass spreadsheets...

Thanks!
You should probably look into some other way of storing your lexicon. Doesn't a spreadsheet really limit how much you can/want to write about each entry?

[Ares Land]

[*]I should be more coherent when dealing with subentries. For example, in the picture I showed, medordìnarìa (know) is a subentry of ordìnarìa (get used to), because they're really just different forms of the same word. But then what should I write in the part of speech-field of the subentry? Or should I just skip that field entirely?

I'd list both as 'verb' myself, really.

It's Lexique Pro by SIL. There is a database file which I can simply open with Notepad. Something could probably be done for point 4 above. But I don't know how to program. T_T

Cheap-ass solution: open that file with Excel or Open Office, using the pipe as a separator. Copy the /ge column. Do a search and replace: '/ge to /de', than another: semicolon to colon. Then you have your /de column.
Proper haughty Unix geek black magic solution: you can probably handle that with awk.

You should probably look into some other way of storing your lexicon. Doesn't a spreadsheet really limit how much you can/want to write about each entry?

Not really: I use 'comments' columns, and I can always add a new column if I need a bit of extra structure. I use Google Sheets so I can access these almost anywhere.
(At some point though I'll need a proper database, and converting that data will be horribly painful.)

[quinterbeck]

How would you transcribe labialised palatals? I'm not sure whether to use ʷ or ᶣ, e.g.:

nasal	ɲᶣ	ɲʷ
voiceless plosive	cᶣ	cʷ
voiced plosive	ɟᶣ	ɟʷ
prenasalised plosive	ⁿɟᶣ	ⁿɟʷ
voiceless fricative	çᶣ	çʷ
approximant	ɥ	ɥ

[jal]

How would you transcribe labialised palatals? I'm not sure whether to use ʷ or ᶣ

I'd say that if you have /ɥ/ but not /w/, it'd make sense to use ᶣ instead of ʷ. On the other hand, in phonemic description, it doesn't matter much, and you could go for more familiar ʷ.


JAL

[Ares Land]

Quick romanization poll:

I've got two 'long' vowels: one variously realized as [a:], [ɑ:], [ɑˤ], [ɔˤ]
Another realized as [o:], [o], [ u ] or even [oʊ]
(The variation is dialectal)

Which of these transcriptions seems most clear to you : au ou or â ô? (I'd rather not use a macron.)

[bradrn]

Quick romanization poll:

I've got two 'long' vowels: one variously realized as [a:], [ɑ:], [ɑˤ], [ɔˤ]
Another realized as [o:], [o], [ u ] or even [oʊ]
(The variation is dialectal)

Which of these transcriptions seems most clear to you : au ou or â ô? (I'd rather not use a macron.)

The latter, definitely. (Though what’s wrong with aa oo?)

[Ares Land]

Quick romanization poll:

I've got two 'long' vowels: one variously realized as [a:], [ɑ:], [ɑˤ], [ɔˤ]
Another realized as [o:], [o], [ u ] or even [oʊ]
(The variation is dialectal)

Which of these transcriptions seems most clear to you : au ou or â ô? (I'd rather not use a macron.)

The latter, definitely. (Though what’s wrong with aa oo?)

That'd be ambiguous with vowels in hiatus. oo would be two syllables [o.o]

[Ahzoh]

I made moods for Vrkhazhian.

Directive:

Code: Select all

NFUT nulaṣrasū (jussive, imperative, cohortative)
FUT  laṣrasū   (commissive)

Indicative:

Code: Select all

NFUT nuṣrasū
FUT  ṣarsū

Subjunctive:

Code: Select all

NFUT nuṣrassū (conditional)
FUT  ṣarassū  (optative)

Although now the verbs can look pretty tongue-twisty:
nulālulū : see\ACT.DIR
nūlulū : see\ACT
nūlullū : see\ACT.SUBJ

[Qwynegold]

[*]I should be more coherent when dealing with subentries. For example, in the picture I showed, medordìnarìa (know) is a subentry of ordìnarìa (get used to), because they're really just different forms of the same word. But then what should I write in the part of speech-field of the subentry? Or should I just skip that field entirely?

