Grammar Formalisms Head-Driven Phrase Structure Grammar (HPSG) - PowerPoint PPT Presentation

Grammar Formalisms Head-Driven Phrase Structure Grammar (HPSG) Laura Kallmeyer, Timm Lichte, Wolfgang Maier Universit¨ at T¨ ubingen 27.06.2007 HPSG 1

Outline Feature Structure Basics The building blocks of HPSG Analyzing raising and control Analyzing long-distance dependencies HPSG 2

Subsumption and Unification { < attr 1 , val 1 >, < attr 2 , val 2 >, 2 3 attr 1 val 1 attr 2 val 2 ..., < attr n , val n > } 6 7 6 7 ... ... 6 7 4 5 val n attr n A ⊑ B , iff subsumption ⊑ : if t ∈ A , then t ∈ B . A ⊔ B = C , iff unification ⊔ : C is the smallest feature structure such that A ⊑ C and B ⊑ C . HPSG 3

Feature structures as Values Feature structures as values: non-recursive: 2 3 2 1 3 per sg 4 index num 6 7 4 5 5 fem gend recursive: " # fi fl » – h i subcat subcat ... = ⇒ Other than FTAG, HPSG uses recursive feature structures! HPSG 4

Structure sharing (1) Re-entrancies (or “structure sharing”): boxed numbers ( 1 , 2 , ...) indicate structure sharing within feature structures: " # " # » – 1 val 1 h i attr 1 attr 1 1 attr 1 attr 2 1 1 attr 2 1 attr 2 1 Re-entrancies produce cyclic structures! HPSG 5

Structure sharing (2) Correspondence between feature structures and graphs: attr1 attr3 2 3 type1 type2 val type1 attr 1 1 6 7 6 7 2 3 6 type2 7 attr2 6 7 6 7 attr 3 val attr 2 1 6 7 attr4 4 5 4 5 attr 4 1 HPSG 6

Types and type hierarchy Types: fixed bundles of attributes 2 3 2 3 index synsem per 1 loc 1 6 7 6 7 6 7 4 5 6 num 2 7 nonloc 2 4 5 gend 3 � specified in the signature Type hierarchy: order on types ≈ attribute inheritance " # sign phon phon synsem synsem word " # phrase 2 3 phon phon phon phon synsem synsem synsem synsem 6 7 4 5 con-struc dtrs � specified in the signature HPSG 7

Signatures, constraints, and grammars (1) The signature Σ is a quadruple (in P&S,94): < types , type hierarchy , attributes , ((types × attributes) × types) > The signature licenses possible feature structures. (language independent) A constraint τ ∈ θ is based on Σ, but also contains variables, conjunction, disjunction, implication, negation ,... Other name: Principle . The constraints further restrict the set of well-formed feature structures. (language independent and dependent) HPSG 8

Signatures, constraints, and grammars (2) A grammar G is a signature Σ and a set of constraints θ . Each G -licensed feature structure must be type-resolved : each type is maximally specific. totally well-typed : exactly the legal attributes of a type must be present. HPSG 9

Phenomena, Model, Formal Theory from M¨ uller (2007) models Phenomenon: Ling. Objects Model: Feature Structures predicts determines licensed by the theory Formal Theory: Feature Descriptions HPSG 10

The architecture of linguistic feature structures 2 3 sign phon phon 6 7 6 7 6 7 2 3 synsem 6 7 6 7 2 3 6 local 7 6 7 6 7 6 7 6 7 category category 6 7 6 7 6 7 synsem local 6 7 6 7 6 7 content content 6 7 6 7 4 5 6 6 7 7 6 context context 7 6 7 6 7 4 5 4 5 nonlocal nonlocal phon : The surface of a sign. synsem : The core of a sign . Contains syn tactic and sem antic information. local : Contains information local to the linguistic sign, such as agreement information nonlocal : Contains non-local references, e.g. for long-distance dependencies HPSG 11

