Surface realization

Surface Realization

Mahalingam.P.R
Semester III, M.Tech CSESIS, RSET

 Introduction Agenda

 Systemic Grammar
 Interpersonal meta-function
 Ideational meta-function
 Textual meta-function

 Functional Unification Grammar
 Functional Description

 Conclusion

2 Surface Realization

Introduction


 Surface realizer receives the fully
specified discourse plan. Discourse Plan
 Generates individual sentences The discourse plan is
generated by the
 Constrained by the lexical and DISCOURE PLANNER
by taking into
grammatical resources consideration the
 Resources communicative goal and
the available Knowledge
 Define the realizer’s potential range of Base.
output The content is structured
appropriately.
 If the plan specifies multiple-
Discourse plan defines:
sentence output, the surface •Choices made for the

realizer is called multiple times. entire
(may
communication
span multiple
sentences)
•Annotations (hypertext,
figures, etc.)


 So, the surface realization component produces
ordered sequence of words as constrained by the
lexicon and grammar.

 Input
 Sentence-sized chunks of the discourse specification

 Influential approaches for surface realization
 Systemic Grammar
 Functional Unification Grammar


 No general consensus as to the level at which the
input to the surface realizer should be specified.

 Some approaches specify only the propositional
content.


What does it do?
 Derive a human readable sentence from a discourse
plan.

 Discourse plan does not give syntax, only functional
information. The Surface Realizer adds syntactical
information and assures that the sentence will
comply with lexical and grammatical constraints.


What doesn’t it do?
 Will not verify that the correctness of the data
provided by the discourse planner or that the
information makes sense.

 Does not deal with more than one sentence at a
time. If the plan calls for many sentences, the
surface realizer will be called once for each sentence
required.


Simple Surface Realization Tools
 Canned Text Systems  Template Systems
- Takes a given input and - The idea of a template is
matches it directly to a pre- that there are premade
made sentence. sentences with fill in the
blank words that are filled
- Commonly used in simple in by the input.
systems such as error
messages or warnings. - These systems work well
- Has no flexibility with Form Letters and
whatsoever. Slightly more advanced
Error or Warning
Messages.
- They are still very
inflexible, but better than
canned text systems.


 The simple surface realization tools eventually gave
way to advanced Feature-based systems

 Systemic Grammar Representation of sentences
as collections of functions. Rules allow mapping
from functions to grammatical forms. (Halliday, 1985)

 Functional Unification Grammar Represents
sentences as feature structures that can be
combined and altered to produce sentences. (Kay,
1979)


“The system will save the document”
 The discourse plan  Other approaches
would specify a saving
action done by a system  Include the specification
entity to a document of the grammatical form
entity.
 In this case, a future
tense assertion
 Specification of lexical
items
 In this case, save,
system and document


 The two approaches take input at different levels.

 Common factor
 Input is functionally specified, rather than syntactically
specified
 Factor typical of generational systems

 Generation systems start with meaning and context
 Specify the intended output in terms of function, rather
than form.


 Can be stated in two ways
 ACTIVE FORM
 PASSIE FORM

 Discourse planners tend not to work with the
syntactic terms.

 They are most likely to keep track of the focus or
local topic of the discourse.
 More natural to define this distinction in terms of focus.


Surface
 If the document is the local topic of Realization
discourse, it would be marked as Approaches
the focus which could trigger the
use of the passive. Systemic Grammar

“The document will be saved by Functional Unification
the system” Grammar

 Both surface realization
approaches categorize grammar in
functional terms.


Systemic Grammar


Systemic-
 A part of Systemic-Functional Functional
linguistics. linguistics

 Represent sentences as collections A branch of linguistics
that views language as a
of functions and maintain rules for resource for expressing
mapping these functions on to meaning in context

explicit grammatical forms.
-An Introduction to
Functional Grammar,
 Well suited for generation
Halliday (1985)
 Widely influential in NLG


 Systemic sentence analysis organize the functions
being expressed in multiple layers.


Mood layer – Layers
simple declarative Transitivity layer Theme layer
structure

“The system will save the
Actor / Doer
Subject
(system)
Theme document”

Process Concepts of theme and
Finite (auxiliary) Rheme
(saving)
rheme were developed by
the Prague school of
Goal – the linguistics
Predicator object being
(verb) acted upon
(document) -Firbas, 1966

Thematic roles apply here
Object too, like AGENT,
Rheme EXPERIENCER,
 a topic of informal discussion INSTRUMENT, and so on.
 different from a theme

 The three layers deal with different sets of functions.
 Meta-functions

• Inter-personal
Mood layer
meta function

Transitivity • Ideational meta
layer function

Theme • Textual meta
layer function


Interpersonal meta-function
 Group the functions that establish and maintain the
interaction between the sentence writer and the
reader.

 Represented by the mood layer
 Determines whether the writer is
 Commanding
 Telling
 Asking

 Examples would be whether the writer is telling the
reader something or is asking a question.


Ideational meta-function
 Concerned with the propositional content of the
expression.

 Transitivity layer determines
 Nature of process being expressed
 Variety of case roles that must be expressed

 Covers much of the semantics.

