This document aims to describe my (mia) rationale for the design of Eberban, both for its grammar and its vocabulary. Eberban is inspired a lot by Lojban and a bit by Toaq, which this document will explain what I liked and don't like and how it shaped Eberban. I'll also detail the train of thoughts for some of the unique aspects of Eberban.
While I looked a bit at lojban before, I really started to learn Lojban around the start of 2019. I really liked many of its ideas but started to be annoyed by many quirks that made it either hard to learn for me or annoying to use.
One thing I liked about Lojban was its phonology and orthography, which I reused for Eberban while changing a few details that I didn't liked. I liked a lot that a symbol as a one-to-one mapping with a range of accepted sounds, which simplifies spelling and avoids homophones (dear/deer in English). Most letters were easy for me to utter (my native language is French and I also speak English), however the consonant x (IPA: [x]) was hard to utter, especialy in clusters with r and l. In thus got rid of it in Eberban.
Regarding the vowels I found the letter y (IPA: [ə]) to cause some issues. In Lojban some consonant clusters are difficult to utter, and it is allowed to use a buffer vowel sound to ease their pronunciation. However it was hard for me to utter another vowel than the 6 Lojban vowels. As y in Lojban have more restricted uses than the other vowels that I didn't needed in Eberban, I got rid of y to allow it as a buffer vowel if needed (while Eberban also disallow some of the most difficult consonant clusters found in Lojban).
Lojban have specific vowel diphthongs, some which I find hard to distinguish with single vowels (mainly au from a). Those diphthongs also comes with some restrictions when making words with them, which for exemple makes borrowing my name as "la .mian." invalid (it must be borrowed as "la .miian." instead). In Eberban I decided to simplify vowel clusters by allowing any cluster of vowels which are uttered one after the other without pauses or glides (glides are still allowed for convenience but are alternative and optional pronunciation).
One general idea that I liked in Lojban is the idea of a Self Segregating Morphology (SSM), which is that the shape of words and some smaller elements are designed in such a way that a stream of sounds or characters can only be parsed in a single sequence of words, which is not the case in many languages like English (attack and a tack are usually homophonous). However there a few aspects of Lojban's SSM that I didn't liked, mainly its use of stress (I have prosody issues), and some rules being difficult for humans (such as the tosmabru problem). It also comes with a convoluted way of making compound words (lujvo in Lojban) that require remembering irregular short forms of words and stringing them together with sometimes the need for a y to separate consonants that are not allowed to follow each other.
In Eberban, no stress is used and word shapes are designed such that string any 2 words cannot form a forbidden cluster. Words don't have a different short form when making compounds, and instead the first word of the compound is prefixed with a vowel indicating the start of a compound and its length ("eberban" is a compound made from words "ber" and "ban", prefixed by "e" which starts a 2-words compound).
In both Lojban and Eberban, words are split into 2 large families: particles and content words, which can be recognized by their morphology (note that some particles can act as content words, but are still particles for specific reasons ).
Lojban handles foreign names and borrowings using 2 different constructs. Names (cmevla in Lojban) must be stressed correctly and necessarily end with a consonant (while all other Lojban words end with a vowel). Borrowings (fu'ivla in Lojban) are done differently and comes in different classes depending of which they contain a prefix or not, and are in my opinion difficult to tell appart from compound words.
In eberban, both are made with a single tool being borrowings, which starts with a u followed by the borrowed word and ended with a pause. This borrowed part have less constraints that native Eberban words to accomodate better for foreign words morphology (even if it still pretty limited), and can end with either a vowel or a consonant. If it needs to be used as a name, it is them prefixed with the word "za", which can also be used before native words.
In both Lojban and Eberban, particles are organized into families, such that grammar rules are expressed in term of families, each particle of a family conveying a different meaning while sharing the same structure.
A big issue for me in Lojban particle families is that particles of a same family most of the time don't look like each other. This is especially an issue as Lojban has a big number of particles, and that a few families exibit some patterns. Eberban on the other hand have less particles and particle families, and all particles in a family share a common prefix (all particles in the family named PA starts with pa, all particles in family VI starts with v, etc). In my opinion, this helps to learn and recognize them, at the cost of them being less distinguishable in high-noise environements. I however prefer to utter longer vowels and make slightly more efforts to articulate than to have trouble learning and recognizing particles (outside from very common particles, I still cannot read Lojban without using a parser or dictionnary to know which family particles are part of).
