[My thoughts on Auton, years later. Largely from an email exchange with Jeremy P Newton, later modified for clarity] "Jeremy P Newton" writes: > Hello, I'd like to learn a bit more about your design of AllNoun. In > particular, I'd like to know what difficulties you encountered, what > features of other conlangs you decided to use or not use, and what you > think could use some improvement. Where to start? I'm going to answer some of your later questions in this stream of text, because it's just easier to organize my answer this way. The original impetus was to design a language that used only one part of speech. It originally came out of some conlang discussion of Glosa. However, it does not borrow any Glosa features. Glosa is widely considered to depend on knowledge of English. Ie, although Glosa words have no official "category", they behave differently in different positions, and intuitively mirror English syntax. I decided that AllNoun should do the one-part-of-speech thing properly. The semantics of a word do not depend on where it is used. (To be very correct, I should say "...of a word-sense...". Word-sense disambiguation was never contemplated.) All words are nominals, and relate to each other only thru the explicit semantics "(", ":", "^", and ")". They do not modify other words, take them as subjects or complements, etc. They specifically don't participate in noun-noun modifiers, and they have no inflection, which makes common nouns all mass nouns. Making it consistent to begin with was pretty easy, actually. Keeping that while making it powerful enough to function as language and making it easy to use were far bigger challenges, which I did not entirely succeed at. One similar conlang, which I did not know of at the time, is Paul Doudna's "Xaq", based on the predicate calculus. He kindly sent me his paper on it, and I found it quite interesting. Problems? My treatment of adjectives is the big one. It treats subsective and intersective adjectives well enough, but its intersective mechanism is not sufficient for nonsective adjectives. For instance, a "former friend" is not any sort of friend, and not easily seen as a member of "the set of all former things". "Alleged thief" is not the intersection of all thieves and all "alleged things" Another difficulty, which Paul Doudna pointed out to me, is that it is unclear how propositions are to be expressed. At the time, I believed that a top-level nominal should be interpreted as existential. Eg, to say "I ate the apple" you'd express "an eating of the apple by me in the past", and it would be understood as existential "(There is) an eating of the apple by me in the past". However, it isn't as expressive as I would like. Paul also suggested it was weak at expressing fictional contexts, which aren't really existential, but I'm not sure it's any worse than natural language. Non-declarative moods (questions, imperatives) worked clumsily, by using the declarative mood. Expressing determinacy ("the", "an", "some") was always clumsy. Determiners more than any other natural words want to modify the nominal adjacent to them, which is totally contrary to AllNoun. So I ended up with a lot of examples that simply did not express determinacy. Relative clauses, while they worked, tended to not be linear. Eg, you could express "apple *that I ate*", but AllNoun did not neccessarily make the relation of the relative clause to the matrix clause immediately clear. Eg, one might say: (apple (eat agent:me time:past patient:^)). ...where it isn't clear until the end of the relative clause that it is about the apple. Of course it's not neccessarily so: (apple (eat patient:^ agent:me time:past)) Importing foreign words, quoted speech, and sometimes names tended to be messy, as they tended to not be uninflected nominals. Improvement? I wouldn't, I'd start fresh. Since I designed AllNoun, I have learned a lot about the deep structure of natural language (mostly English). Now I'd be more interested in regularizing natural grammar than constructing a grammar. One thing that AllNoun gave me was an appreciation for the central function of prepositions and other specifiers in natural language, a function that I believe is underappreciated by linguists. > (Just out of curiosity, have you heard of the conlang "Taxy-Turvy"?) I had not heard of it. > While reading the grammar, I had some nagging suspicions about > consistency. That is, what complicated meanings or idioms are > hiding in those English glosses? Well, honestly, I don't think consistency as such was a problem. The basic relations and atoms were very simple, and the rules were few. The consistency tended to be quite constricting, tho. I did find that I tended to import adjectives as meaning "ADJECTIVE thing(s)", and prepositions as meaning "thing(s) that something is PREPOSITION", eg "on" => "thing(s) that something is on" = "support". I don't think either method of borrowing violated consistency, really, but they always felt a little strange. And I always feared they'd confuse users. > The : also seems to inherit some > of the many different uses of the possesive and "of". Yes. In a philosophical sense, I believe that relatedness is basic in communication. Any communicator in any language has to, at some basic underlying level, recognize a relation just because it is named, without additional underlying mechanism. > Are there > some changes one could make that would prevent the AllNoun user > from just creating equivalents to the usual English parts of speech > without really considering the deeper meaning of the sentence? I don't believe there are. I noticed some tendency for people to create extra-syntactic modifier relationships, which of course is ubiquitous in natural language. There wasn't much I could do except try to hook them on the vision of simplicity.