Did you ever think the current method is less than optimal? I'm talking about the use of lerfu pro-sumti that refer back to the last sumti whose first letter is the same as the lerfu. It's probably the best we currently have, but it has some big flaws. Not only is it hard to keep track of the first letter of all the words that have been uttered, but it also happens that we'd like to refer back to, say, {lo mlatu}, but then before we can do so, another sumti that starts with M appeared in the meantime, so that {my} can no longer refer to the cat. The only way out is to repeat the entire sumti, which can be much longer than just {lo mlatu}. In any case, it's not reliable enough.

I have an idea  — and it's surely insane — that could put an end to this shortcoming.

The lerfu pro-sumti could be repurposed. In a big way.

Instead of having them refer to a previous word that begin with the corresponding letter, each lerfu would have a defined meaning, a type of thing that it refers to, reliably.  There would be one that refers back to the most recent living thing, the most recent person, object, abstract thing and many more. Below is a list of possible referents. I based the letters on gismu to make them easier to remember:

Let us look at some of the amazing things we can do with this.

    (1) na go'i .i mi pu no roi zvati sy
        "No. I have never been there."

    (2) fy dunda lo kargu junla ny
        "She gave him an expensive watch."

    (3) mi .e lo mensi be mi pu klama sy .i ku'i fy na nelci sy
        "Me and my sister went there, but she didn't like it."

    (4) ko xrugau dy lo pu te tolcri be fi do
        "Put it back where you found it."

    (5) my cinri sidbo
        "That is an interesting idea."

I can think of more things to put in the above table, this is just a proof of concept, for you to try out.

Though let me warn you. If you use this for a while, you might never want to go back. It's so powerful! It almost feels like cheating, that's how good it is. Suddenly, all the prior clumsy make-shift anaphora mechanisms are no longer needed, and what used to be a difficult area of the language has turned into a cakewalk.

But of course this whole idea is crazy.

In a recent post I talked about algorithmically created, gismu-like fu'ivla. For thos who thought that a word like kankazo would be hard to recognize (which I personally don't mind, as gismu are largely the same), the method described in this post might present a nice alternative.As I already said in the aforementioned blog post, I find CVC/CVCV and CCVCVCV the prettiest fu'ivla shapes (and I like very few others). If we look closely, we'll find that both shapes' first five letters are gismu-shaped:

    CVCCVCV
    CCVCVCV

followed by a lone CV in both cases. This gave me an idea.

If the first five letters form a gismu, then the old gismu-making algorithm can be used as is in order to create the first five letters of the fu'ivla. Then, and this is new, the trailing CV would be dependent on the kind of word that is being made; like the head of stage-3 fu'ivla, this CV would be taken from the related gismu (the first two letters for CVCCV gismu and the second and third for CCVCV gismu).
For example, say we want to make a new word for "frog". We first run the algorithm and get {tenca} (yes, I did run the algorithm). Then we add {-ba} (from {banfi}) and get our fu'ivla {tencaba}.
The first five letters give the word a (potentially) recognizeable sound (easier to remember for some), and the final CV tells the general type of the word, in the case of {-ba} it tells us that it's a banfi, an amphibian.
This method would be a hybrid between pure stage-3 (which I personally find quite ugly) and my previous idea for generated stage-4 fu'ivla.

Here is a short list of class suffixes:
    -ma     mabru         mammal
    -fi        finpe            fish
    -ba      banfi            amphibian
    -la       mlatu           cat
    -ri        tricu             tree
    -pa      spati             plant
    -ci        cipni           bird
    -ru       grute           fruit

and so on.

What do you think?

(note that CVCCVCV is prone to tosmabru, so it might be better to limit the fu'ivla of this system to CCVCVCV, which means that only CCVCV gismu candidates can be used)

Clearly, {lo} is the most common word by far in any (modern) text, but {cu}, {gi'e} and {cusku} also tend to be extremely common. I hope you find the clouds pretty, click on them to see a bigger image. Oh, and I think you can guess of which texts they are clouds without much difficulty (unless you've been living under a rock). :)
To get a fullscreen view of the images, click them, and then on the website, choose open in window and then maximize the window to fullscreen. Otherwise it can be hard to see some of the smaller (less common) words.

Imagine not knowing the definition of {banzu}, and being told that there is a gismu that means "enough/sufficient", what place structure would you intuitively expect it to have? Turns out not everyone has the same intuition here!

Personally I always thought {banzu} was a wee bit clumsy. The x1 can be anything, but it feels like sumti-raising most of the time. I tend to use a {ni} in banzu1:

    (1) lo ni mi certu cu banzu lo nu mi snada
        "The amount of [I am skilled] is enough for [I succeed] to happen"
        "I am skilled enough to succeed."

But there is also usage like:

    (2) lo xance cu banzu lo nu bevri
        "A hand is enough to carry it."

It's basically sumti-raising, since "having a hand" or "using a hand" is what is actually meant.

For a long time now, I have felt that {banzu} should parallel {sarcu}. {sarcu} is the "necessary condition", and with {banzu} there could be the "sufficient condition", which is a useful and necessary (no pun intended) word to have. The place structure would simply be:

    (a) x1 (nu) is a sufficient condition/is enough for x2 (nu) to happen under conditions x3

Everybody knows I like parallelism in Lojban, so such a definition would sit well with me. It's a pretty counterpart to {sarcu}.

