Understanding Natural Language with Functional Discourse Grammar
Recommend Documents
May 5, 2011 ... for achieving natural language understanding through a capable grammatical
framework. The UNL system utilizes a processing work-.
latest manifestation of Simon Dikâs. Functional Grammar, and it appears as a response to criticism of the earlier model, particularly in reaction to the small part.
John H. CONNOLLY. 1. ⢠ABSTRACT: An important and ...... In: OWEN, J. L. Context and communication behaviour. Reno: Context Press, 1997. p.97-110.
Jul 6, 2015 - Logical forms? What is the largest city in California? argmax(λx.city(x) < loc(x,CA), λx.population(x
The integration of speech recognition with natural language understanding
raises issues of how to adapt natural language processing to the characteristics
of ...
Sep 5, 2013 - Building Automatic Indonesian Mind Map Generator ... that employs several Indonesian natural language understanding tools as its main.
generation of flexible natural language utterances which pin- point possible problems with a user's request and which fur- ther go on to outline the user's most ...
formation on Conferences) comprising a dialog man- ager which is able to ..... Procceedings of SIGIR '91, October 13-16, 1991, Chica- go, USA. Wahlster, W.
tion of natural language generation to augment the interaction functionalities of .... ticularly text involving controlled and appropriate deployment of text-forming ...
from The School of Electronic Engineering, Dublin City University, Dublin, Ireland, EC. 1 ..... If you speak to your car mechanic about what is wrong with your.
Nov 21, 2007 - Tell us about your experience on SpringerLink and you could win a MacBook Pro. ... We use cookies to improve your experience with our site.
Indian Institute of Information Technology, âInnovatorâ Towers, ITPL,. Whitefield ... grammar of a poetry construct of the Tamil language as a context free grammar (CFG). This ... Terminals define the basic symbols of which strings in the languag
We would equally like a number of functional programming language features for this application. Checking for type equality (or unifying two types) without.
IVANA LUČIĆA 3, 10 000 ZAGREB, CROATIA, E-MAIL: [email protected] ...
The practical part of the LFG model consists of the computer program written in ...
using the inside-outside algorithm. Computer Speech and Language, 4:35â56, 1990. [12] David J.C. MacKay. em ensemble learning for hidden markov models.
Mar 1, 2017 - The aim of this paper is to provide a unified analysis of partitive constructions ... Keywords: English Partitives; Functional Discourse Grammar; ...
Page 3. DISCOURSE GRAMMAR of the GREEK NEW TESTAMENT iii ..... goal of
this project is to break this trend, providing practical solutions to grammatical.
Introduction to Discourse Analysis: Theory and Method (1999) which serve as a
... of Frances Christie (2002) who used the functional grammar framework in ...
12 An asymmetrical approach to markedness views members of a given set as each uniquely marking the presence of some dis
8 | English1,2,3. ⢠We commonly use the simple present to ask and answer questions which begin with âHow oftenâ e.g How often do you go to the dentist?
Spontaneous, unplanned spoken and formal written language overlap, but also
differ substantially in the ... In spoken English that and zero are the most common
relativisers: (1) I liked the book ..... Oxford: Oxford University Press. Miller, J. and
19. Functional Grammar. The term functional grammar has been used before,
notably by Dik (1978). I risk adding to the number of its meanings here, and thus
...
and W. J. Wilbur. Automatic ..... Alex, and Claire Grover. Exploring the .... Hong Yu, Vasileios Hatzivassiloglou, Andrey Rzhetsky, and W. John Wilbur. Au-.
Understanding Natural Language with Functional Discourse Grammar
13 Feb 2009 ... 5 Demonstration ... Original Functional Grammar by Simon Dik (1940–1995).
Functional Discourse Grammar introduced by Hengeveld &.
Understanding Natural Language with Functional Discourse Grammar
Understanding Natural Language with Functional Discourse Grammar Reinier Lamers Universiteit Utrecht
February 13, 2009
Understanding Natural Language with Functional Discourse Grammar
Outline
1
Introduction
2
Relevance
3
Functional Discourse Grammar
4
The Algorithm
5
Demonstration
6
Conclusions
Understanding Natural Language with Functional Discourse Grammar Introduction
The Project
I am writing my thesis at the Intelligent Systems group They are designing a companion robot for the kitchen domain
←− They’re using Philips’s iCat
It has to understand language And I have to make the “understanding” part
Understanding Natural Language with Functional Discourse Grammar Introduction
Converting Syntax to Semantics
“Understanding” means here: “converting syntax to semantics”. Syntax:
Semantics: ∀x∃y (. . .) Why design it ourselves?