An example 2 3 word ˙ she ¸ phon 6 7 6 7 6 2 synsem 3 7 6 7 6 2 3 7 local 6 7 6 7 6 7 6 7 2 cat 3 6 7 6 6 7 7 6 7 6 6 » – 7 7 6 noun 7 6 6 6 7 7 7 6 7 category head 6 6 7 7 6 6 7 7 case nom 6 6 6 4 5 7 7 7 6 6 6 7 7 7 �� subcat 6 7 6 6 7 7 6 7 6 6 7 7 6 2 3 7 ppro 6 6 7 7 6 7 6 6 7 7 6 7 2 3 ref 6 6 7 7 6 7 6 7 synsem 6 local 6 7 7 6 7 6 7 3rd content per 6 6 6 7 7 7 6 6 7 7 index 6 6 6 1 7 7 7 6 6 7 7 sing 6 num 7 6 6 7 7 6 4 4 5 5 7 6 6 7 7 6 7 fem 6 6 gend 7 7 6 7 6 6 7 7 6 7 6 6 2 3 7 7 6 context 7 6 6 7 7 6 7 8 9 6 6 6 7 7 7 psoa 2 3 6 6 6 6 7 7 7 7 context < = 6 6 6 6 7 7 7 7 female backgr reln 6 7 6 4 4 4 5 5 5 7 6 7 6 7 inst 1 6 : ; 7 4 5 4 5 . . . nonlocal HPSG 12

Category category : Contains syntactic information 2 3 cat » – 6 noun 7 6 7 head 6 7 nom case 6 7 4 5 subcat �� head : Syntactic features (roughly part-of-speech) subcat : Subcategorization information HPSG 13

Content and Context content : Contains semantic information 2 3 ppro 2 3 6 ref 7 6 7 6 3rd 7 per 6 7 6 7 index 1 6 7 6 7 sing num 6 7 4 5 4 5 fem gend context : Contains pragmatic information 2 3 context 8 9 6 7 2 3 psoa 6 7 > > 6 < = 7 6 reln female 7 backgr 6 7 6 7 4 5 4 5 > inst > 1 : ; HPSG 14

Abbreviations of feature structures (1) 2 3 synsem 2 3 6 2 3 7 » – noun 6 7 4 head 6 7 6 case 7 cat case 6 6 7 7 6 7 6 5 7 6 7 NP i [ per , num , gend ] [ case ] 6 subcat �� 7 � 6 7 6 7 loc 6 7 6 7 6 7 2 3 6 per 7 per 6 7 6 7 6 7 6 7 num num cont | index i 6 6 7 7 6 7 4 4 5 5 4 5 gend gend 2 3 synsem 2 3 2 3 6 » – 7 verb 6 7 head 6 7 vform 6 vform 7 6 7 VP[ vform ]: i 6 7 cat � 6 7 6 6 7 7 loc 6 7 6 D E 7 4 5 6 7 synsem 6 subcat 7 6 7 6 7 4 5 4 5 cont i HPSG 15

Abbreviations of feature structures (2) " # » – noun h i Noun[nom] head head � nom case 2 3 2 3 index i " # » – index i 6 » – 7 rel 4 cont 6 7 content 5 � 6 7 rel ( 1 ) restr 4 5 arg1 1 " # » – verb h i Verb[fin] head head � vform fin 2 3 2 3 rel 6 arg1 7 1 h i 6 6 7 7 rel ( 1 , ..., n ) cont content 5 � 6 6 7 7 ... ... 6 7 4 5 4 argn n HPSG 16

