Grammar Implementation with Lexicalized Tree Adjoining Grammars and - PowerPoint PPT Presentation

Outline of today’s course The derivation tree 1 Design principles for elementary trees 2 Sample derivations 3 NP and PP complements Sentential complements and long-distance dependencies Modifiers Feature based TAG 4 Summary and outlook 5 Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 31 9

Linguistic analyses with LTAG What is an elementary tree, and what is its shape? Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 32 10

Linguistic analyses with LTAG What is an elementary tree, and what is its shape? syntactic/semantic properties of ? elementary trees ⇐ = linguistic objects Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 33 10

Linguistic analyses with LTAG What is an elementary tree, and what is its shape? syntactic/semantic properties of ? elementary trees ⇐ = linguistic objects ⇒ Syntactic design principles from Frank (2002): Lexicalization Fundamental TAG Hypothesis (FTH) Condition on Elementary Tree Minimality (CETM) θ -Criterion for TAG Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 34 10

Linguistic analyses with LTAG What is an elementary tree, and what is its shape? syntactic/semantic properties of ? elementary trees ⇐ = linguistic objects ⇒ Syntactic design principles from Frank (2002): Lexicalization Fundamental TAG Hypothesis (FTH) Condition on Elementary Tree Minimality (CETM) θ -Criterion for TAG ⇒ Semantic design principles [Abeillé & Rambow (2000)] Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 35 10

Linguistic analyses with LTAG What is an elementary tree, and what is its shape? syntactic/semantic properties of ? elementary trees ⇐ = linguistic objects ⇒ Syntactic design principles from Frank (2002): Lexicalization Fundamental TAG Hypothesis (FTH) Condition on Elementary Tree Minimality (CETM) θ -Criterion for TAG ⇒ Semantic design principles [Abeillé & Rambow (2000)] ⇒ Design principle of economy Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 36 10

Syntactic design principles (1): Lexicalization Each elementary tree has at least one non-empty lexical item, its lexical anchor . Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 37 11

Syntactic design principles (1): Lexicalization Each elementary tree has at least one non-empty lexical item, its lexical anchor . ⇒ All widely used grammar formalisms support some kind of lexicalization! Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 38 11

Syntactic design principles (1): Lexicalization Each elementary tree has at least one non-empty lexical item, its lexical anchor . ⇒ All widely used grammar formalisms support some kind of lexicalization! ⇒ TAG → LTAG: Lexicalized Tree-Adjoining Grammar Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 39 11

Syntactic design principles (1): Lexicalization Each elementary tree has at least one non-empty lexical item, its lexical anchor . ⇒ All widely used grammar formalisms support some kind of lexicalization! ⇒ TAG → LTAG: Lexicalized Tree-Adjoining Grammar [Schabes & Joshi (1990); Joshi & Schabes (1991)] Recall: reasons for lexicalization Formal properties: A finite lexicalized grammar provides finitely many analyses for each string (finitely ambiguous). Linguistic properties: Syntactic properties of lexical items can be accounted for more directly. Parsing: The search space during parsing can be delimited (grammar filtering). Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 40 11

Syntactic design principles (2): FTH Fundamental TAG Hypothesis (FTH); [Frank (2002)] Every syntactic dependency is expressed locally within an elemen- tary tree. Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 41 12

Syntactic design principles (2): FTH Fundamental TAG Hypothesis (FTH); [Frank (2002)] Every syntactic dependency is expressed locally within an elemen- tary tree. “syntactic dependency” valency/subcategorization binding filler-gap constructions ... Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 42 12

Syntactic design principles (2): FTH Fundamental TAG Hypothesis (FTH); [Frank (2002)] Every syntactic dependency is expressed locally within an elemen- tary tree. “syntactic dependency” valency/subcategorization binding filler-gap constructions ... “expressed within an elementary tree” terminal leaf (i.e. lexical anchor) nonterminal leaf (substitution node and footnode) marking an inner node for obligatory adjunction Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 43 12

Syntactic design principles (2): FTH Examples of ill-formed elementary trees: S S VP VP NP ↓ NP ↓ V V NP ε ate persuades Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 44 13