 In other words, identify items like who the actors are,
what the goals are for the sentence, and type of
process being performed.

Textual meta-function
 Concerned with the way the expression fits into the
current discourse.

 Includes issues of thematization and reference.

 Tries to fit the expression with a given theme and
reference.

 Represented by the theme layer
 Explicitly marks the system as the theme of the sentence


 Explicit concern for interpersonal and textual issues
as well as traditional semantics
 Feature of systemic linguistics that is attractive for NLG.

 Many choices that generation systems make depend
on context of communication
 Formalized by the interpersonal and textual meta-
functions.


System network


Grammar represented using a directed, acyclic and/or
graph, called a system network

 Curly braces
 AND  parallel systems
 Vertical lines
 OR  disjoint systems


 Every clause (represented as the highest level
feature) will simultaneously have a different set of
features for mood, transitivity and theme.

 “The system will save the document”
 Indicative, declarative clause expressing an active
material process with an unmarked theme.


Realization Statements
 A systemic grammar uses realization statements to
map from the features specified in the grammar (like
Indicative, Declarative) to syntactic form.
 Each feature in the network can have a set of
realization statements specifying constraints on the
final form of the expression.
 Shows as italicized statements below each feature
 Realization statements allow the grammar to
constrain the structure of the expression as the
system network is traversed.


Some simple operators
 +X
 Insert the function X
 The grammar here
specifies that all clauses
will have a predicator.


 X=Y
 Conflate the functions X
and Y. This allows the
grammar to build a
layered function
structure by assigning
different functions to the
same portion of the
expression.
 Active clauses conflate
the actor with the subject
 Passive clauses conflate
the goal with the subject


 X>Y
 Order function X
somewhere before
function Y.
 Indicative sentences
place the subject
somewhere before the
predicator.


 X:A
 Classify the function X with the
lexical or grammatical feature
A.
 Signal a recursive pass through
the grammar at a lower level.
 Grammar would include other
networks similar to the clause
network that would apply to
phrases, lexical items and
morphology.
 Indicative feature inserts a
subject function that must be a
noun phrase.
 Phrase further specified by
another pass through the
grammar.


 X!L
 Assign function X the
lexical item L.
 Finite element of the
passive is assigned the
lexical item “be”


Procedure for generation
-Given a fully specified system network
1. Traverse the network from left to right, choosing
the appropriate features and collecting the
associated realization statements.
2. Build an intermediate expression that reconciles
the constraints set by the realization statements
collected during the traversal.
3. Recurse back through the grammar at a lower level
for any function that is not fully specified.


 We can use the following specification as input.

(
:process save-1
:actor system-1
:goal document-1
:speechact assertion
:tense future
)


 save-1 knowledge base
instance is identified as
the process of the
intended expression.
 Assume all knowledge base
objects to be KLONE-styled
instances
 Actor and goal similarly
specified as system-1
and document-1
respectively.
 Input also specifies that
the expression be in the
form of an assertion in
the future tense.


Generation Process
 Start at clause feature
 Insert a predicator
 +predicator
 Classify predicator as a
verb
 predicator:verb
 Proceed to mood system
 Correct option for a system
chosen by a simple query
or decision network
associated with that system
 Decision based on the
relevant information from
input specification and from
Knowledge Base.


 Mood system chooses the
indicative and declarative
features
 Input specifies assertion.
 Realization statements
associated with the
indicative and declarative
features will insert subject
and finite functions
 order them as subject, then
finite and then predicator.
 +subject
 subject > predicator
 +finite
 finite > predicator
 subject > finite


 The resulting function structure is as follows:


 Assume save-1 is marked
as a material process in
the knowledge base.
 Transitivity function
chooses the material
process feature
 Insert goal and process
functions
 Conflates the process with
the finite/predicator pair
 +goal
 +process
 process= finite,predicator


 Since there is no indication
in either the input or
knowledge base to use a
passive, the system chooses
the active feature, which
 Inserts the actor and
conflates it with the subject
 +actor
 actor=subject
 Inserts the object, conflating
it with the goal and ordering
it after the predicator
 +object
 object=goal
 predicator>object


 This results in the following functional structure.


 There is no thematic
specification in the input
 Thematic network chooses
unmarked theme
 Inserts theme and rheme
 Conflate theme with subject
 Conflate rheme with
finite/predicate/object group
 +theme +rheme
 theme=subject
 rheme=predicator,object


 This results in the full function structure as:


 The generation process
recursively enters the
grammar a number of times
at lower levels to fully
specify the phrases, lexical
items, and morphology.
 This is due to the presence
of the following statements
 When the network found that
it is an indicative statement
 finite : auxiliary
 subject : noun phrase
 When active voice was
identified
 object : noun phrase


 Noun phrase network
 Create the lexical items The system and the document
 Auxiliary network systems
 Create the lexical item will
 The choice of lexical items system, document and
save can be handled in a number of ways, most
typically by retrieving the lexical item associated with
the relevant knowledge base instances.
 The noun phrase and auxiliary network systems
work similar to the clause network we have seen till
now.


