Teams that solve the The Artifact meta begin to receive artifacts written in Perflontus (puflantu), which is used in the four main round metas as well as the metameta. The artifacts are intended to give teams known or deducible texts against which to begin learning how to translate Perflontus. What follows is a summary of language features, starting from the most basic. For puflantu vocabulary, see here.
For the remainder of this write-up, and in other write-ups where it becomes relevant, Perflontus text will be written in monospace.
Design notes: Perflontus takes a lot of design cues from Chinese, Japanese, Korean, and Spanish (in some order), along with a few etymological decisions from Russian. I felt that German would seep a bit into the language regardless of what we wanted through its relationship with English. I tried to refer to the Asian languages, both because two of them are SOV and because I wanted to give native English speakers a little less of a "home" advantage in picking up Puflantu.
Mouth Sounds and Word Composition
Perflontus uses the English alphabet plus apostrophe (henceforth '). Most of the sounds are what a Spanish speaker would expect them to be. More details are available in the Artists meta solution.
Syllables in Perflontus are always contain a single vowel, which may be preceded by up to one consonant and followed by up to one consonant. This implies that consonant clusters of 3 or more do not exist in Perflontus, that initial consonants are never clustered, and that whenever two vowels are adjacent there is a syllable break between them. For purposes of most of the language rules, ' is a consonant that is always a syllable unto itself. 1
Stress in Perflontus always falls on the vowel preceding a word's final consonant cluster. 2 If no such vowel exists, it falls on the word's first syllable instead.
Nouns and the things that modify them always end in vowels. Verbs and adverbs 3 always end in consonants. To this end the suffix -s is frequently used to adverbialize words, including at the end of particles like -li(s) and -ka(s).
Design notes: A lot of the details here were nailed down early. The phoneme inventory intentionally excludes /θ/ and /ð/, and includes /ɲ/ and /ŋ/, in an effort to sound less English. The bell was originally pitched as a funny inclusion for the FAQ ("We'll add 'a bell' to the recommended materials with no further explanation") and took a life of its own from there. Early versions of our consonant chart include "Manual" row as a manner of articulation just for this.
Perflontus is head-final and uses a subject-object-verb (SOV) sentence structure 4 . Commas are used to set off long phrases. 5 Taking cues from other SOV languages, we tried to place adverbs between the object and the verb as often as possible.
I eat cookies.
cookie-PL eat-1S 6
The particle -ka(s) is used at the end of verbs to indicate the end of a dependent clause. The role of the modified word within the dependent clause is indicated by the use of the relative pronouns al and ol; the former later expanded its role to include adverbs used in the dependent clause, e.g. albwr "at that time" from al- + bwri "time". Such an adverb, if it exists, is placed at the beginning of the dependent clause instead of as close to the verb as possible.
I eat when I'm hungry.
that-time hungry be-1S-DEP-ADV eat-1S
Albwr ruzeqo zwm-kas wlwmoc.
Design notes: Properly handling dependent clauses was easily the most time-consuming part of design, since it's the part of syntax that really lets you build sentences with normal amounts of complexity. (A real linguist might say something here about generative grammar.) It's not really obvious how heavily we rely on these things until one tries to go without them (including constructs like rephrasing "with X" as "that possesses X").
Ultimately the structure I went with is very Englishy, and in a lot of cases -ka(s) can be understood as "that" or "which" in a first pass (as opposed to, say, Chinese, which uses a genitive particle to mark dependent clauses as a property that possesses the modified word). The placement of referring adverbs like albwr and alhan at the start of dependent clauses happened pretty early and I didn't realize it was inconsistent with our other adverb placement until it was already everywhere. At least it's inconsistent in a systematic way?
Perflontus inflects its verbs in a variety of ways, as follows:
- Proclitic particles for tense (future/past), aspect (imperfective/perfective), and mood (hypothetical/counterfactual).
- Prefixes for reversal (which is distinct from negation), imperative mood, augmentation, and diminution. (The latter two also migrated to being valid nounjective prefixes.)
- Up to three infixes for the subject pronoun, negation, and object pronoun, in that order, placed before the root verb's final vowel.
- Nounjectivifying 7 suffixes for gerund/participle/infinitive (mixing), agent (mixer/mix-capable), patient (mixee/mixable), locative (mix-hosting), instrument (mix-enabling), causative (mix-causing), and resultative (mixture).
Each verb has a root form that is never used alone; one of the last three inflections is always applied. To this end the root form is formally notated with an asterisk indicating the infix position (e.g. wl*oc), though we ended up only having one irregular verb where this notation was a useful reminder (to be, az / z*).