Complex primitives Joshi (2004): Complicate locally, simplify globally. “[...] start with complex (more complicated) primitives, which capture directly some crucial linguistic properties and then introduce some general operations for composing these complex structures (primitive or derived). What is the nature of these complex primitives? In the conventional approach the primitive structures (or rules) are kept as simple as possible. This has the consequence that information (e.g., syntactic and semantic) about a lexical item (word) is distributed over more than one primitive structure. Therefore, the information associated with a lexical item is not captured locally, i.e., within the domain of a primitive structure.” [Joshi (2004)] Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 45 14

Syntactic design principles (3): CETM Condition on Elementary Tree Minimality (CETM); ; [Frank (2002)] The syntactic heads in an elementary tree and their projections must form the extended projection of a single lexical head. Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 46 15

Syntactic design principles (3): CETM Condition on Elementary Tree Minimality (CETM); ; [Frank (2002)] The syntactic heads in an elementary tree and their projections must form the extended projection of a single lexical head. Examples of ill-formed elementary trees: S S NP ↓ VP NP ↓ VP AP ↓ VP V NP ↓ S V persuades NP VP ε arrived V NP ↓ to eat Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 47 15

Syntactic design principles (4): θ -Criterion for TAG Thematic role ( θ -role) the semantic relationship of an argument with its predicate is ex- pressed through the assignment of a role by the predicate to the argument. Different theta-roles have different labels, such as A gent, Theme, Patient, Goal, Source, Experiencer etc. Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 48 16

Syntactic design principles (4): θ -Criterion for TAG Thematic role ( θ -role) the semantic relationship of an argument with its predicate is ex- pressed through the assignment of a role by the predicate to the argument. Different theta-roles have different labels, such as Agent, Theme, Patient, Goal, Source, Experiencer etc. example: Bart kicked the ball. kicked � predicate Bart � Agent ball � Theme/Patient Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 49 16

Syntactic design principles (4): θ -Criterion for TAG Thematic role ( θ -role) the semantic relationship of an argument with its predicate is ex- pressed through the assignment of a role by the predicate to the argument. Different theta-roles have different labels, such as Agent, Theme, Patient, Goal, Source, Experiencer etc. example: Bart kicked the ball. kicked � predicate Bart � Agent ball � Theme/Patient The ball was kicked by Bart. kicked � predicate Bart � Agent ball � Theme/Patient Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 50 16

Syntactic design principles (4): θ -Criterion for TAG θ -Criterion (TAG version) a. If H is the lexical head of an elementary tree T, H assigns all of its θ -roles in T. b. If A is a frontier non-terminal of elementary tree T, A must be assigned a θ -role in T. [Frank (2002)] = ⇒ Valency/subcategorization is expressed only with nonterminal leaves! Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 51 17

Syntactic design principles (4): θ -Criterion for TAG θ -Criterion (TAG version) a. If H is the lexical head of an elementary tree T, H assigns all of its θ -roles in T. b. If A is a frontier non-terminal of elementary tree T, A must be assigned a θ -role in T. [Frank (2002)] = ⇒ Valency/subcategorization is expressed only with nonterminal leaves! S S VP VP NP ↓ NP ↓ S ∗ V V try sleeps Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 52 17

Further design principles Semantic design principles Predicate-argument co-occurrence: Each elementary tree associated with a predicate contains a non- terminal leaf for each of its arguments. Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 53 18

Further design principles Semantic design principles Predicate-argument co-occurrence: Each elementary tree associated with a predicate contains a non- terminal leaf for each of its arguments. Semantic anchoring: Elementary trees are not semantically void (to, that.) Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 54 18

Further design principles Semantic design principles Predicate-argument co-occurrence: Each elementary tree associated with a predicate contains a non- terminal leaf for each of its arguments. Semantic anchoring: Elementary trees are not semantically void (to, that.) Compositional principle: An elementary tree corresponds to a single semantic unit. Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 55 18

