[YG Conlang Archives] > [engelang group] > messages [Date Index] [Thread Index] >
Jorge Llamb�as jjllambias@hidden.email [engelang], On 29/09/2014 19:32:
On Mon, Sep 29, 2014 at 10:10 AM, And Rosta and.rosta@hidden.email <mailto:and.rosta@hidden.email> [engelang] <engelang@yahoogroups.com <mailto:engelang@yahoogroups.com>> wrote:
3 is something I've given a lot of thought to for many years, and it's not clear to me that we've found a good enough solution or indeed that a good enough solution exists to be found. All known solutions look to overload working memory in various ways. The problem could do with more discussion, and with me getting round to documenting my loglang's solution; that might help us grope our way to a solution.
If I remember right, your idea was that instead of variables, you
would have predicate modifiers, somewhat similar to Lojban's SE, but
much more powerful, since they could not only re-order argument
places but also conflate them, and they didn't act only on atomic
predicates but they could act on any formula. Is that more or less
right? Do you have some estimate of how many of these modifiers would
be required to make things work?
Let me essay a fuller but still short and simplified account. All words are predicates (except magic zoi-like words). Each predicate is inflected to show which of its argument places are saturated implicitly and which explicitly (by merger with another argument place (of another predicate)). For the first word of the sentence, all explicitly saturated argument places (ESAPs) are added to a virtual list. For the next word, for each of the word's ESAPs the word is inflected to (a) show which ESAP on the virtual list it merges with, with the merged ESAP removed from the virtual list, or (b) show which ESAP on the virtual list it merges with, with the merged ESAP left on the virtual list, or (c) show the ESAP does not merge with any ESAP on the virtual list, and is added to the list And so on, iteratively, for each next word. The sentence ends when the virtual list is empty. A complication is that besides intersective merger (for a things that is red and is a door, say) I also have joi-like additive merger (for a thing that consists of xorxes and And, say). The load on working memory comes from the virtual list, an ordered list of all variables that have appeared in the sentence and will reappear later. From brainstorming with Alex and Sai, I thought of some ameliorations if not outright solutions, such as resorting to giving names to variables that have to be retained over a long span of text. Another idea was using a computer to discover the optimal word order to keep the length of the virtual list within reasonable limits. On my to-do list is a rendition of the North Wind and the Sun into Livagianoid logical form, which I hope Alex might work out the optimal word order for, so we can get a sense of how different that is from the starting word-order. A further unergonomic feature of the system is that most of the info in a sentence is in the inflections, not in the predicates (i.e. stems), though I haven't yet been able to assess how extreme the imbalance between stem and inflection is. --And.