Functional Unification Grammar


 Functional Unification Grammar uses unification to
manipulate and reason about feature structures.
 With a few manipulations, the same technique can be
applied to NLG.

 Basic Idea
 Build the generation grammar as a feature structure with
lists of potential alternations
 Then unify this grammar with an input specification built
using the same sort of feature structure.


 Unification process
 takes the features specified in the input
 reconciles them with those in the grammar
 produces a full feature structure which can then be
linearized to form sentence output.


 A simple functional
unification grammar.
 Expressed as an
attribute-value matrix
 Supports simple
transitive sentences in
present or future tense
 Enforces subject-verb
agreement on number


 At highest level, the grammar
provides alternatives for
sentences, noun phrases and verb
phrases
 CAT S
 CAT NP
 CAT VP
 Alternation feature provided by the
ALT feature on the left.
 Curly braces indicate that any of the
enclosed alternatives may be
chosen and followed
 This level also specifies a pattern
indicating the order of the features
specified at this level
 Actor
 Process
 Goal


 At sentence level, grammar
supports the following features.
 Actor  NP
 Process  VP
 Goal  NP
 Subject-verb agreement
 Enforced using the number
feature inside the process
feature.
 Number of processes must unify
with the path {actor number}
 Path  list of features
specifying a path from the root
to a particular feature.
 Here, number of process must
unify with the number of actor.


 While the path is given
explicitly, we can also
have relative paths
 Like the number feature
of the head feature of the
NP.
 The path here,
{↑↑number }, indicates
that the number of the
head of the NP must
unify with the number of
the feature 2 levels up.

Use of {↑↑number}
 VP level is similar to the NP
level except that it has its
own alternation between
future and present tense.
 Tense is specified in the
input feature structure.
 Unification will select the
alternation that matches and
then proceed to unify
associated values.
 If tense is present
 For example, the head will
be single verb.
 If tense is future
 Insert modal auxiliary “will”
before the head verb.


 This grammar is similar to the systemic grammar in
the point that it supports multiple levels, that are
entered recursively during the generation process.

 The details of the particular sentence we want to
generate is given in an input feature structure.


Functional Description (FD)
 The input feature structure.
 It defines the input specifications for the particular
sentence we want to generate.
 It is a feature structure just like the grammar.


 Here, we see a sentence specification with
 a particular action  the system
 a particular goal  the document
 Process  saving of the document by the system in the
future
 The input structure specifies the particular verbs and
nouns to be used as well as the tense
 Different from input to systemic grammar
 In systemic grammar, lexical items retrieved from
knowledge base entries associated with actor and goal.
 Tense, not included in systemic grammar, is computed by
a decision network that determines relative points in time
relevant to the content of the expression.


 Since tense is also to be included in the input feature
structure (Functional Description), more decisions
have to be made by the discourse planning
component.

 To produce the output, the input is unified with the
grammar.
 May require multiple passes through the grammar.


 The preliminary unification unifies the input FD with
the S level in the grammar
 First alternative at the top level
 This results in the structure:


 The features specified in the input structure have
been unified and merged with the features at the top
level of the grammar.
 Features associated with actor include the lexical item
system from the input FD and category NP from the
grammar.
 Process feature combines the lexical item and tense from
the input FD with the category and number features from
the grammar.


 Generation mechanism
now recursively enters the
grammar for each of the
sub-constituents.

 It enters the NP level
twice  for actor and
goal

 It enters the VP level once
 for the process.

Final FD


 Every constituent feature
that is internally complex
has a pattern
specification.
 Every simple constituent
feature has a lexical
specification

 The system now uses the
pattern specifications to
linearize the output,
producing
“The system will save the
document”


 The example didn’t specify the actor to be plural. We
can do that by adding the feature-value pair
number plural
to the actor structure in the input FD.
 Subject-verb agreement would then be enforced by
the unification process.
 Grammar requires that the number of heads of NP
and VP match with the number of the actor that
was specified in the input FD.


Conclusion


 The two surface generation grammars illustrate the
nature of computational grammars for generation.
Both used functional categorizations.
 Bidirectional grammar
 Single grammar for both generation and understanding
 Currently under investigation
 Haven’t found widespread use in NLG
 Additional semantic and contextual information required as input
to the generator


Sample NLG programs

KPML FUF/SURGE

 A text generation  A text generation system
system based off of the and English Grammar
using Functional
earlier Penman system. Unification.
Uses Systemic-  FUF – Functional
Functional Linguistics Unification Formalism is
Principles. an implementation of
 http://www.fb10.uni- Functional Unification
bremen.de/anglistik/langpro/kpml/REA Grammar developed by
DME.html
Elhadad (1992,1993)
 http://www.cs.bgu.ac.il/research/projects/s
urge/index.htm


THANK YOU…


Surface realization

Recomendados

Recomendados

Mais conteúdo relacionado

Mais procurados

Mais procurados (20)

Destaque

Destaque (19)

Semelhante a Surface realization

Semelhante a Surface realization (20)

Último

Último (20)

Surface realization