Predicting implicit and explicit questions Matthijs Westera COLT - PowerPoint PPT Presentation

Predicting implicit and explicit questions Matthijs Westera COLT kick-off workshop

Predicting implicit and explicit questions Matthijs Westera COLT kick-off workshop = Computational Linguistics & Linguistic Theory

The data of linguistic theory:

The data of linguistic theory: g m a t i c s ) n t i c s & P r a ( S e m a

The data of computational ling:

Main interest messy crisp

Main interest messy crisp enables

Main interest messy crisp enables How?

Main interest messy crisp enables How? - ... ? - ... ? - ... ?

Neural networks

Neural networks crisp

Neural networks messy crisp

Neural networks messy crisp ? ?

Neural networks messy crisp ? ? cf. Aina et al. (2018) [SemEval]

Distributional & formal semantics

Distributional & formal semantics cat animal red

Distributional & formal semantics messy cat animal red

Distributional & formal semantics messy ∃ x (R ED ( x ) ∧ C AT ( x )) cat animal red

Distributional & formal semantics messy crisp ∃ x (R ED ( x ) ∧ C AT ( x )) cat animal red

Distributional & formal semantics messy crisp ∃ x (R ED ( x ) ∧ C AT ( x )) cat animal ? ? red

Distributional & formal semantics messy crisp ∃ x (R ED ( x ) ∧ C AT ( x )) cat animal ? ? red Westera & Boleda (under review)

Main interest messy crisp enables How?

Main interest messy crisp enables How? - how do NN models do it?

Main interest messy crisp enables How? - how do NN models do it? - how do DS and FS relate?

Main interest messy crisp enables How? - how do NN models do it? - how do DS and FS relate? - what does ling. theory say?

Pragmatics

Pragmatics € PAINT

Pragmatics

Main interest messy crisp enables How? - how do NN models do it? - how do DS and FS relate? - what does ling. theory say?

Main interest messy crisp enables How? - how do NN models do it? - how do DS and FS relate? - ...

Main interest messy crisp enables How? - how do NN models do it? - how do DS and FS relate? - how are goals identifjed?

Discourse goals Pragmatic theory:

Discourse goals Pragmatic theory: ● Discourse is organized around questions:

Discourse goals Pragmatic theory: e.g., Westera 2017 (PhD thesis) ● Discourse is organized around questions:

Discourse goals Pragmatic theory: e.g., Westera 2017 (PhD thesis) ● Discourse is organized around questions: – Some explicit (“What color do you want?”) – Most are implicit:

Discourse goals Pragmatic theory: e.g., Westera 2017 (PhD thesis) ● Discourse is organized around questions: – Some explicit (“What color do you want?”) – Most are implicit: So, there's this party tonight... August, no! But it's perfect! It’s my last night in Aberdeen! I'll never be in this town again! No, we can't. BookCorpus/590124__come-as-you-are.txt

Discourse goals Pragmatic theory: e.g., Westera 2017 (PhD thesis) ● Discourse is organized around questions: – Some explicit (“What color do you want?”) – Most are implicit: ? ? o g So, there's this party tonight... e w August, no! l l a h But it's perfect! It’s my last night in Aberdeen! I'll never be in this S town again! No, we can't. BookCorpus/590124__come-as-you-are.txt

Question prediction Get a model to predict implicit questions (goals), by training on explicit questions.

Question prediction Get a model to predict implicit questions (goals), by training on explicit questions. = f v i s i o n ! r e e s u p e r

Question prediction Get a model to predict implicit questions (goals), by training on explicit questions. = f v i s i o n ! r e e s u p e r Data: Only dialogue contains enough questions. ●

Question prediction Get a model to predict implicit questions (goals), by training on explicit questions. = f v i s i o n ! r e e s u p e r Data: Only dialogue contains enough questions. ● Extract dialogues from movie scripts and books. ● – E.g., 1M dialogues from BookCorpus.