Because verbs necessarily have personal pronouns infixed in them, the pronouns are often elided, except when they are desired for emphasis or clarity.
I eat cookies.
1S cookie-PL eat-1S
Wm torelwe wlwmoc.
1S cookie-PL eat
Wm torelwe wloc.
Cookies are eaten.
Two of the verbs (and possibly more) are "reversed" from what an English speaker would expect. This feature is borrowed from Spanish, just because. The verbs that appeared in artifacts were som*un "to please" (i.e. "gustar") and pel*ig "to worry". That is, one says "X pleases me" instead of "I like X", and "X worries me" instead of "I worry about X".
The affixes/clitics are as follows. If multiple affixes with the same position are used, they appear in the order listed in the table:
|Tense-aspect-mood proclitics (if multiple are used, they appear as one word)|
|im||past tense (PST)||im wl*oc||ate|
|et||future tense (FUT)||et wl*oc||will eat|
|av||imperfective aspect (IMPF)||av wl*oc||is eating|
|os||perfective aspect (PRF)||os wl*oc||has eaten|
|fel||hypothetical mood (HYP)||fel wl*oc||might eat|
|gaq||counterfactual mood (CNT)||gaq wl*oc||would eat|
|'||imperative mood (IMP)||
|leave (request/instruction) (explicit 2P subject = command)|
|ag||augmentative (AUG)||agj*uf||overcook (j*uf cook)|
|sa||diminutive (DIM)||sam*id||glance (m*id see)|
|vo||reverse (REV)||vol*ay||return (l*ay leave)|
|ey||negative (NEG)||mer*ey*in||cannot (mer*in to be able to)|
|Suffixes (Verb → Nounjective)|
|a||gerund/participle/infinitive (GER)||wl*oca||eating / to eat|
|afe||agent (AGT)||wl*ocafe||eater / eat-capable|
|who||patient (PAT)||wl*ocwho||eatee / edible|
|ice||locative (LOC) 8||wl*ocice||place where (one) eats|
|ede||causative (CAU) 9||wl*ocede||eat-causing|
|aqo||instrumentative (INS)||wl*ocaqo||for eating|
Nouns in Perflontus also double as adjectives. For those things that in English would clearly be one or the other, it is up to the listener/reader to determine the sense of the corresponding type. For example we used furwnwe as "fluffy (plural)" in the Bakeoff but as "clouds" in the "Dumb" lyrics clip.
Nouns take relatively few additional morphemes; the most frequent are the -w and -we suffixes, which replace the noun's final vowel when using the dual and plural numbers respectively. In addition, nounjectives can be negated by infixing -ay- prior to their last vowel, similar to how it's done with verbs.
The ag-, sa-, and vo- prefixes also retain the same meanings as when they are applied to verbs, but their usage on nounjectives is more a matter of forming new root words than as generally applicable prefixes.
The rules surrounding ' are a little more complicated than this due to other language rules/constraints, but this is close enough.
This may not have been honored as much in the Artists meta as I'd have liked, since I was more preoccupied with speaking slowly without being completely monotone.
For a very loose definition of "adverb". There are definitely adjective-modifying words that didn't end up getting the adverbial transformation because I felt it would be too confusing.
This was rigidly imposed for a lot of artifact creation but started slipping as our adverbial phrases got longer and more complex (e.g. by allowing those adverbial phrases to come after the verbs they modify, instead of before).
This compensates somewhat for the lack of syntax-marking particles, as in Japanese or Korean. I didn't start doing this much until later in the translation process, when I felt my fundamentals and vision for the language were clear enough that I could afford to be slack in natural ways.
Our translation process made extensive use of glosses, which represent a sort of halfway translation point where all the syntax of a sentence has been converted to the target language, but none of the vocabulary has. This makes it easier for us to check that grammar is correct without having to memorize the lexicon, and it also allows us to search over glossed text when a word or grammatical feature needs to be replaced. Here the -PL marker stands for "plural", and the -1S stands for "first person subject".
Iqa ehyo natho aza merateyin.
We ended up not using this one as much as we could have, since it's somewhat ambiguous and hard to come up with example for. In retrospect it might have been good to extend this suffix to nouns and replace the X henqo(s) "X interior" pattern. A more organic example might be something like Rwmo rali qekadice et zat "My house will be the meeting-place."
Another suffix that got neglected; we could have used wlocede "appetite" or zumuzede "soporific".