to as a linguistic theory to to a member of the family of linguistic frameworks that are called generative grammars a grammar which is formalized to a high degree and thus makes exact predictions about the objects (sentences) that are considered to belong to a language. does not have phrase structure rules or transformations. Instead, it is declarative, non-derivational, and constraint-based. Kordula De Kuthy April 29, 2009 Sets of constraints which hold simultaneously determine the collections of admissible linguistic structures without defining an order of the derivation or generation of signs. 1 / 31 2 / 31 The Beginnings to to Is there such a thing as one single coherent framework that was created sometime in the early days of, remained unchanged, and is employed by all linguists working in? Two different formalisms have been informally presented by Carl Pollard and Ivan Sag: One presented in their first book of 1987, Information-based Syntax and Semantics (Pollard & Sag 1987) the other one in their second book of 1994, Head-Driven Phrase Structure Grammar (Pollard & Sag 1994) began in the mid 1980s inspired by several other frameworks: Government and Binding (Chomsky 1981) ideas of the combinatorial system of from Categorial Grammar combination of feature structure and phrase structure from Lexical Functional Grammar (LFG) (Bresnan 1982) and Generalized Phrase Structure Grammar(GPSG) (Gazdar et al. 1985) 3 / 31 4 / 31
87-94 to grammars from a linguistic perspective to 87 is a typical instance of a unification-based grammar formalism. Underlying intuition is that linguist specify pieces of partial information about a language in their grammars. All pieces of partial information about a language are then combined by operations such as unification to obtain all available information about that language. 94 can be called an object-based grammar formalism or a constraint-based grammar formalism. An grammar, from a linguistic perspective, consists of a) a lexicon: licensing basic words b) lexical rules: licensing derived words c) immediate dominance (ID) schemata: licensing constituent structure d) linear precedence (LP) statements: constraining word order e) a set of grammatical principles: expressing generalizations about It envisions an architecture in which linguists use a logical language on order to specify language as a collection of total objects. 5 / 31 6 / 31 grammars from a formal perspective to The signature of an grammar to The signature An grammar formally consists of I. the signature as declaration of the domain, and II. the theory constraining the domain. defines the ontology ( declaration of what exists ): consists of which kind of objects are distinguished, and which properties of which objects are modeled. the type hierarchy (or sort hierarchy) and the appropriateness conditions, defining which type has which appropriate attributes (or features) with which appropriate values. 7 / 31 8 / 31
An example type hierarchy to to Signs sign PHON list(phonstring) SYNSEM synsem word phrase Models of What do the mathematical structures used as model for theories look like? The objects are modelled by typed feature structures, which can be notated as directed graphs. Since these models represent objects in the world (and not knowledge about objects in the world) they are total with respect to the ontology declared in the signature. Formally speaking, the feature structures are Part of speech p-o-s totally well-typed: Every type has every one of the attributes and their values which are appropriate for it. sort-resolved: Every type is maximally specific. Note that type and sort are often used synonymously, as well as attribute and feature. adj adv det noun prep 9 / 31 10 / 31 How do we express a theory? to (cont.) to A description language and its abbreviating AVM (attribute value matrix) notation is used to talk about sets of objects. consists of three building blocks: Type descriptions single out all objects of a particular type, e.g., word Attribute-value pairs describe objects that have a particular property. The attribute must be appropriate for the particular type of object, and the value can be any kind of description, e.g., [ SPOUSE [ NAME mary ]] Complex descriptions are obtained through conjunction ( ), disjunction () and negation ( ). In the AVM notation, conjunction is implicit. Tags (structure sharing) to specify token identity, e.g. 1 11 / 31 12 / 31
