features unification
play

Features & Unification Ling 571 Deep Processing Techniques for - PowerPoint PPT Presentation

Features & Unification Ling 571 Deep Processing Techniques for NLP January 31, 2011 Roadmap Features: Motivation Constraint & compactness Features Definitions & representations Unification


  1. Features & Unification Ling 571 Deep Processing Techniques for NLP January 31, 2011

  2. Roadmap — Features: Motivation — Constraint & compactness — Features — Definitions & representations — Unification — Application of features in the grammar — Agreement, subcategorization — Parsing with features & unification — Augmenting the Earley parser, unification parsing — Extensions: Types, inheritance, etc — Conclusion

  3. Constraints & Compactness — Constraints in grammar — S -> NP VP — They run. — He runs.

  4. Constraints & Compactness — Constraints in grammar — S -> NP VP — They run. — He runs. — But… — *They runs — *He run — *He disappeared the flight

  5. Constraints & Compactness — Constraints in grammar — S -> NP VP — They run. — He runs. — But… — *They runs — *He run — *He disappeared the flight — NP -> Det Nom — This flight

  6. Constraints & Compactness — Constraints in grammar — S -> NP VP — They run. — He runs. — But… — *They runs — *He run — *He disappeared the flight — NP -> Det Nom — This flight — These flights

  7. Constraints & Compactness — Constraints in grammar — S -> NP VP — They run. — He runs. — But… — *They runs — *He run — *He disappeared the flight — NP -> Det Nom — This flight — These flights — *This flights

  8. Constraints & Compactness — Constraints in grammar — S -> NP VP — They run. — He runs. — But… — *They runs — *He run — *He disappeared the flight — NP -> Det Nom — This flight — These flights — *This flights — Violate agreement (number), subcategorization

  9. Enforcing Constraints — Enforcing constraints

  10. Enforcing Constraints — Enforcing constraints — Add categories, rules

  11. Enforcing Constraints — Enforcing constraints — Add categories, rules — Agreement: — S-> NPsg3p VPsg3p, — S-> NPpl3p VPpl3p,

  12. Enforcing Constraints — Enforcing constraints — Add categories, rules — Agreement: — S-> NPsg3p VPsg3p, — S-> NPpl3p VPpl3p, — Subcategorization: — VP-> Vtrans NP , — VP -> Vintrans, — VP->Vditrans NP NP

  13. Enforcing Constraints — Enforcing constraints — Add categories, rules — Agreement: — S-> NPsg3p VPsg3p, — S-> NPpl3p VPpl3p, — Subcategorization: — VP-> Vtrans NP , — VP -> Vintrans, — VP->Vditrans NP NP — Explosive!, loses key generalizations

  14. Features — person: 1 st , 2 nd , 3 rd — I, we; you; he, she, they — am, are, is

  15. Features — person: 1 st , 2 nd , 3 rd — I, we; you; he, she, they — am, are, is — number: sg, pl — I am; we are

  16. Features — person: 1 st , 2 nd , 3 rd — I, we; you; he, she, they — am, are, is — number: sg, pl — I am; we are — case: nom, acc — I, he; me, him

  17. Features — person: 1 st , 2 nd , 3 rd — I, we; you; he, she, they — am, are, is — number: sg, pl — I am; we are — case: nom, acc — I, he; me, him — gender: masc, fem, neut

  18. Features — person: 1 st , 2 nd , 3 rd — I, we; you; he, she, they — am, are, is — number: sg, pl — I am; we are — case: nom, acc — I, he; me, him — gender: masc, fem, neut — animacy: +/- — etc

  19. Why features? — Need compact, general constraints — S -> NP VP

  20. Why features? — Need compact, general constraints — S -> NP VP — Only if NP and VP agree

  21. Why features? — Need compact, general constraints — S -> NP VP — Only if NP and VP agree — How can we describe agreement, subcat?