Further design principles Semantic design principles Predicate-argument co-occurrence: Each elementary tree associated with a predicate contains a non- terminal leaf for each of its arguments. Semantic anchoring: Elementary trees are not semantically void (to, that.) Compositional principle: An elementary tree corresponds to a single semantic unit. Design principle of economy The elementary trees are shaped in such a way, that the size of the elementary trees and the size of the grammar is minimal. Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 56 18

Modification and functional elements How to insert modifiers (e.g. easily ) and functional elements (complementizers, determiners, do-auxiliaries, ...)? Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 57 19

Modification and functional elements How to insert modifiers (e.g. easily ) and functional elements (complementizers, determiners, do-auxiliaries, ...)? either as co-anchor in the elementary tree of the lexical item they are associated with S S Comp S VP NP ↓ VP VP AP that NP ↓ sleeps sleeps easily Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 58 19

Modification and functional elements How to insert modifiers (e.g. easily ) and functional elements (complementizers, determiners, do-auxiliaries, ...)? either as co-anchor in the elementary tree of the lexical item they are associated with S S Comp S VP NP ↓ VP VP AP that NP ↓ sleeps sleeps easily or by separate auxiliary trees (e.g., XTAG grammar) S VP ⇒ Footnodes/Adjunctions indicate both complementation and modification. S ∗ VP ∗ Comp AP ⇒ Enhancement of the CETM: [see that easily Abeillé & Rambow (2000)] Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 59 19

Outline of today’s course The derivation tree 1 Design principles for elementary trees 2 Sample derivations 3 NP and PP complements Sentential complements and long-distance dependencies Modifiers Feature based TAG 4 Summary and outlook 5 Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 60 20

Sample derivations: NP and PP complements (1) Adam gave Eve the apple. Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 61 21

Sample derivations: NP and PP complements (1) Adam gave Eve the apple. Elementary trees: S NP NP NP NP NP ∗ NP ↓ VP N N N Det V NP ↓ NP ↓ Adam Eve apple the gave Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 62 21

Sample derivations: NP and PP complements (1) Adam gave Eve the apple. Elementary trees: S NP NP NP NP NP ∗ NP ↓ VP N N N Det V NP ↓ NP ↓ Adam Eve apple the gave Derivation tree: gave 1 23 22 adam eve apple 0 the Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 63 21

Sample derivations: NP and PP complements (2) Adam gave the apple to Eve. Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 64 22

Sample derivations: NP and PP complements (2) Adam gave the apple to Eve. Elementary trees: S NP NP NP NP NP ∗ NP ↓ VP N N N Det V NP ↓ PP Adam Eve apple the gave P NP ↓ to Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 65 22

Sample derivations: NP and PP complements (2) Adam gave the apple to Eve. Elementary trees: S NP NP NP NP NP ∗ NP ↓ VP N N N Det V NP ↓ PP Adam Eve apple the gave P NP ↓ to Derivation tree: gave 1 22 232 adam apple eve 0 the Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 66 22

Sample derivations: Sentential complements (3) Adam hopes that Eve comes. Elementary trees: S S NP NP S VP VP N N Comp S ∗ NP ↓ NP ↓ V S ∗ V Eve Adam that comes hopes Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 67 23

Sample derivations: Sentential complements (3) Adam hopes that Eve comes. Elementary trees: S S NP NP S VP VP N N Comp S ∗ NP ↓ NP ↓ V S ∗ V Eve Adam that comes hopes Derivation tree: comes 0 1 eve that 0 hopes 1 adam Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 68 23

Sample derivations: long-distance dependency (4) What i did Adam say (that) Eve ate _ i ? NP S NP NP S N Aux S N N NP i ↓ S VP Eve COMP S what did NP ↓ Adam ϵ V S ∗ VP NP ↓ say V NP ϵ i ate Derivation tree: ate 221 1 2 did_say eve what 21 adam Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 69 24

Sample derivations: Modifiers (5) The good student participated in every course during the semester. Elementary trees: S NP NP NP NP ↓ VP N N N course V PP semester student part. P NP ↓ NP NP N VP in NP ∗ NP ∗ N ∗ VP ∗ Det Det AP PP every the good P NP ↓ during Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 70 25

Sample derivations: Modifiers Derivation tree: part_in 2 1 222 course during stud 1 0 0 22 every good the semester 0 the Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 71 26