In Lojban there is a main verb and other parts of the sentence acts like nouns (either by being particles acting like nouns, or by being a verb transformed into a noun using a particle).
Eberban words either acts like verbs (also called predicates) or structure the sentence or text shape (all structuring words are particles, but some particles acts like verbs), and there is nothing that looks like a noun. Verbs are chained one after the other, and adjacent verbs share a common thing, like an invisible noun that is managed for you by the language.
| mian | etiansa | meon | zman | ||||
|---|---|---|---|---|---|---|---|
| There is some X which | is a cat | and which | eats some Y | which | is an apple | and which | is red |
Verbs in the chain are grouped in a right-first order to form new verbs. The
sentence above is grouped in order mian (etiansa (meon zman)), forming the
following intermediary verbs:
- meon zman: is a apple which is red
- etiansa meon zman: eats an apple which is red
- mian etiansa meon zman: is a cat which eats an apple which is red
Unlike many natural languages (natlangs), Eberban verbs don't have cases, but instead have ordered "places" (also called slots or parameters) for which the first 4 are commonly refered to using the vowels E, A, O and U in this order (the alphabetical order of vowels in Eberban is IEAOU, and all particles refering to those places will use the corresponding vowel, while I can be used alongside them to form some patterns). How verbs interacts with adjacent verbs depends on 2 concepts:
- Transitivity: the transitivity of a verbs determines which place will be used to interact with the chain of predicates on its right. For non-particle words, transitivity is encoded by the last letter of the word. If it is a vowel the verb is transitive and the A place will be used, otherwise it is intransitive and the E place will be used. Transitivity can be modified explicitly in a sentence using a particle of family SI.
- Type of the place: all places have a type which represents the kind of thing that must be used. It can either be a verb with itself its own typed places, or a non-verb which called an atom (which usually represent individuals or concepts). If the place selected by the transitivity rules is an atom place, then the atom is shared with the E place of the following chain (another place can also be selected using a particle in family SI). If the place selected is a verb place, then the following chain is stated to be equivalent with verb in this place. ("mi gali [mi jvin]" means "I am happy that [I dance]")
If one wants to make verbs interact in a different way from the default, particles in family VI and FI can be used to explicly state which place to use and how it related to the following chain. As VI and FI starts their own subchains, they can also be used to interact with multiple places of the same verb ("mi vani va [spua mi] fo [spua mo]" means "I go from [a home of mine] to [a home of yours]").
Eberban aims to encode Higher-Order Logic statements. Each verb can be seen as a function such that for every possible combinaison of inputs, it returns a trivalent truth value: either True, False or Unknown. A long term goal for Eberban is to have an interpreter such that given some facts (irrefutable truths, also called axioms) and a proposition, the interpreter could state if the proposition is provably true, false, or unable to prove.
Those functions are pure functions: the output depends solely on the input arguments and it can have no side effet. However this make a problem arise: how to express sentences like "In the future: [You dance]". How "dance" is able to know the time if its only parameter is "you"? One solution could be to have an additional parameter for the time, but that would require to explicitly provide the time to every verb that depends on it, and would require to have a place for many other concepts; which are both not ergonomic.
The strategy choosen for Eberban is to have an hidden parameter called the context parameter which is automatically managed by the language. Each verb will automatically provide the message it receives to verbs it depends on (the following chain and any verbs that are used in its definition). Additionally, the language contain 2 special verbs mue and mua that allow respectively to extract the current context and evaluate a verb with an arbitrary context. Thanks to those verbs, vocabulary expressing time can modify the context it provides to following chain to encode the time relation. To encode concepts like the time passing by between sentences, the language have grammar to register context changes that will be applied between each sentences.
The context parameter is a powerful tool that can be used to implement many concept that "just work" in other languages like tenses, and to design ergonomic vocabulary that uses the context parameter to "carry information around" to reduce verbosity.