Documentation for the Grammar Matrix Customization Information Structure Library

Introduction

This document explains how to fill out the Information Structure page of the Grammar Matrix Customization questionnaire and presents background information on the information structure library of the Grammar Matrix Customization System (Bender et al., 2002; Bender and Flickinger, 2005; Bender et al., 2010). General instructions on using the questionnaire can be found here.

Citing the Information Structure Library

The standard reference for the Information Structure Library and its implementations is Song 2014 (A Grammar Library for Information Structure). The full reference and .bib entry can be found here.

Options

The InformationStructure page consists of four parts.

First, in [Focus], a user can specify the canonical position of focus in the user's language. A cross-linguistic survey reports that there are four positions for expressing focus; clause-initial, clause-final, preverbal, and postverbal. What is to be noted is that a sentence in the basic word order does not have any specific information structrual values. A sentence in basic word order is a default form in the language, which can be interpreted as conveying a variey of information structures. For example, if a language basically employs the SVO order, and the canonical position of focus is postverbal, [O] in SVO is not specified as [focus], though it occurs postverbally. A user also can add a focus marker, which include an affix, an adposition, and a modifier (e.g. clitics). An affix and an adpoistion are not specified in the Information Structure page, which should be created in the Morphology page and the Lexcion page respectively. If a user checks out an affix or an adposition of expressing focus in Information Structure, but an affix or an adposition that involve focus or super/sub-types of focus as a value of [information-strucural meaning] in Morphology or Lexcion, there comes a validation error. A modifier can be directly created on the Information Structure page. It can occur before and/or after nouns and/or verbs.

Second, [Topic] has two types of options. As for the constraint on position, topic always occupies the sentence-initial position (i.e. topic-first) in some languages. If a user's language has such a constraint, the checkbox should be marked. After that, a topic marker can be added, which operates in the same way as [Add a Focus Marker] above.

Third, in some languages, contrastive focus can be differently marked from non-contrastive focus. The checkbox under [Contrastive Focus] is not selected, there can exist two types of foci; one is semantic-focus for non-contrastive focus, and the other is contrast-focus for contrastive focus. In that case, a user can choose a specific position for contrastive focus. Likewise, a user can add a contrastive focus marker in the same manner.

Finally, there is an option for [Contrastive Topic], which can barely have a constraint on positioning across languages. So, there is no checkbox for the positioning constraint. Instead, it is reported that some languages (e.g. Vietnamese) employ a contrastive topic marker, which can be specified using the button [Add a Contrastive Topic Marker].

For more information on the implementation of information structure in Grammar Matrix please refer to the section Analyses.

Motivation

There are several motivations in the Information Structure Library.

First, just as other linguistic phenomena, there can some mismatches between form and meaning. That means information structural markings are not necessarily compatible with information structural meanings. For example, the topic marker wa in Japanese, as the name itself implies, is a merely marker. The topic marker can be used to express aboutness topic, contrastive topic, or sometimes contrastive FOCUS. In the library, information structural markings are specified as mkg under CAT, while information structural meanings are specified in ICONS (Individual CONstraintS) under CONT (i.e. mrs).

Second, just as there are semantically empty lexical items, there are also some informatively empty lexical items. In principle, semantically empty items, devoid of predicate (i.e. RELS <! !>), are also informatively empty (i.e. ICONS <! !>). The items include expletives (there, it), case-marking adpoistions (ga, wa in Japanese), by in passive constructions, relative pronouns, etc.

Third, information structural values of a constituent would be less specified or preferentially underspecified. For example, 'The book, Kim reads.' can be interpreted as either 'It is the book that Kim reads.' or 'As for the book, Kim reads it.' In the former, 'the book' conveys a focus meaning, while in the latter it conveys a topic meaning. The MRS representation for the sentence should be able to capture the ambiguity that 'The book' shows. Hence, 'The book' in the sentence is specified as containing focus-or-topic.

Finally, unless there is a linguistic clue to identify the information structural meaning, the value of information structure remains underspecified.

Analyses

There are three type hierarchies for information structure.

First, mkg shows information structural MarKinGs. Note that mkg is a morphosyntactic feature under CAT, which can be different from values of info-str.

mkg := avm &
  [ FC luk,
        TP luk ].
fc := mkg & [ FC + ].
non-tp := mkg & [ TP - ].
tp := mkg & [ TP + ].
non-fc := mkg & [ FC - ].
fc-only := fc & non-tp.
fc-+-tp := tp & fc.
unmkg := non-tp & non-fc.
tp-only := tp & non-fc.