Understanding Natural Language with Functional Discourse Grammar Relevance
Relevance of Converting Syntax to Semantics
Speech Text Syntax Semantics Pragmatics Beliefs, Desires, Intentions
Understanding Natural Language with Functional Discourse Grammar Functional Discourse Grammar
A Short History of Functional Discourse Grammar
Original Functional Grammar by Simon Dik (1940–1995) Functional Discourse Grammar introduced by Hengeveld & Mackenzie, 2008 FDG principles: No transformations or filters Discourse Act as basic unit Single theory of language Maximal Depth Not a Theory of Discourse
Understanding Natural Language with Functional Discourse Grammar Functional Discourse Grammar
The Levels of FDG
FDG analyzes utterances at four Levels That has to do with the Single Theory Principle The four levels are the Phonological, Morphosyntactic, Representational and Interpersonal Levels Syntax to Semantics ∼ Morphosyntactic Level to Representational Level
Understanding Natural Language with Functional Discourse Grammar Functional Discourse Grammar
Tree Notation for FDG Levels Levels consists of layers may contain other layers: Tree Structure How I draw them as trees:
id
operators
head
modifiers
Φ
...
...
...
...
...
Understanding Natural Language with Functional Discourse Grammar The Algorithm
Outline
A quick outline of the algorithm:
Understanding Natural Language with Functional Discourse Grammar The Algorithm
Outline
A quick outline of the algorithm: 1
Run through the Morphosyntactic Level tree and generate pieces of Representational Level tree
Understanding Natural Language with Functional Discourse Grammar The Algorithm
Outline
A quick outline of the algorithm: 1
Run through the Morphosyntactic Level tree and generate pieces of Representational Level tree
2
Try to glue the pieces together in a consistent way
Understanding Natural Language with Functional Discourse Grammar The Algorithm
Outline
A quick outline of the algorithm: 1
Run through the Morphosyntactic Level tree and generate pieces of Representational Level tree
2
Try to glue the pieces together in a consistent way
3
Fill in some references to other parts of the tree
Understanding Natural Language with Functional Discourse Grammar The Algorithm
Outline
A quick outline of the algorithm: 1
Run through the Morphosyntactic Level tree and generate pieces of Representational Level tree
2
Try to glue the pieces together in a consistent way
3
Fill in some references to other parts of the tree
4
Add Prolog’s backtracking to taste
Understanding Natural Language with Functional Discourse Grammar The Algorithm
Treewalk I Run through tree in depth-first order For every node (excluding list or restrictor nodes), apply node function to subtree rooted in it Node function returns list of tree chunks - pieces of the Representational Level Result of treewalk procedure: ordered list of chunks Adjwi
operators
head
fi
modifiers
node function
operators
head
∅
∅
easy
easy
modifiers
Understanding Natural Language with Functional Discourse Grammar The Algorithm
Treewalk II
Lexicon: The node function often consults the lexicon The lexicon contains only truly lexical items It stores identifier, spelling, lexeme class, combination schemes and class-dependent properties
Why such a complex treewalk procedure? Not every utterance is as form-preserving as the previous example... Some features depend on co-occurrence (count/mass)
Understanding Natural Language with Functional Discourse Grammar The Algorithm
Composition
Those tree chunks are glued together during the composition stage Basic procedure: 1 2 3
Start with an empty tree Add pieces to it as they fit, until there are no pieces any more Is the tree complete? If yes, solution. Else, backtrack.
“Fitting” under point 2 above is defined by a mostly language-specific matching function. Universal bits for position, function, more? Matching function compares criteria stored on the chunks. So node function and matching function cooperate.
Understanding Natural Language with Functional Discourse Grammar The Algorithm
Coreference resolution
For resolving anaphora and vaguely similar things: the coreference resolution stage Some nodes marked as dangling references by node function, resolved now Only used for implicit subjects in implementation
Understanding Natural Language with Functional Discourse Grammar Demonstration
Demonstration
Understanding Natural Language with Functional Discourse Grammar Conclusions
Conclusion
The computer program formalizes a linguistic theory that was not very rigorous Using a three-step algorithm with backtracking, we can convert English syntax to semantics It works!
Understanding Natural Language with Functional Discourse Grammar Conclusions
Conclusion
The computer program formalizes a linguistic theory that was not very rigorous Using a three-step algorithm with backtracking, we can convert English syntax to semantics It works! For a small subset of English at least
Understanding Natural Language with Functional Discourse Grammar Conclusions
Discussion
Will it scale to any linguistic phenomenon? Isn’t Functional Discourse Grammar too complex? How does this compare to the Categorial Grammar + Montague-like approach? Isn’t there an easier way to convert one tree to the other?