I'd list both as 'verb' myself, really.

Aha. Should there then be a note of some kind saying that "this is the progressive form of the verb" or something, so that the user isn't lead to believe that it's just a simple derivation? Or should I assume the user knows enough about the grammar to be able to tell that that's a conjugated verb?
I'm saying "user" as if it's going to be anyone else than me...

It's Lexique Pro by SIL. There is a database file which I can simply open with Notepad. Something could probably be done for point 4 above. But I don't know how to program. T_T

Cheap-ass solution: open that file with Excel or Open Office, using the pipe as a separator. Copy the /ge column. Do a search and replace: '/ge to /de', than another: semicolon to colon. Then you have your /de column.
Proper haughty Unix geek black magic solution: you can probably handle that with awk.

Unfortunately this doesn't work because everything is on a separate line, so I can't get the different fields to appear on separate columns. :/

[Qwynegold]

How would you transcribe labialised palatals? I'm not sure whether to use ʷ or ᶣ, e.g.:

ʷ can signify simultaneous labialization and velarization, so ᶣ would be less ambiguous. But almost everyone uses ʷ without caring about its velarization, and ʷ is a more familiar symbol, so...

Quick romanization poll:

I've got two 'long' vowels: one variously realized as [a:], [ɑ:], [ɑˤ], [ɔˤ]
Another realized as [o:], [o], [ u ] or even [oʊ]
(The variation is dialectal)

Which of these transcriptions seems most clear to you : au ou or â ô? (I'd rather not use a macron.)

The letters with circumflexes. Conlangers do sometimes use them for various diphthongs.

[Qwynegold]

I've been rewriting my dictionary for about a week now. I'm still at the letter A. The only automated thing I've done is using search-and-replace to change the parts of speech, from things like "V., intr" or "N., fem" into "V.intr" and "N.f".

Writing example sentences is the worst thing ever. People seem to like whimsical example sentences a lot, but mine are all really boring. Do you think this matters?

[Ares Land]

Writing example sentences is the worst thing ever. People seem to like whimsical example sentences a lot, but mine are all really boring. Do you think this matters?

Ah, yeah, they're a lot of work... A bit of originality helps I think. Sample sentences get a lot harder to make, but you get reusable vocabulary and constructions.
They're also less boring to write and they help sustain the reader's attention.

Whimsical sentences are really zompist's trick which a lot of us stole; but there are other options. For Old Hieratic Tarandim I use religious language, using both the Bible and the Popol Vuh as models, and it works really well.

[WeepingElf]

Writing example sentences is the worst thing ever. People seem to like whimsical example sentences a lot, but mine are all really boring. Do you think this matters?

Ah, yeah, they're a lot of work... A bit of originality helps I think. Sample sentences get a lot harder to make, but you get reusable vocabulary and constructions.
They're also less boring to write and they help sustain the reader's attention.

Whimsical sentences are really zompist's trick which a lot of us stole; but there are other options. For Old Hieratic Tarandim I use religious language, using both the Bible and the Popol Vuh as models, and it works really well.

If you look at sample texts in scholarly descriptions of natlangs, you will soon recognize that they are usually taken from actual speech acts such as conversations between native speakers or books written in the language discussed. This is because elicited examples are considered second-class evidence (they are likely to elicit an ideal model of the language that exists in the mind of the informant but is not really held up in actual speech acts) while constructed examples are not considered evidence at all (because they only show the describing linguist's own ideas of the language).

Many, if not most conlang grammars fall short at this: the examples clearly show that they are constructed. Of course, with a conlang, there are no "actual speech acts", but a good conlanger ought to be able to come up with something which gives the impression of one. (I have to admit that my grammar of Old Albic indeed contains examples that are obviously constructed, but these are meant to be replaced by ones drawn from texts I shall write in the language.)

That said, Zompist's Xurnese example about a sculptor hoping to govern a province one day is one of my favourite conlang text examples, as it tells us something about an important facet of Xurnese culture: the country is ruled by an art academy. Almost everywhere else, it would be a non sequitur: why should a sculptor expect to be appointed a province? But in Xurno, it makes sense!