  22. Why features? — Need compact, general constraints — S -> NP VP — Only if NP and VP agree — How can we describe agreement, subcat? — Decompose into elementary features that must be consistent — E.g. Agreement

  23. Why features? — Need compact, general constraints — S -> NP VP — Only if NP and VP agree — How can we describe agreement, subcat? — Decompose into elementary features that must be consistent — E.g. Agreement — Number, person, gender, etc

  24. Why features? — Need compact, general constraints — S -> NP VP — Only if NP and VP agree — How can we describe agreement, subcat? — Decompose into elementary features that must be consistent — E.g. Agreement — Number, person, gender, etc — Augment CF rules with feature constraints — Develop mechanism to enforce consistency — Elegant, compact, rich representation

  25. Feature Representations — Fundamentally, Attribute- Value pairs — Features: atomic symbols from a finite set

  26. Feature Representations — Fundamentally, Attribute- Value pairs — Features: atomic symbols from a finite set — Values may be — Atomic symbols from a finite set Attribute-value matrix (AVM)

  27. Feature Representations — Fundamentally, Attribute- Value pairs NUMBER PL — Features: atomic symbols from a finite set — Values may be — Atomic symbols from a finite set Attribute-value matrix (AVM)

  28. Feature Representations — Fundamentally, Attribute- Value pairs NUMBER PL — Features: atomic symbols from a finite set PERSON 3 — Values may be — Atomic symbols from a finite set Attribute-value matrix (AVM)

  29. Feature Representations — Fundamentally, Attribute- Value pairs NUMBER PL — Features: atomic symbols from a finite set PERSON 3 — Values may be NUMBER PL — Atomic symbols from a finite set PERSON 3 Attribute-value matrix (AVM)

  30. Feature Representations — Fundamentally, Attribute- Value pairs NUMBER PL — Features: atomic symbols from a finite set PERSON 3 — Values may be NUMBER PL — Atomic symbols from a finite set PERSON 3 Attribute-value matrix (AVM) CAT NP NUMBER PL PERSON 3

  31. Feature Representations — Fundamentally, Attribute-Value pairs — Features: atomic symbols from a finite set — Values may be — Atomic symbols from a finite set — Values may also be feature structures themselves Attribute-value matrix (AVM) CAT NP NUMBER PL AGREEMENT PERSON 3

  32. Feature Representations — Feature path: — Sequence of features through a feature structure leading to a particular value CAT NP NUMBER PL AGREEMENT PERSON 3

  33. Feature Representations — Feature path: — Sequence of features through a feature structure leading to a particular value CAT NP NUMBER PL AGREEMENT PERSON 3 <AGREEMENT NUMBER> -> PL

  34. Feature Representations — Feature path: — Sequence of features through a feature structure leading to a particular value CAT NP NUMBER PL AGREEMENT PERSON 3 <AGREEMENT NUMBER> -> PL <AGREEMENT PERSON> -> 3

  35. Feature Representations — Reentrant feature structures — Features share some feature structure as value — Not merely equal values — Shared substructure — Feature paths lead to same node CAT S NUMBER PL HEAD AGREEM ’ T 1 PERSON 3 SUBJECT AGREEMENT 1

  36. Head-Subject Agreement CAT S NUMBER PL 1 HEAD AGREEM ’ T PERSON 3 SUBJECT AGREEMENT 1

  37. Feature representations — Feature structures can also be represented as DAGs — Directed, acyclic graphs — Edges are features — Nodes values

  38. Reentrant DAG

  39. Unification — Two key roles:

  40. Unification — Two key roles: — Merge compatible feature structures

  41. Unification — Two key roles: — Merge compatible feature structures — Reject incompatible feature structures

  42. Unification — Two key roles: — Merge compatible feature structures — Reject incompatible feature structures — Two structures can unify if

  43. Unification — Two key roles: — Merge compatible feature structures — Reject incompatible feature structures — Two structures can unify if — Feature structures are identical — Result in same structure

  44. Unification — Two key roles: — Merge compatible feature structures — Reject incompatible feature structures — Two structures can unify if — Feature structures are identical — Result in same structure — Feature structures match where both have values, differ in missing or underspecified — Resulting structure incorporates constraints of both

  45. Subsumption — Relation between feature structures — Less specific f.s. subsumes more specific f.s. — F .s. F subsumes f.s. G iff — For every feature x in F , F(x) subsumes G(x) — For all paths p and q in F s.t. F(p)=F(q), G(p)=G(q)

Recommend


More recommend