Outline of today’s course The derivation tree 1 Design principles for elementary trees 2 Sample derivations 3 NP and PP complements Sentential complements and long-distance dependencies Modifiers Feature based TAG 4 Summary and outlook 5 Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 72 27

Feature structures Idea: Instead of atomic categorial symbols, feature structures are used as non-terminal nodes. Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 73 28

Feature structures Idea: Instead of atomic categorial symbols, feature structures are used as non-terminal nodes. Two reasons: generalizing agreement and case marking (via underspecifica- tion) modelling adjunction constraints (TAG specific) ⇒ smaller grammars that are easier to maintain Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 74 28

Feature structures Idea: Instead of atomic categorial symbols, feature structures are used as non-terminal nodes. Two reasons: generalizing agreement and case marking (via underspecifica- tion) modelling adjunction constraints (TAG specific) ⇒ smaller grammars that are easier to maintain case assignment: Joe saw her. / *Joe saw she. Joe expected her to come. (ECM) / *Joe expected she to come. Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 75 28

Feature structures Idea: Instead of atomic categorial symbols, feature structures are used as non-terminal nodes. Two reasons: generalizing agreement and case marking (via underspecifica- tion) modelling adjunction constraints (TAG specific) ⇒ smaller grammars that are easier to maintain case assignment: Joe saw her. / *Joe saw she. Joe expected her to come. (ECM) / *Joe expected she to come. person/number agreement: You sing. / *You sings. She sings. / *She sing. This woman sings. / *This woman sing. These women sing. / *These women sings. Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 76 28

Feature structures Idea: Instead of atomic categorial symbols, feature structures are used as non-terminal nodes. Two reasons: generalizing agreement and case marking (via underspecifica- tion) modelling adjunction constraints (TAG specific) ⇒ smaller grammars that are easier to maintain case assignment: Joe saw her. / *Joe saw she. Joe expected her to come. (ECM) / *Joe expected she to come. person/number agreement: You sing. / *You sings. She sings. / *She sing. This woman sings. / *This woman sing. These women sing. / *These women sings. Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 77 28

Features structures a list of features (e.g. case ) and values (e.g. nom) Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 78 29

Features structures a list of features (e.g. case ) and values (e.g. nom) feature structures are ofen represented as atribute value matrices (AVM) Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 79 29

Features structures a list of features (e.g. case ) and values (e.g. nom) feature structures are ofen represented as atribute value matrices (AVM) sings:    cat V      vform finite             num sg     agr         pers 3     Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 80 29

Features structures a list of features (e.g. case ) and values (e.g. nom) feature structures are ofen represented as atribute value matrices (AVM) sings:    cat V      vform finite             num sg     agr         pers 3     feature values: atomic (e.g. for vform ) feature structures (e.g. for agr ) Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 81 29

Features structures a list of features (e.g. case ) and values (e.g. nom) feature structures are ofen represented as atribute value matrices (AVM) sings:    cat V      vform finite             num sg     agr         pers 3     feature values: atomic (e.g. for vform ) feature structures (e.g. for agr ) combining constituents ⇒ unify their feature structures Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 82 29

Unification unification is a (partial) operation on feature structures intuitively: the operation of combining two feature structures such that the new feature structure contains all the information of the original two, and nothing more Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 83 30

Unification unification is a (partial) operation on feature structures intuitively: the operation of combining two feature structures such that the new feature structure contains all the information of the original two, and nothing more     cat vp           cat vp cat vp           � � � � e.g.       = ⊔  num pl          agr num pl agr pers 3 agr               pers 3       Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 84 30

Unification unification is a (partial) operation on feature structures intuitively: the operation of combining two feature structures such that the new feature structure contains all the information of the original two, and nothing more     cat vp           cat vp cat vp           � � � � e.g.       = ⊔  num pl          agr num pl agr pers 3 agr               pers 3       partial operation ⇒ unification can fail         cat np cat np     e.g. ⊔ = FAIL         num sg num pl     Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 85 30

