ICONS-KEY in clauses

I used the Grammar Matrix information structure library to produce a Russian grammar fragment with free word order (which is not officially supported), and am currently trying to make it so it it covers all the basic sentences that I have which include intransitive, transitive, ditransitive, and sentences with clausal complements and embedded questions.

The library produces phrase structure rules which constrain the verb’s arguments’ ICONS-KEY for focus or topic (depending on what the user said in the questionnaire with respect to whether focus is clause-final or whatever). They also have ``periphery’’ constraints which are there to prevent multiple parses. They basically look like this:

head-periph-subj-phrase := decl-head-subj-phrase & head-initial-head-nexus &
  [ NON-HEAD-DTR.SYNSEM [ R-PERIPH +,
                          LOCAL.CONT.HOOK.ICONS-KEY semantic-focus ],
    SYNSEM [ R-PERIPH +,
             L-PERIPH + ],
    HEAD-DTR.SYNSEM [ R-PERIPH -,
                      L-PERIPH +,
                      LOCAL [ CAT.VAL.COMPS < >,
                              CONT.HOOK.ICONS-KEY contrast-or-topic ] ] ].

The rule above assumes that it is the final one to apply and that the head complement rule had already applied.

My questions are:

(1) What is the meaning of the clause’s own ICONS-KEY?

This is what I get (it is coming from basic-non-rel-clause, I think), and with that in place, I cannot immediately use the clause as a clausal complement, since the head-complement rule wants ICONS-KEY of type topic or focus. I could probably add something like that to the clause type but would that be meaningful?

(2) Additionally, what should happen to the periphery constraints? Would it make sense to have a special head-complement rule for complementizers which ignores them? (I seem to be needing such a rule anyway as free word order does not apply to adpositions).

(3) What about the ICONS diff list? It seems to be empty everywhere. Shouldn’t stuff be put on it and percolate up the tree? How to access the information structure information once the clause has been formed?

Correction: the clause’s ICONS-KEY value does not come from non-rel-clause; it probably comes from headed-phrase?

headed-phrase := phrase &
  [ SYNSEM.LOCAL [ CAT.HEAD head & #head,
                   AGR #agr,
                   COORD -,
                   COORD-REL #crel ],
    C-CONT.HOOK.ICONS-KEY #icons,
    HEAD-DTR.SYNSEM.LOCAL local &
		 [ CAT.HEAD #head,
		   CONT.HOOK.ICONS-KEY #icons,
		   AGR #agr,
		   COORD -,
		   COORD-REL #crel ] ].

We discussed this in today in Emily’s office, here’s the summary:

(1) The above examples are probably not really relevant because it is not fully clear that this kind of grammar should even be customized. For the free word order option, the GM currently says “pragmatically determined”. This probably means that there is somehow no “basic” order for structures unmarked for focus, whereas something like Russian can be modeled as an SVO language with clause-final focus. Then additional orders will be licensed by extraction+filler gap rules.

(2) What kind of behavior is in fact envisioned for the free word order with information structure constraints? Perhaps I can find some insights in partially-free orders, such as v-initial and v-final, for which there are regression tests in the system.

@sanghoun Could you maybe tell us whether we are on the right track here?

Dear Olga,

(1) Several years ago when I worked with Varya with her small-size Russian grammar, we made a similar approach to yours. So, basically, I think this approach is good enough. I’ve also tried to understand what is genuinely a free word order language and what are the basic properties of such a language, but I’ve found no clear idea.

(2) The notion of “free word order” is a little bit different from that of “v-final”, I think. Korean and Japanese employ the so-called scrambling, and therefore the order between the subject and the object does not affect the acceptability and meaning difference of sentences. In contrast, “free word order” (if it is really existent) means NO word order in a sense. This may implies that we may have to build up various phrase structure rules for all types of sentences one by one manually.

I am quite interested in this topic. Let me follow up your additional discussion and questions.

Sanghoun

1 Like