    (3) lo nu mi sanga cu banzu lo nu mi gleki
        "Me singing is enough for me to be happy."    (also yet another case where nu/ka polymorphism would be useful)
   
This also adds an additional nuance to the Lojban conditionals:

    (4) banzu lo nu mi do fraxu kei fa lo nu do nupre lo ka na za'u re'u tai tarti
        "I will forgive you if only you promise me not to behave like that again."

On the other hand, latro'a, another Lojbanist, happened to not pay very close attention to the actual definition of {banzu}, and intuitively kept using a definition that they felt was appropriate. That definition was along the lines of:

    (b) x1 (object) is sufficiently/enough of x2 (ka/ni) for x3 to happen

In my opinion, this is also a useful definition! This definition is especially useful for lujvo, something which the actual {banzu} is terrible at, which I had to realize time and again upon trying to use it in lujvo. The natural tendency for me was to place {banzu} as the veljvo tertau, but then, where does the seltau go? The x1 is where it should go, but that isn't how lujvo tend to be formed. "Good enough" could be anything of {xagbanzu}, {bazyxau} or {selbazyxau}. In any case, it feels clumsy. This last definition makes making lujvo very easy, "good enough" is simply {banzu lo ka xamgu} -> {xagybanzu} (meanwhile I shed a tear over the fact that {banzu} has no short terminal rafsi).

The big point here is that I believe that both definitions, definition (a) and (b), are useful and should have a brivla. They can't both be {banzu}, however. Either we give one of them {banzu} and make a lujvo for the other, or we create a new gismu for one of them. I don't really have a preference for one option over the other, I just hope that both meanings end up with a nice brivla. Or rather: both definitions getting a brivla is a sufficient condition for my happiness! ;)

I don't have a good introduction, so I'm going to jump right in.
Borrowing words straight from Latin (or the Latin-Greek pseudo-language used in taxonomy), while probably more culturally neutral than a lazy borrowing from modern English, is still quite a euro-centric practice. Moreover, I find it a bit strange sometimes to use Linnean names for "everyday" animals or plants/fruit. Not to mention that in order to make them legal lojban words, they have to mangled quite a bit sometimes. For example, when the source word contains no consonant clusters, it's necessary to switch the order of some of the letters to artificially create clusters. Take the word for "birch" I recently made. The Linnean name is "Betula". I had to turn it into {.eptula}. Not too bad, but the above points apply; it's Latin (ish), but modified, to the degree that it becomes questionable if the relation to the Linnean source word is even helpful or obvious enough anymore (not that most people even know the Linnean names of things!).

So what I'm trying to say is that it might not be optimal to borrow such general words as those for animals from a single language, be it Latin or English. On the other hand, local concepts deserve to be borrowed directly from wherever they are local to, as there is usually only one language (or a group of related langauges) that even have a word for it. My idea for non-local words is to create fu'ivla using an algorithm much like the one used for making gismu. More on that later, first I want to talk about fu'ivla shapes.

I generally like type-4s much better than type-3s. Type-3s are too different from the rest of the native lojban words, they stick out too much. It's distracting, and the extra gismu head adds unnecessary syllables.
Type-4s, however, can be very pretty. There are lots of possible shapes, most of them aren't all that nice, but there are two in particular that really float my boat.

My favorites are fu'ivla and zi'evla of the shape CCVCVCV and CVC/CVCV, e.g. sorpeka or tceraso.

What is so great about them? They go well with the other native lojban brivla, like lujvo or gismu. They don't stick out like a sore thumb, but fit in harmoniously.

So, naturally, I would suggest using these forms over any other.

Let's create an example algorithmic fu'ivla that uses the words from the six most widely spoken languages. How about a word for "carrot". (yeah, I already made a fu'ivla for that recently: {ka'orta}, but it's based on Latin again, though I do like it more than {najgenja}).

So, carrot.

I just used some random dictionaries to find the translations below, and I don't have a way to verify their accuracy, so please don't be too harsh, this is just an example afterall.

Chinese: xulu'obo
Bengali: gaiara
English: karot (whether or not this should be karyt is really not important for this example)
Spanish: zana'oria
Russian: morkov
Arabic: iazarun

Using the (by now outdated) weightings...

Chinese 0.347
Hindi 0.196
English 0.160
Spanish 0.123
Russian 0.089
Arabic 0.085

We get lots of fu'ivla candidates, of which we only need to pick the best.

I haven't worked out the best scoring algorithm, but I think we could simply make some adjustments to the gismu algorithm and use that. For this example, I did everything by hand, so the result is not guaranteed to correspond to the one the algorithm would spit out. Let's say it tells us that the top three candidates are:

— kankazo
— grukoru
— ka'urko (whoops that one slipped in there by accident, it actually has a different shape, but came to me naturally, so maybe such a shape is also worth considering. This shape is CVVCCV, like ka'orta)

What's left is to decide is when to use CCVCVCV and when CVC/CVCV (and when CVVCCV). Maybe there are cases where one of them is better (has a higher score) than the other. Otherwise it would be an arbitrary choice which one to pick.

Anyway, I think I really like this idea. It sort of adds a second layer of "gismu" to the lexicon, except they aren't gismu and so can be easily recognized as not being part of the core vocab, which means they don't need to be learned as early. At the same time, they are as culturally neutral as possible, or at least as neutral as gismu, and much more neutral than Linnean borrowings. Plus, they feel like real Lojban words!