Unification unification is a (partial) operation on feature structures intuitively: the operation of combining two feature structures such that the new feature structure contains all the information of the original two, and nothing more     cat vp           cat vp cat vp           � � � � e.g.       = ⊔  num pl          agr num pl agr pers 3 agr               pers 3       partial operation ⇒ unification can fail         cat np cat np     e.g. ⊔ = FAIL         num sg num pl     underspecified feature values             cat np cat np cat np       e.g. ⊔ =             case nom | acc case acc case acc       Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 86 30

Unification: definition Unification ( F ⊔ G ) The unification of two feature structures F and G is (if it exists) the smallest feature structure that is subsumed by both F and G . Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 87 31

Unification: definition Unification ( F ⊔ G ) The unification of two feature structures F and G is (if it exists) the smallest feature structure that is subsumed by both F and G . That is, (if it exists) F ⊔ G is the feature structure with the following three properties: Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 88 31

Unification: definition Unification ( F ⊔ G ) The unification of two feature structures F and G is (if it exists) the smallest feature structure that is subsumed by both F and G . That is, (if it exists) F ⊔ G is the feature structure with the following three properties: (1) F ⊑ ( F ⊔ G ) Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 89 31

Unification: definition Unification ( F ⊔ G ) The unification of two feature structures F and G is (if it exists) the smallest feature structure that is subsumed by both F and G . That is, (if it exists) F ⊔ G is the feature structure with the following three properties: (1) F ⊑ ( F ⊔ G ) (2) G ⊑ ( F ⊔ G ) Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 90 31

Unification: definition Unification ( F ⊔ G ) The unification of two feature structures F and G is (if it exists) the smallest feature structure that is subsumed by both F and G . That is, (if it exists) F ⊔ G is the feature structure with the following three properties: (1) F ⊑ ( F ⊔ G ) (2) G ⊑ ( F ⊔ G ) (3) If H is a feature structure such that F ⊑ H and G ⊑ H , then ( F ⊔ G ) ⊑ H . If there is no smallest feature structure that is subsumed by both F and G , then we say that F ⊔ G is undefined. Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 91 31

Unification: definition Unification ( F ⊔ G ) The unification of two feature structures F and G is (if it exists) the smallest feature structure that is subsumed by both F and G . That is, (if it exists) F ⊔ G is the feature structure with the following three properties: (1) F ⊑ ( F ⊔ G ) (2) G ⊑ ( F ⊔ G ) (3) If H is a feature structure such that F ⊑ H and G ⊑ H , then ( F ⊔ G ) ⊑ H . If there is no smallest feature structure that is subsumed by both F and G , then we say that F ⊔ G is undefined. Subsumption ( F 1 ⊑ F 2 ) A feature structure F 1 subsumes ( ⊑ ) another feature structure F 2 , iff all the information that is contained in F 1 is also contained in F 2 . Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 92 31

Reentrancies the paths that both lead to the same node ⇒ to the same value Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 93 32

Reentrancies the paths that both lead to the same node ⇒ to the same value ⇒ hence, they share that value Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 94 32

Reentrancies the paths that both lead to the same node ⇒ to the same value ⇒ hence, they share that value this property of sharing value(s) is called reentrancy Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 95 32

Reentrancies the paths that both lead to the same node ⇒ to the same value ⇒ hence, they share that value this property of sharing value(s) is called reentrancy in AVMs: expressed by coindexing the shared values (boxed numbers) Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 96 32

Reentrancies the paths that both lead to the same node ⇒ to the same value ⇒ hence, they share that value this property of sharing value(s) is called reentrancy in AVMs: expressed by coindexing the shared values (boxed numbers) within feature structures:     � �� �     1 val 1 attr 1 attr 1 1     attr 1 attr 2     1 1     attr 2 attr 2 1 1     Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 97 32

Reentrancies FTAG uses acyclic reentrancies! Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 98 33

Reentrancies FTAG uses acyclic reentrancies! between feature structures (in a tree): Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 99 33

Reentrancies FTAG uses acyclic reentrancies! between feature structures (in a tree): � � � � attr 1 attr 1 1 1 Kallmeyer, Lichte, Osswald & Petitjean (HHU Düsseldorf) 100 33