Second, sform shows the Sentential FORMs, which inherits from binary-headed-phrase. This hierarchy constrain combining two phrases with mkg values. For example, topic-comment requires [MKG tp] of non-head-daughter, which prevent focus-marked constituents from being used as the non-head-daughter of topic-comment constructions.

sform := basic-binary-headed-phrase.
focality := sform &
  [ SYNSEM.LOCAL.CAT.MKG fc-only,
        NON-HEAD-DTR.SYNSEM.LOCAL.CAT.MKG fc ].
topicality := sform.
narrow-focus := focality &
  [ HEAD-DTR.SYNSEM.LOCAL.CAT.MKG non-fc ].
wide-focus := focality.
topicless := topicality &
  [ NON-HEAD-DTR.SYNSEM.LOCAL.CAT.MKG non-tp ].
topic-comment := topicality &
  [ SYNSEM.LOCAL.CAT.MKG tp,
    NON-HEAD-DTR.SYNSEM.LOCAL.CAT.MKG tp ].
focus-bg := narrow-focus & topicless &
  [ HEAD-DTR.SYNSEM.LOCAL.CAT.MKG unmkg ].
all-focus := wide-focus & topicless &
  [ HEAD-DTR.SYNSEM.LOCAL.CAT.MKG fc ].
frame-setting := topic-comment.
non-frame-setting := topic-comment &
  [ HEAD-DTR.SYNSEM.LOCAL.CAT.MKG non-tp ].

Finally, information structural meanings, such as focus, topic, contrast, background, etc., are subtypes of info-str. The info-str hierarchy is designed in a flexible way. What is of importance is how much specified the information structural meaning of a constituent is.

icons := avm.

info-str := icons &
  [ CLAUSE individual,
    TARGET individual ].

; constraints on information structure
non-topic := info-str.
contrast-or-focus := info-str.
focus-or-topic := info-str.
contrast-or-topic := info-str.
non-focus := info-str.

focus := non-topic & contrast-or-focus & focus-or-topic.
contrast := focus-or-topic & contrast-or-focus & contrast-or-topic.
topic := non-focus & focus-or-topic & contrast-or-topic.

bg := non-topic & non-focus.

semantic-focus := focus.
contrast-focus := contrast & focus.
contrast-topic := contrast & topic.
aboutness-topic := topic.

NOTE

The information structure values are represented in ICONS. Currently, ACE(AceTop) and AGREE(AgreeTop) fully process ICONS. So MRSes coming from LKB/PET do not have ICONS.

In some languages, both focus and topic can occur in the clause-initial position. In that case, topic always precedes focus, and a counterexample to this has not yet been reported across languages. The current version of Information Structure library does not place an ordering constraint on such a co-occurrence of focus and topic.

Modules for constraining positions of information structural components in FREE word order languages are currently under development. All types of word order is potentially used in a free word order language, and the Word Order page in customization system specifies "free" as "pragmatically determined word order". We assume that "pragmatically determined" means "information structurally conditioned". Thus, variations in word order of such a language need to be deeply investigated one by one. The current version does not place such an elaborated constraint on word order variations in the languages. If word order is checked as "free" in Word Order, and multiple positions are checked, the grammar may not work properly in terms of information structure. For example, a position for focus is specified but it is not clause-initial, and the checkbox for the topic-first constraint is checked at the same time, information structure values might be entangled.

Modifiers of expressing information structure in the current system are semantically empty, which means they require trigger rules in trigger.mtr. Unfortunately, there is no way to constrain them in transfer. Tentatively, they are just filtered out in transfer.

References

Song, Sanghoun. 2017. Modeling information structure in a cross-linguistic perspective. Language Science Press.

Song, Sanghoun. 2014. A Grammar Library for Information Structure. Doctoral Dissertation, University of Washington.

Song, Sanghoun and Bender, Emily M. 2012. Individual Constraints for Information Structure. In Stefan Müller (ed.), Proceedings of the 19th International Conference on Head-Driven Phrase Structure Grammar, pages 329-347, Stanford, CA: CSLI Publications.

bibtex:

@book{song:17,
author = {Song, Sanghoun},
title = {Modeling information structure in a cross-linguistic perspective},
series = {Topics at the Grammar-Discourse Interface},
year = {2017},
publisher = {Language Science Press}
}

@phdthesis{song:14,
author = {Song, Sanghoun},
title = {A Grammar Library for Information Structure},
school = {University of Washington},
year = {2014}
}

@inproceedings{song:bender:12,
title = {Individual Constraints for Information Structure},
author = {Song, Sanghoun and Bender, Emily M.},
pages = {329--347},
editor = {Stefan M{\"u}ller},
booktitle = {Proceedings of the 19th International Conference on Head-Driven Phrase Structure Grammar},
address = {Stanford, CA},
publisher = {CSLI Publications},
year = {2012}
}

MatrixDoc/InformationStructure (last edited 2018-05-07 15:32:14 by GuyEmerson)

(The DELPH-IN infrastructure is hosted at the University of Oslo)