conceptual coherence in the generation of referring expressions - PowerPoint PPT Presentation


PPT – conceptual coherence in the generation of referring expressions PowerPoint presentation | free to download - id: 76e94e-YmRjN


The Adobe Flash plugin is needed to view this content

Get the plugin now

View by Category
About This Presentation

conceptual coherence in the generation of referring expressions


conceptual coherence in the generation of referring expressions Albert Gatt & Kees van Deemter University of Aberdeen {agatt, kvdeemte} – PowerPoint PPT presentation

Number of Views:28
Avg rating:3.0/5.0
Slides: 30
Provided by: agatt
Learn more at:


Write a Comment
User Comments (0)
Transcript and Presenter's Notes

Title: conceptual coherence in the generation of referring expressions

conceptual coherence in the generation of
referring expressions
  • Albert Gatt Kees van Deemter
  • University of Aberdeen
  • agatt,

  • Gatt and Van Deemter 2007 Lexical Choice and
    conceptual perspective in the generation of
    plural referring expressions. Journal of Logic
    Language and Information (JoLLI) 16 (4),

some received wisdom
  • Choice is ultimately dependent on the
    perspective you decide to take on the referent
  • Will it be more effective for me to refer to my
    sister as my sister or as that lady or as the
    physicist ? (Levelt 99, p. 226)

the rest of this talk
  • Generation of Referring Expressions
  • Perspective and Conceptual Coherence
  • reference to sets (van Deemter 2002)
  • experimental work
  • An algorithm
  • evaluation
  • Extensions
  • local (Conceptual) Coherence in discourse

Generation of Referring Expressions (GRE)
  • Part of micro-planning (Reiter/Dale 00)
  • At this stage, the content of a message is being
    determined, including descriptions of domain
    objects (Noun Phrases)
  • The task of GRE
  • given a set of intended referents, look up
    properties of these referents that will
    distinguish them from their distractors in a
    Knowledge Base

Content determination strategies
entity base type occupation specialisation girth
e1 woman professor physicist Plump
e2 woman lecturer geologist thin
e3 man lecturer biologist plump
e4 man postgraduate thin
  • Most algorithms inspired by the Gricean maxims
    (Grice 75)
  • especially Brevity (Dale 89, Gardent 02)
  • But compare
  • ?? ?x professor(x) V plump(x)
  • ?? ?x professor(x) V plump(x) man(x)
  • ? ?x biologist(x) V physicist(x)
  • All are equally brief, but not all have an
    equally good ring to them.

the Conceptual Coherence constraint
  • Sets (and disjunction) ?x p(x) V q(x) the p
    and the q
  • reference to a plurality suggests to the listener
    that there is a relationship holding between
    elements of the pluralities
  • p and q should be related or similar
  • semantic relatedness allows the listener to
    conceptualise the plurality more easily (Sanford
    and Moxey, 95)
  • Gatt and van Deemter (02)
  • Peoples preference for descriptions of this form
    were highly correlated to the semantic similarity
    of disjuncts
  • Best results achieved with a distributional
    definition of similarity (Lin 98)
  • sim(w,w) is a function of how often w and w
    occur in the same grammatical relations in a

Lins definition of distributional similarity
  • Let w1, w2 be two words of the same grammatical
  • E.g. dog, cat
  • Let F(w) GR1,,GRn be the set of grammatical
    relations w occurs in, where
  • GRi ltwi, R, x, I(w,x)gt
  • wi the target word, R the relation, x the
    co-argument of w
  • I(wi,x) is the probability of wi and x occurring
    in this construction (as mutual information).
  • Example ltdog, modified-by, straygt
  • sim(w1, w2) is calculated using the GR triples
    that w1 and w2 share.
  • We use SketchEngine, a large-scale implementation
    of this theory, based on the BNC (Kilgarriff,

experiment 1 multimodal sentence completion
  • General idea
  • To refer to a set, people will prefer to use a
    plural that respects the conceptual coherence
  • If this is impossible, then they will prefer to
    refer to the individuals in the set separately.
  • Experimental domains
  • 3 targets (a,b,c) 1 distractor (d)
  • sim(a,b) could be high or low
  • sim(a,c) sim(b,c) low
  • Expectation
  • if 2 of the targets have semantically similar
    types, they will be referred to in a plural
  • reference to any two targets in case similarity
    is constant is no better than chance

experiment 1 example domain
  • Experimental domain
  • Participants completed the sentences by clicking
    on the pictures.
  • Manipulation of similarity of two of the objects
  • Hypothesis
  • If a,b are similar, they are more likely to be
    referred to in the plural.