Question prediction Get a model to predict implicit questions (goals), by training on explicit questions. = f v i s i o n ! r e e s u p e r Data: Only dialogue contains enough questions. ● Extract dialogues from movie scripts and books. ● – E.g., 1M dialogues from BookCorpus. Models (e.g.): Simple word-level RNN. (Westera 2018 [SemDial]) ●

Question prediction Get a model to predict implicit questions (goals), by training on explicit questions. = f v i s i o n ! r e e s u p e r Data: Only dialogue contains enough questions. ● Extract dialogues from movie scripts and books. ● – E.g., 1M dialogues from BookCorpus. Models (e.g.): Simple word-level RNN. (Westera 2018 [SemDial]) ● Sentence-level RNN on pre-trained sent. embeddings. ●

Example (aim) So, there's this party tonight… So, there's this party tonight… August, no! August, no! But it's perfect! It’s my last night in Aberdeen! I'll never be in this But it's perfect! It’s my last night in Aberdeen! I'll never be in this town again! town again! BookCorpus/590124__come-as-you-are.txt No, we can't. No, we can't.

Example (aim) So, there's this party tonight… So, there's this party tonight… Shall we go? Shall we go? August, no! August, no! But it's perfect! It’s my last night in Aberdeen! I'll never be in this But it's perfect! It’s my last night in Aberdeen! I'll never be in this town again! town again! BookCorpus/590124__come-as-you-are.txt No, we can't. No, we can't.

Example (aim) So, there's this party tonight… So, there's this party tonight… Shall we go? Shall we go? August, no! August, no! Why do you even bring it up? Why do you even bring it up? But it's perfect! It’s my last night in Aberdeen! I'll never be in this But it's perfect! It’s my last night in Aberdeen! I'll never be in this town again! town again! BookCorpus/590124__come-as-you-are.txt No, we can't. No, we can't.

Example (aim) So, there's this party tonight… So, there's this party tonight… Shall we go? Shall we go? August, no! August, no! Why do you even bring it up? Why do you even bring it up? But it's perfect! It’s my last night in Aberdeen! I'll never be in this But it's perfect! It’s my last night in Aberdeen! I'll never be in this town again! town again! BookCorpus/590124__come-as-you-are.txt Can we go? Can we go? No, we can't. No, we can't.

Example (aim) So, there's this party tonight… So, there's this party tonight… Shall we go? Shall we go? August, no! August, no! Why do you even bring it up? Why do you even bring it up? But it's perfect! It’s my last night in Aberdeen! I'll never be in this But it's perfect! It’s my last night in Aberdeen! I'll never be in this town again! town again! BookCorpus/590124__come-as-you-are.txt Can we go? Can we go? No, we can't. No, we can't. Why can’t we go? Why can’t we go?

Summary messy crisp enables - how do NN models do it? How? - how do DS and FS relate? - can we predict discourse goals?

Predicting implicit and explicit questions Matthijs Westera COLT - PowerPoint PPT Presentation

Predicting implicit and explicit questions Matthijs Westera COLT kick-off workshop Predicting implicit and explicit questions Matthijs Westera COLT kick-off workshop = Computational Linguistics & Linguistic Theory The data of

Not Smooth High degree approximation Explicit y=f(x) Implicit f(x,y)=0 Parametric

Session 14: Explicit and Implicit Cooperation 1. Nash equilibrium versus collective action.

Implicit Graphs Implicit Graph: Only a subset, possibly only one, of the vertices is

2.1 Explicit & Implicit Surfaces Hao Li http://cs599.hao-li.com 1 Administrative

2.1 Explicit & Implicit Surfaces Hao Li http://cs621.hao-li.com 1 Administrative

Combining Implicit and Explicit Topic Representations for Result Diversification Jiyin He, Vera

Explicit vs Implicit CommunicaGon in Control Anant Sahai UC

Higher prices Hotels set prices Higher prices help others Low speeds Driving Higher speeds

A COMPARISON OF IMPLICIT AND EXPLICIT FINITE ELEMENT METHODS FOR THE MESO-LEVEL SIMULATION OF DRY

HYPERBOLIC EXPLICIT-PARABOLIC LINEARLY IMPLICIT FINITE DIFFERENCE METHODS FOR DEGENERATE

Second order Implicit-Explicit Total Variation Diminishing schemes for the Euler system in the

Second order Implicit-Explicit Total Variation Diminishing schemes for the Euler system in the

