This message was posted by Felix Wan in the old forum, I think it is important enough to be remembered

Noun Classes,Sunday 28-Jan-2001 05:55:47

Vorlin roots are mostly nouns convertible to other parts of speech through suffixes -a -e -i -o and -u. That is the very core and beloved feature of the language. Resolving issues around that in the voroz (or voroza) way is important. It would be nice if all the nouns were the same and suffix application could be all consistent. However the fact of life is that nouns do behave differently. The idea of "Noun Classes" is to allow nouns in different classes to behave differently while forcing nouns in the same class to behave analogously. We shall see why this approach may help in the design of Vorlin.

Insights from Another Project

One lesson we can learn is from la lojban. It is an interesting thought provoking constructed language yet I found it extremely difficult to learn. It managed to reduce common nouns adjectives and verbs into one word class called the gismu. A gismu is a "predicate" sort of a verb which relates its arguments in specific "place structure". For example:

x1 tircu x2: x1 is a tree of species x2.
x1 cmene x2 x3: x1 is a name of x2 used by namer x3.
x1 xamgu x2 x3: x1 is good for x2 by standard x3.
x1 viska x2 x3: x1 sees x2 under conditions x3.
x1 klama x2 x3 x4 x5: x1 moves to x2 from x3 via x4 using x5.

Although every gismu follows the same rules of syntax and of word derivation each xi of one gismu may not have any semantic resemblence with the xi of another one. Essentially every gismu is irregular in a unique way because each relation is unique. The user must memorize not only the word but also the sequence of arguments that can be used with it.

(The above paragraph basically repeats what was discussed in the thread "Prepositional phrases of place" (16-Jan-01 09:25:53). I find it suitable as an introduction to this article.)

The Lojban project serve different purposes. For Vorlin although Rick Harrison has offically given up designing an optimal IAL as stated in his Farewell to Auxiliary Languages) I believe the idea of an easy to learn language in his Proposed Guidelines for the Design of an Optimal International Auxiliary Language is what makes Vorlin attractive. Making Vorlin easy to learn should still be our objective.

Why Can Noun Classes Simplify Learning?

Suppose you have 1000 document files on your computer. If you store them all in one folder it is very difficult to find one out. Usually you will create several folders and place related documents under separate folders. Some aspects in language design behaves like that: Too many root words makes memorizing the root words difficult yet too few root words makes memorizing the extended meanings of the root words difficult. "Noun Classes" behaves similarily: introducing a new class means a new set of rules yet with too few classes it will be more difficult to force words to conform with the rules thus creating more exceptions or more unnatural usage.

We say that Lojban is predicate-based because its root words in the unmarked form are predicates. Similarily we say that Vorlin is noun-based because its root words in the unmarked forms are mostly nouns. English for example and most natural languages do not have a scheme:

Since bom gut and vid are all nouns the are syntactically identical in sentences. However if we must apply the same rules on their derivates and compounds we may get something undesirable:

x1 guti. = x1 is good. So "gut" is being good "gutant" is something good. x1 bomi. = x1 is a tree. So "bom" is being a tree "bomant" is the tree.

Perhaps one day I may create a language like that but the original design of Vorlin is not so.

Proposal of Noun Classes

The goal is to select semantically natural classes. Semantically related words require similar derivations and classification done in that way is easier to remember.

I believe a good point to start the classification is to learn from natural languages. The noun-adjective-verb distinction of English root words is not there without a reason: the unmarked form is likely to be the most frequently used concept. Analysis of the Chinese language words is more difficult because they do not change form in different parts of speech but the noun-adjective-verb distinction is similar. To avoid confusion with the parts of speech we shall call those classes "substance" "quality" and "action".

Because of the avoidance of the verb "to be" it seems desirable to have simple verbs and prepositions to express relations. Vorlin also has a unique way to deal with all the emotions. So beside the three basic classes I also want to propose two more classes: "relation" and "emotion" which I find unfit into the first three classes.

The rest of this article is a brief analysis on the behavior of different classes. There will be my opinions on how to derive words from each class. Alternative points of view are welcome.

Action Nouns

This class can be subdivided naturally according to the number of arguments the action usually take:

1. Intransitive actions: sed (sitting) lop (running) sof (sleeping).
2. Transitive actions: bev (drinking) faj (eating) mim (imitation).
3. Actions with more arguments: gev (giving).

The first rule for action words is that the root form always denotes the action. English words are sometimes unclear in this aspect: "a drink" may refer to the action (bev) or the beverage (bevit). If we are careful on that for example not to define "bev" as beverage we can create a large class of consistent words with many useful derivatives.

For deriving verbs the 1999 Grammar recommends that intransitives only take "-i" and the transitives only "-o" not the other way round. That sounds like a good idea and we will stick to this rule for the time being. One implication is that we cannot derive words from intransitives with "-it" the patient marker though "-ant" is allowed for both types of words. Some affixes are designed for the actions: -ep -ob -akso -ikso -enda -ende. They should assume their designed functions. For example:

Remember that once a root is classified as intransitve it will not have a -it form. So we cannot define *sedit as "seat" or *ganit as "song". Whether fajit should mean "food" the usual direct object or just "something that is eaten" is up to us. Perhaps the latter deserves a longer expression like "wel man fajo da diµ". Another similar semantic decision we need to make is the causatives: should the "ya" in "ya fajakso re ku." take more responsiblity than that in "ya kazo ke ku fajo."? Is it appropriate that the former is an action of "feeding" and ther later can be just a indirect cause of eating?