Abbreviations in action 2 word 3 D E 6 phon walks 7 6 7 6 7 6 2 3 7 " # verb 6 7 head 6 7 7 6 6 vform fin 7 6 6 7 7 6 6 7 7 6 6 7 7 2 2 3 3 2 3 " # 6 6 7 7 noun 6 6 7 7 head 6 6 7 7 6 6 7 7 6 7 cat case nom 6 6 7 7 synsem | loc | cat 6 6 7 7 6 7 6 6 7 7 * 6 6 7 7 + 4 5 6 7 6 7 6 6 7 7 subcat �� 6 6 subcat loc 7 7 6 6 7 7 6 6 7 7 6 6 7 7 6 6 2 3 7 7 6 6 7 7 " # 6 6 per 3 7 7 6 6 7 7 6 6 7 7 6 6 cont 4 index 7 7 6 6 7 7 5 4 4 num sg 5 5 4 4 5 5 2 3 word D E 6 7 phon walks 6 7 6 7 � 2 3 6 7 head Verb[fin] 6 7 6 synsem | loc | cat 7 D E 4 5 NP 1 [ 3sg ] [ nom ] 4 subcat 5 HPSG 17

Phrases - Syntactic composition = ⇒ Structures of type phrase have the feature dtrs . 2 phrase 3 D E 6 7 phon Kim, walks 6 7 6 7 6 " # 7 head Verb[fin] 6 7 synsem | loc | cat 6 7 6 subcat �� 7 6 7 6 7 2 3 6 head-comp-struc 7 6 7 6 7 2 3 6 6 phrase 7 7 6 7 6 7 6 7 6 D E 7 6 6 7 7 6 phon walks 7 6 6 7 7 6 7 6 7 6 6 head-dtr 7 7 6 6 2 3 7 7 6 head Verb[fin] 7 6 7 6 6 7 7 dtrs 6 6 7 7 6 synsem | loc | cat 7 D E 6 4 4 5 5 7 6 subcat NP[ nom ] 7 6 7 6 7 6 7 6 7 6 6 7 7 6 7 6 7 2 D E 3 6 * 4 phon Kim + 7 6 7 6 7 6 7 comp-dtrs 6 7 6 7 5 4 5 4 synsem NP[ nom ] 5 " # Tree notation: head Verb[fin] subcat �� C H 2 3 head Verb[fin] NP[ nom ] D E 4 5 subcat NP[ nom ] Kim walks HPSG 18

Principles The Word Principle word ⇒ LE 1 ∨ ... ∨ LE n The Head Feature Principle " # synsem | loc | cat | head 1 dtrs | head-dtr | synsem | loc | cat | head 1 The Subcategorization Principle (binary version) 2 3 D E synsem | loc | cat | subcat 2 6 7 2 3 6 D E 7 2 | 1 4 head-dtr | synsem | loc | cat | subcat 6 7 6 dtrs 7 5 4 5 comp-dtr | synsem 1 HPSG 19

The Immediate Dominance Schemata (1) The ID Principle Every headed phrase must satisfy exactly one of the ID schemata. ≈ phrase structure rules For example (from P&S,94): Head-Subject Schema Head-Complement Schema Head-Subject-Complement Schema Head-Marker Schema Head-Adjunct Schema Head-Filler Schema HPSG 20

The Immediate Dominance Schemata (1) = ⇒ including Head Feature Principle and Subcategorization Principle! " # head 1 Head-Subject Schema: �� subcat C H 2 phrase 3 head 1 6 7 2 6 7 D E 4 5 subcat 2 Head-Complement Schema: 2 3 head 1 D E 4 5 subcat 2 H C C C 2 3 head 1 ... n D E 3 4 5 2 , 3 ,..., n subcat HPSG 21

Raising and control (1) - A semantic distinction (1) a. John seems to leave. b. John tries to leave. “seems”: 2 3 word 2 3 6 7 Verb[fin] head 6 7 6 7 cat D E 6 4 5 7 1 [ 3sg ] [ nom ], VP[ inf ,subcat< 1 >]: 3 subcat 6 7 6 7 4 5 seem ( 3 ) cont “tries”: 2 3 word 2 3 6 Verb[fin] 7 head 6 7 6 7 cat D E 6 4 5 7 NP 1 [ 3sg ] [ nom ], VP[ inf ,subcat<NP 1 >]: 3 subcat 6 7 6 7 4 5 try ( 1 [ ref ] , 3 ) cont = ⇒ no PRO and no ECM! = ⇒ by and large: control and raising only differ semantically! HPSG 22