Implicit-Explicit Runge-Kutta schemes for the Boltzmann-Poisson equation for semiconductors

GCMs with Implicit and Explicit Representation of Cloud Microphysics: Simulation of Extreme

From Explicit to Implicit Dynamic Frames in Java Dynamic Logic and KeY . Wojciech Mostowski

Implicit and Explicit Risk Perception, Affect, and Trust: An Investigation of Food Traffic

Explicit and Implicit Discourse Relations: An Extrinsic Evaluation Peter Bourgonje and Manfred

Making the implicit explicit: Theories informing simulation-based education #simtheory

Explicit vs. Implicit Parallel Programming Language, Directive, Library Expose, Express,

IMplicit-EXplicit schemes for BGK kinetic equations Sandra Pieraccini, Gabriella Puppo

A Comparison of Implicit and Explicit Links for Web Page Classification Dou Shen 1 Jian-Tao Sun 2

Shonan workshop number 90, 22-25 November 2016 Implicit and explicit semantics integration in

4/5/2017 Explicit and Implicit Biases in Early Childhood Education Becoming Aware of

@ .To increase understanding of the The difference between "conscious/explicit behavior"

Predicting implicit and explicit questions Matthijs Westera COLT - PowerPoint PPT Presentation

Predicting implicit and explicit questions Matthijs Westera COLT kick-off workshop Predicting implicit and explicit questions Matthijs Westera COLT kick-off workshop = Computational Linguistics & Linguistic Theory The data of

Not Smooth High degree approximation Explicit y=f(x) Implicit f(x,y)=0 Parametric

Session 14: Explicit and Implicit Cooperation 1. Nash equilibrium versus collective action.

Implicit Graphs Implicit Graph: Only a subset, possibly only one, of the vertices is

2.1 Explicit &amp; Implicit Surfaces Hao Li http://cs599.hao-li.com 1 Administrative

2.1 Explicit &amp; Implicit Surfaces Hao Li http://cs621.hao-li.com 1 Administrative

Combining Implicit and Explicit Topic Representations for Result Diversification Jiyin He, Vera

Explicit vs Implicit CommunicaGon in Control Anant Sahai UC

Higher prices Hotels set prices Higher prices help others Low speeds Driving Higher speeds

A COMPARISON OF IMPLICIT AND EXPLICIT FINITE ELEMENT METHODS FOR THE MESO-LEVEL SIMULATION OF DRY

HYPERBOLIC EXPLICIT-PARABOLIC LINEARLY IMPLICIT FINITE DIFFERENCE METHODS FOR DEGENERATE

Second order Implicit-Explicit Total Variation Diminishing schemes for the Euler system in the

Second order Implicit-Explicit Total Variation Diminishing schemes for the Euler system in the

Implicit-Explicit Runge-Kutta schemes for the Boltzmann-Poisson equation for semiconductors

GCMs with Implicit and Explicit Representation of Cloud Microphysics: Simulation of Extreme

From Explicit to Implicit Dynamic Frames in Java Dynamic Logic and KeY . Wojciech Mostowski

Implicit and Explicit Risk Perception, Affect, and Trust: An Investigation of Food Traffic

Explicit and Implicit Discourse Relations: An Extrinsic Evaluation Peter Bourgonje and Manfred

Making the implicit explicit: Theories informing simulation-based education #simtheory

Explicit vs. Implicit Parallel Programming Language, Directive, Library Expose, Express,

IMplicit-EXplicit schemes for BGK kinetic equations Sandra Pieraccini, Gabriella Puppo

A Comparison of Implicit and Explicit Links for Web Page Classification Dou Shen 1 Jian-Tao Sun 2

Shonan workshop number 90, 22-25 November 2016 Implicit and explicit semantics integration in

4/5/2017 Explicit and Implicit Biases in Early Childhood Education Becoming Aware of

@ .To increase understanding of the The difference between &quot;conscious/explicit behavior&quot;

2.1 Explicit & Implicit Surfaces Hao Li http://cs599.hao-li.com 1 Administrative

2.1 Explicit & Implicit Surfaces Hao Li http://cs621.hao-li.com 1 Administrative

@ .To increase understanding of the The difference between "conscious/explicit behavior"