To solve the problem of "strange" prepositions like "*faju" I think we can forbid "-u formation for the actions for the time being. Prepositions makes most sense with the "relations".

How about the "-a" and "-e" words? There are two options:

1. Treat the "-a" as meaning "characterized by" or "pertaining to" so it is different from the "-enda" form. For example: "sofenda kat" is a sleeping cat what is a "sofa kat then? Can a sleeping bag use the word "sofa" or need some special preposition?
2. Treat the "-a" as if it is a contraction of "-enda" so the adjective form is just the active particle of the verb. Perhaps we can eliminate the need of "-enda" after all.

For the suffixes -iz -uc and -oz the lexicon has specification for "actions": the action being done with low moderate or high intensity. Later we will need to decide whether the suffixed words are actions or qualities and work on the semantic implications. If the resulting words are actions that will require -iz words to be nouns and -iza words to be adjectives. Finally let's look at verbs with more than two arguments. The typical example is "gev" which has a direct object and an indirect object. Some old discussions suggest that the beneficiary marker "be" acts like regular prepositions. That means Vorlin verbs can take up to two arguments by itself; other arguments are to be expressed through prepositional phrases. So "gev" can follow same derivations as all other transitive actions. Notice that "gevit" is not "gift" because the normal direct object is ususally not free. To name the receiver of "gev" we may consider this scheme: We create a noun for "be" say "ben" a "relation" which means the "being benefitial to". We pretend that "be" is a short-hand for "benu". So the "beneficiary" is "benit". and we can call the receiver of "gev" as "gevbenit". This scheme works for other prepositional arguments too.

Emotion Nouns

Emotion Nouns are discussed with transitive actions in the 1999 Grammar. The are similar to the transitive actions but I recommend a few changes:

1. The root form is the emotion not an action of feeling that emotion. That is for brevity of the more frequently used concept. In fact can we really distinguish between the two concepts?
2. The word may take both the "-i" and the "-o" form because a person may feel an emotion towards something or just feel the emotion. The "-o" form without object may not work as well here as for transitive actions.

For example:

More examples:

ya luko vido ti. = I am glad to see you. ku luki. = He is happy. ya vilo vido ti. = I want to see you. tis kat fobikso ya. = Your cat frightens me. ne fobi. = Don't be afraid. du filo cina dag. = Do like this day. = Have a nice day.

The most natural way to use the "-a" and "-e" seems to be like shorthands of "-enda" and "-enda" since a "luka ful" is a bird that is happy: "wel luki da ful". The prepositonal "-u" form for example "*vilu" still sounds "strange". However since we have separated the emotions from the actions we can consider the merits separately.

Quality Nouns

Quality nouns are usually derived from adjectives in natural languages. Their stative verb form is discussed with the intransitive actions in the 1999 Grammar. This class whould take the "-a" "-e" and "-i" suffixes naturally but should not take "-o" or "-u". They also take "-iz" "-uc" and "-oz" more naturally. For example:

An important difference from intranstive actions is the way to define the root. "The state of being good" and "the quality of goodness" should be close enough. We may decide later if we need special morphemes to turn "goodness" and "happiness" from quality and emotion into abstract substance. "The state of being red" and "the quality of redness" should be close enough but is "red color" the same concept? We may decide that later too.

Relation Nouns

Relation Nouns are also similar to the transitive actions but there is no activity going on. The "-o" form simply states the relation between the subject and the object. It appears that we do not need the "-i" form. The null object solution as for transitive actions seems to work here. The most important trait of "relations" is that they form natural prepositions.

The "-a" and "-e" forms are officially contractions of "-enda" and "-ende" and are mostly used in forming compounds. Since this is almost a "catch-all" class further work is required to indentify smaller classes. For example the "locatives": lok top duµ etc. may require a separate class. They may similar to hav otherwise but it is more convenient to define the root word as "the location" rather than a "relation". Substance Nouns

Finally we discuss these noun-nouns. They are usually a diµ or treated like one. The "substance" here includes plants animals artifact and abstract matters not belonging to other classes. They do not represent an action a state or any relation. Again this catch-all class is likely to have subdivisions. We need to be very careful in assigning roles of different suffixes to this class because the meanings may not be obvious. We may need to spend time to review practices in the 1999 Grammar and those invented by fans.

For example can we assign "sali" as "being salt" like the qualities? If we do so "sala" will become "wel sali da" which is quite different from the most natural "saline" - "characterized by salt". Should we then assign "sali as "being characterized by salt"? Either definition has its own logic and most probably both will cause confusion. We may be in a better position without a word like "sali".

Can "salo" mean "to salt" as in applying the substance to other thing? If we have both "sali" and "salo" we will surely be confused what is a "salant". If we forbid "sali" "salo" looks attractive but we may be better off without it. The same concept can be safely expressed with an additional action morpheme. For example: "sal{addition}o".

It is possible that after all this class should only allow "-a" among the five vowels. This class is the least productive in forming derivates. When we assign morphemes we should take that into consideration giving higher priority for three-letter words to say transitive actions unless the noun in the class is proven productive otherwise: for example it represents a category of many things.

Conclusion

The Noun Classes described here is not exhaustive. They act as a skeleton that welcomes modification. If we find exceptions that do not fit any class we may put them on an exception list and create a new class when there are enough members. I hope that this article can stimulate all your thoughts towards a better way of constructing and define Vorlin morphemes.

Filiks ,Felix Wan,felix@ksb.net