An example AVM - The pronoun she word PHON <she> local cat [ ] noun CAT CASE nom SUBCAT ppro ref SYNSEM LOC PER third CONT INDEX 1 NUM sing GEND fem RESTR {} context psoa CONTEXT BACKGR RELN female INST 1 to The theory of an grammar A theory is a set of description language statements, often referred to as the constraints. The theory singles out a subset of the objects declared in the signature, namely those which are grammatical. A linguistic object is admissible with respect to a theory iff it satisfies each of the descriptions in the theory and so does each of its substructures. to 13 / 31 14 / 31 Connection between, Model and Empirical Domain to Signs to phenomena linguistic objects predicts modelling specify set of descriptions constraints formal theory model feature structures are licensed word synsem LOCAL local NON-LOCAL non-local sign PHON list(phonstring) SYNSEM synsem [ ] phrase DTRS constituent-structure local CATEGORY category CONTENT content CONTEXT context category head SUBCAT list(synsem) MARKING marking 15 / 31 16 / 31
Motivating SUBCAT to Syntactic category information to (1) a. I laugh. (<NP>) b. I saw him. (<NP NP>) c. I give her the book. (<NP NP NP>) d. I said that she left. (<NP S[that]>) Cannot always be derived from semantics: (2) a. Paul ate a steak. (<NP>) b. Paul ate. (<NP NP>) (3) a. Paul devoured a steak. (<NP>) b. * Paul devoured (<NP NP>) [ ] functional SPEC synsem head marker determiner adjective VFORM vform AUX boolean INV boolean substantive PRD boolean MOD mod-synsem [ ] [ ] noun preposition... CASE case PFORM pform 17 / 31 18 / 31 Properties of particular categories to Motivating VFORM to vform (4) a. Peter will win the race. (base form) b. * Peter will won the race. c. * Peter will to win the race. finite infinitive base gerund present-part. past-part. passive-part. case pform (5) a. Peter has won the race. (past participle) b. * Peter has win the race. c. Peter has to win the race. ( different ) nominative accusative of to... (6) a. Peter seems to win the race. (to-infinitive) b. * Peter seems win the race. c. * Peter seems won the race. 19 / 31 20 / 31
Motivating CASE to Indices to (7) a. He left. (nom) b. * Him left. index PERSON person NUMBER number GENDER gender (8) a. She sees him. (acc) b. * She sees he. person referential there it number gender first second third singular plural masculine feminine neuter 21 / 31 22 / 31 Semantic representations to Auxiliary data structures to content quant [ ] laugh give LAUGHER ref GIVER ref GIVEN ref GIFT ref psoa nom-obj INDEX index RESTRICTION set(psoa) drink think DRINKER ref THINKER ref DRUNKEN ref THOUGHT psoa boolean true false list... empty-list non-empty-list FIRST REST list Alternative names for the attributes FIRST (FT) and REST (RT) of non-empty-list are (HD) and TAIL (TL). 23 / 31 24 / 31
Abbreviations for describing lists to Abbreviations for common AVMs to empty-list is abbreviated as e-list, <> non-empty-list is abbreviated as ne-list [ ] FIRST 1 REST 2... 1 FIRST 1 REST [ FIRST 2 REST 3 ] is abbreviated as is abbreviated as is abbreviated as 1 2... 1 1, 2 3 Pollard and Sag (1994) make use of the following abbreviations for describing synsem objects: Abbrev. abbreviated AVM synsem [ ] noun NP 1 CATEGORY LOCAL SUBCAT CONTENT INDEX 1 synsem [ ] S: 1 CATEGORY LOCAL SUBCAT CONTENT 1 synsem VP: 1 CATEGORY LOCAL SUBCAT synsem CONTENT 1 25 / 31 26 / 31 The Lexicon to An example lexicon to The basic lexicon is defined by the Word Principle as part of the theory. It is an implicational statement defining which of the ontologically possible words are grammatical: word lexical-entry 1 lexical-entry 2... with each of the lexical entries being descriptions, such as e.g.: word PHON <laughs> [ ] CAT VFORM fin SYNSEM LOC SUBCAT NP[nom] 1 [3rd,sing] [ ] laugh CONTENT LAUGHER 1 word PHON <drinks> [ ] CAT VFORM fin SUBCAT NP[nom] S L 1 [3rd,sing], NP[acc] 2 drink CONT DRINKER 1 DRUNKEN 2 PHON <drink> [ ] CAT VFORM fin SUBCAT NP[nom] S L 1 [plur], NP[acc] 2 drink CONT DRINKER 1 DRUNKEN 2 27 / 31 28 / 31
to to PHON <give> [ ] VFORM fin CAT NP[nom] 1 [plur], NP[acc], 2 SUBCAT PP[to] S L 3 give GIVER 1 CONT GIFT 2 GIVEN 3 PHON <to> [ ] preposition CAT PFORM to S L SUBCAT NP[acc] 1 [ ] CONT INDEX 1 PHON <think> [ ] CAT VFORM fin SUBCAT NP[nom] S L 1 [plur], S[fin]: 2 think CONT THINKER 1 THOUGHT 2 29 / 31 30 / 31 to References to PHON <poets> [ ] noun CAT SUBCAT SYNSEM LOC [ PER third CONT INDEX ] NUM plur Bresnan, J. (ed.) (1982). The Mental Representation of Grammatical Relations. Cambridge, MA: MIT Press. Chomsky, N. (1981). Lectures on Government and Binding. Dordrecht: Foris Publications. Gazdar, G., E. Klein, G. K. Pullum & I. A. Sag (1985). Generalized Phrase Structure Grammar. Cambridge, MA: Harvard University Press. Pollard, C. & I. A. Sag (1987). Information-based Syntax and Semantics, Vol. 1: Fundamentals. No. 13 in CSLI Lecture Notes. Stanford, CA: CSLI Publications. Pollard, C. & I. A. Sag (1994). Head-Driven Phrase Structure Grammar. Chicago, IL: University of Chicago Press. PHON <wine> [ ] noun CAT SUBCAT SYNSEM LOC [ PER third CONT INDEX ] NUM sing 31 / 31 31 / 31