Archive:Frame: Difference between revisions

From The Toaq Wiki
mNo edit summary
No edit summary
Line 1: Line 1:
A '''frame''' is a class of verbs that all, in a certain specific sense, have the "same" argument places, giving rise to the same grammatical behavior. There are two senses of "same", giving rise to two types of frames: '''semantic''' and '''serial''' frames.
A '''frame''' is a class of verbs that all, in a certain specific sense, have the "same" argument places, giving rise to the same grammatical behavior. There are two types of frame: '''semantic''' and '''serial''' frames.


== Semantic frames ==
== Semantic frames ==
Line 77: Line 77:


TODO: what happens to {{t|jıe}}?
TODO: what happens to {{t|jıe}}?
== Why complicate things? ==
If semantic frames can be predictably mapped to serial frames (serial behaviors), we could just describe the rules for serialization in terms of semantic frames. In fact, this is pretty much what the [[refgram]] does.
<div style="border:1px solid black;padding:5px 15px;background:#ffd">
* If there is exactly one non-concrete argument place in the '''type signature''', subordinate and merge there.
* If the '''type signature''' is <code>(c c)</code>, perform genitival serialization in the second argument place.
* If the '''type signature''' is <code>(c)</code>, perform adjectival serialization.
* Otherwise, the verb cannot serialize.
</div>
So it might seem unnecessary to talk about serial frames as a separate "thing".
However, for a long time it has been unclear if there is such a predictable mapping (and it still is a little), and for a long time it was thought that verbs like {{t|she}} can participate in serials, and it was hotly debated whether its <code>(0 0)</code> type signature should have <code>(x 0)</code> or <code>(0 x)</code> serial behavior.
At the very least, a separate notation for serial behaviors still lets us theorize about these things.

Revision as of 17:00, 28 September 2021

A frame is a class of verbs that all, in a certain specific sense, have the "same" argument places, giving rise to the same grammatical behavior. There are two types of frame: semantic and serial frames.

Semantic frames

A semantic frame is a class of verbs that all have the same amount of argument places, of the same types, in the same order.

The type of an argument place says what kind of thing is allowed to go in there:

  • c ("concrete") means anything goes. Usually concrete arguments like "me" or "cats" or "a house", but also clauses/events whenever it makes sense in that verb.
    • For example, de's first slot is type c. One can say dẻ súq (You're beautiful) or dẻ sûaq súq (The event of your singing is beautiful).
  • 0 means a place that must be filled with a proposition or event.
    • Often such slots are filled with a rising-falling tone content clause, or a pronoun referring to one.
    • In the dictionary, look for the words "that ___ is the case" to recognize these slots.
  • 1 means the place must be filled with a property.
    • Often such slots are filled with a rising-falling tone content clause with one ja in it.
    • In the dictionary, loook for the words "satisfying property ___" to recognize these slots.
  • 2 means the place must be filled with a binary relation.
    • Often such slots are filled with a rising-falling tone content clause with two instances of ja in it.
    • In the dictionary, loook for the words "relation ___" to recognize these slots.

Thus, a semantic frame can be identified with its type signature, which is just all the types of the argument places listed in a row — traditionally in parentheses.

For example: leo ("tries to") and juoq ("should") both take one concrete argument followed by one property argument. This is expressed by the type signature (c 1). These verbs have the same type signature, so they belong to the same semantic frame.

Furthermore, each semantic frame in Toaq has an arbitrary representative chosen for it, used as a handy way to refer to the frame. For example, the semantic frame of all verbs with type signature (c 1) is called the LEO (semantic) frame. LEO consists of all the verbs whose argument places are just like leo's.

We say that “juoq is in the LEO frame” or “juoq is in LEO”. We also often just say that “juoq is (c 1)”.

Table of semantic frames

TODO

Serial frames

A serial frame is a class of verbs that all exhibit the same serialization behavior.

Again, each such frame is identified by a serial signature and a representative.

But this time, rather than describing "what can this argument slot be filled with?", the elements of the signature describe: "what happens to this argument slot, when this verb is the first (left-hand) verb in a serial verb?"

These are the possible serial types an argument place can have:

  • x or c means that this argument place remains untouched and will still be there in the resulting serial verb.
  • 0 means that this place will be subsumed by all of the right-hand verb's arguments:
    • x wants 0 to be the case + x sits on x = x wants x to sit on x
  • 1 means that this place will be subsumed by all of the right-hand verb's arguments, merging with its first one:
    • x tries to satisfy 1 + x sits on x = x tries to sit on x
  • 2 means that this place will be subsumed by all of the right-hand verb's arguments, merging with its first two:
    • x all reciprocally satisfy 2 + x agrees with x that 0 is the case = x all agree that 0 is the case
  • e means that this place will disappear, merging with "baq right-hand-verb".
    • x takes care of e + x is a cat = x takes care of cat(s).
    • Such a slot is known as an exhibitor slot (hence e), and the resulting serial is a genitival serial.
  • a means that this adjectival place will disappear, and the following verb's first place is modified, attributively when appropriate or otherwise predicatively, by this adjective:
    • a is small + x is a cat = x is a small cat.

The signature is again written by writing down all the types in parentheses, like (x x 0) or (x e).

There is at most one non-x in a serial signature. This is because all the other slot types define the serialization behavior for the verb, and a verb must have one unambiguous serialization behavior! So while (c 0 0) is a valid semantic frame, there cannot be a (x 0 0) serial behavior, as it wouldn't be clear which of the 0 slots accepts the right-hand verb arguments.

Some verbs cannot participate in serials, and are not part of any serial frame.

Table of serial frames

TODO

How do they relate?

(This is the author's unofficial theory.)

There is a predictable partial function from semantic frame signatures to serial frame signatures.

If the semantic frame has more than one "digit" (0, 1 or 2) in it, then the verb cannot serialize. Example: she.

If it has exactly one "digit": the serialization behavior is "subordinating", and the serial frame is obtained by replacing all c with x. Example: tua, leo, taq, mıujeq, huaq, toı.

If there are no "digits":

  • Semantic frame (c) corresponds to serial frame (a). Example: nuı, kue(?!).
  • Semantic frame (c c) corresponds to serial frame (x e). Example: hea, kıaı, tuı.
  • Otherwise (i.e. (c c c) and beyond, like kuq), the verb cannot serialize.

TODO: what happens to jıe?

Why complicate things?

If semantic frames can be predictably mapped to serial frames (serial behaviors), we could just describe the rules for serialization in terms of semantic frames. In fact, this is pretty much what the refgram does.

  • If there is exactly one non-concrete argument place in the type signature, subordinate and merge there.
  • If the type signature is (c c), perform genitival serialization in the second argument place.
  • If the type signature is (c), perform adjectival serialization.
  • Otherwise, the verb cannot serialize.

So it might seem unnecessary to talk about serial frames as a separate "thing".

However, for a long time it has been unclear if there is such a predictable mapping (and it still is a little), and for a long time it was thought that verbs like she can participate in serials, and it was hotly debated whether its (0 0) type signature should have (x 0) or (0 x) serial behavior.

At the very least, a separate notation for serial behaviors still lets us theorize about these things.