Complete the following by clicking on the
The _____________ and the _____________ cost 5.
The _____________ also costs 5.
experiment 1 results
Proportion of plural references to designated
targets a,b when
experiment 2 sentence continuation
  • Does similarity play a role in content
  • Distinguishing properties nouns (12) or
    adjectives (12 ).
  • Expectation
  • Participants will select similar properties in
    the plural description, even though the
    alternative would also constitute a successful

A university building was robbed last night. The
police have detained three suspects for
questioning, all of whom work or study at the
university. 1. One of them is a postgraduate. He
is a physicist. 2. Another is, a Greek, an
undergraduate. 3. Also among the suspects is a
cleaner. He is an Italian. Both
______________________ were held in custody, but
the physicist was released last night.
experiment 2 results
Proportion of references using pairwise similar
Friedman 45.89, p lt .001 trend as expected
Friedman 36.3, p lt .001 trend in the opposite
summary of findings so far
  • In referential situations, people prefer to
    produce plural descriptions if the entities can
    be conceptualised under the same perspective.
  • This holds for types, but not modifiers
  • Types correspond to concepts, and are the way
    we carve up the world and categorise objects
  • Modifiers correspond to properties of those
  • Results have been corroborated in other

a GRE algorithm
  • Main knowledge source
  • The relation sim (which is based on the BNC-based
    lexical database of Kilgarriff 03)
  • Input
  • Knowledge base
  • Target referents (R )

step 1
  • Lexicalise properties in the KB
  • Identify types (nominal properties) and modifiers
  • The set of types and the similarity relation
    define a semantic space S ltT, simgt
  • Definition 1 Perspective
  • A perspective P is a convex subset of S, i.e.
  • ? t, t, t ? T
  • t, t ? P sim(t, t) sim(t, t) ? t ? P
  • Computed using a clustering algorithm (Gatt 06),
    which recursively groups together semantic
    nearest neighbours.

perspective graph
  • Aim find a description for R that minimises the
    distance between perspectives from which
    properties are selected.
  • Weight of a description, w(D) the sum of
    distances between perspectives represented in D.
  • w( the professor and the plump man ) 1
  • w( the biologist and the physicist ) 0

descriptive coherence
  • Definition 2 Maximal coherence
  • D is maximally coherent if there is no D
    coextensive with D such that w(D) lt w(D)
  • implies finding a shortest connection network in
    the perspective graph (intractable!)
  • Definition 3 Local coherence
  • D is locally coherent if there is no D
    coextensive with D s.t.
  • D is obtained by replacing a perspective in D
  • w(D) lt w(D)

search procedure
  • N ? Ø //the perspectives represented in D
  • root ? perspective with most referents in its
  • starting from root do
  • Check types and modifiers.
  • If a property excludes distractors
  • add it to D
  • add the perspective to N
  • If R is not distinguished, go to the next
    perspective, which is

(V is the set of perspectives).
  • Do people prefer coherence over brevity?
  • (Two Gricean maxims Be brief vs. Be orderly)
  • Method subjects (N 39) shown 6 discourses.
  • Each discourse introduces 3 entities
  • Followed by 2 possible continuations
  • Subjects had to indicate their preferred
  • Each of the 6 discourses represented a condition
  • Brevity descriptions equally (in-)coherent, but
    one is brief
  • Coherence descriptions equally (non-)brief only
    one is coherent
  • Trade-off coherent description is non-brief

results preference for brevity
both descriptions coherent x2 .023, p .8
both descriptions non-coherent x2 .64, p .4
results preference for coherence
both descriptions non-minimal x2 13.56, p lt .001
both descriptions minimal x2 16.03, p lt .001
results trade-off
coherent non-minimal x2 39.0, p lt .001
  • Perspectives on perspectives
  • Aloni (2002) answers to questions wh x? must
    conceptualise the different x using one and the
    same perspective (relevant given hearers
    information state and the context)
  • GvD (2007)
  • perspective defined (via lexical similarity)
  • when its impossible to use the same perspective,
    use perspectives that are similar

  • Many experiments were done
  • to find a suitable notion of similarity/coherence
  • to discover how coherence and brevity relate
  • Surprise coherence more important than brevity
    (in the cases investigated!) no effect of
    brevity was found
  • Yet, different algorithmic interpretations would
    be possible
  • Algorithms are almost always under-determined by
    the empirical evidence

This is a lexical approach
  • Advantage empirical resources for determining
    similarity/coherence are readily available
  • Disadvantage the relation with classic (i.e.,
    linguistic or real-AI) notions of perspective
    and relevance is unclear
  • (A modest move towards reconciliation the root
    could be chosen using some notion of relevance)

A limitation
  • Ambiguity/polysemy is not taken into account
  • For example, we might generate
  • the river and the/its bank
  • These issues investigated in Imtiaz Khans PhD
  • One remark river might disambiguate bank

An open question
  • Why doesnt coherence play the same role for
    modifiers as for types?

Work in progress discourse coherence
  • The algorithm has been extended with a discourse
    model, which keeps track of previous references.
  • Currently experimenting with a model of discourse
  • when generating a new reference, prefer the
    perspectives that have been used in the previous
    n utterances
  • Problems
  • verifying this empirically
  • defining n (cf. Centering Theory)