Anaphora Resolution
Sobha Lalitha Devi
AU-KBC Research Centre
MIT Campus of Anna University
Chennai-44
sobha@au-kbc.org
Contents






Introduction to Anaphora and
Anaphora Resolution
Types of Anaphora
Process of Anaphora Resolution
Tools
Applications
References
05/24/2011
Summer School, IIIT HYderabad
2
What is Cohesion
COHESION is the internal continuity or network of
points of continuity within a text.
Text is not just a string of sentences. It is not simply a large
grammatical unit
“something of the same kind as a sentence, but differing
from it in size- a sort of super-sentence· A semantic unit”
Halliday & Hassan
05/24/2011
Summer School, IIIT HYderabad
3
Cohesive Relationships
Cohesive relationships between words and
sentences have certain definable qualities that
allow us to recognize the super sentence
Nature of cohesive relation
Type of cohesion
Relatedness of form
Substitution and ellipsis; Lexical
collocation
Relatedness of reference Reference, Lexical reiteration
Semantic connection
Conjunction
05/24/2011
Summer School, IIIT HYderabad
4
Relatedness of Form



Substitution:
"Nice teapots! I'll take one.“
Ellipsis:
"Turn on. Tune in. Drop out." ['you' is
elided]
Collocation:
"John went to the bank. He wanted to swim
in the river." ['river' disambiguates 'bank']
05/24/2011
Summer School, IIIT HYderabad
5
Relatedness of Reference
Exophora: (extra linguistic feature: deitic markers , this
that)
“what is this?”
Anaphora:
"I used to have the key. But I lost it.”
Cataphora:
"It is your turn, John”
Reiteration:
"He speaks only to the Huxleys; the Huxleys speak only
to the Darwins; and the Darwins speak only to God.“
05/24/2011
Summer School, IIIT HYderabad
6
Semantic Connection
Conjunction:
"You tell me that you've got ev'rything you want, and your
bird can sing, but you don't get me, you don't get me!
You say you've seen seven wonders, and your bird is
green, but you can't see me, you can't see me!
When your prized possessions, start to tear you down, then
look in my direction, I'll be round, I'll be round."
[Beatles -- Lee Campbell]
05/24/2011
Summer School, IIIT HYderabad
7
Comparison
Halliday & Hassan also classify comparison
as a form of cohesion
"She's more fun than a barrel of monkeys!"
"He's as tall as a six foot four inch tree."
05/24/2011
Summer School, IIIT HYderabad
8
CONTEXT DEPENDENCE
The interpretation of most expressions
depends on the context in which they are used

Developing methods for interpreting context
dependent expressions useful in many
applications

We focus here on dependence of nominal
expressions on context introduced
LINGUISTICALLY,
for which we will use the term ANAPHORA

05/24/2011
Summer School, IIIT HYderabad
9
Introduction
What is
Anaphora
Antecedent
Anaphora Resolution
1. Sabeer Bhatia arrived at Los Angeles International
Airport at 6 p.m. on September 23, 1998. His flight
from Bangalore had taken 22hrs and he was
starving.
[RD, NOV 2000]
05/24/2011
Summer School, IIIT HYderabad
10
Etymology of Anaphora
ANA- Back, Upstream, Back upstream
Phora-
Act of Carrying
Anaphora - Act of Carrying Back
05/24/2011
Summer School, IIIT HYderabad
11
What is Anaphora
Anaphora, in discourse, is a device for
making an abbreviated reference (containing
fewer bits of disambiguating information,
rather than being lexically or phonetically
shorter) to some entity (or entities) in the
expectation that the receiver of the
discourse will be able to disabbreviate the
reference and, thereby, determine the
identity of the entity.
(Hirst 1981)
05/24/2011
Summer School, IIIT HYderabad
12
Cataphora

When “anphora” precedes the antecedent
Because she was going to the
departmental store, Mary was asked to
pick up the vegetables.
05/24/2011
Summer School, IIIT HYderabad
13
Relevance from the Linguistics
point of view



Binding Theory is one of the major results of
the principles and parameters approach
developed in Chomsky (1981) and is one of
the mainstays of generative linguistics.
The Binding Theory deals with the relations
between nominal expressions and possible
antecedents.
It attempts to provide a structural account of
the complementarity of distribution between
pronouns, reflexives and R-expressions.
05/24/2011
Summer School, IIIT HYderabad
14
Dichotomy Between Linguistic and
NLP




The Binding Theory (and its various formulations)
deals only with intra-sentential anaphora,
A very small subset of the anaphoric phenomenon
that practical NLP systems are interested in
resolving.
A much larger set of anaphoric phenomenon is the
resolution of pronouns inter-sententially.
This problem is dealt with by Discourse
Representation Theory and more specifically by
Centering Theory (Grosz et al., 1995)..
05/24/2011
Summer School, IIIT HYderabad
15
Types of Anaphors
The Prime Minister is yet to arrive and he is
expected at the central hall at any time. [The
Times of India, Feb 2001]
This book is about Anaphora Resolution. The
book is designed to help beginners in the
field and its author hopes that it will be
useful.
VP Anaphor
John screamed, as did Mary .
05/24/2011
Summer School, IIIT HYderabad
16
Pronominal anaphora
Vajpayee hits back forcefully when he told the
opposition today “sometimes we fall prey to
the media and sometimes you do. [Indian
Express 2001]
Possessive
Priyanka eats only chicken sandwiches
before going to take any exam; nothing else
goes down her gullet that day.[Indian Express, 13
March 2001]
05/24/2011
Summer School, IIIT HYderabad
17
Reflexive Pronoun
Finally ,Danian heaved himself up and lay on a waiting
stretcher.
Demonstrative Pronoun
John had lots of packing to do before he shifted his
house. This was something he never liked….
Relative Pronoun
Stumper Sameer Dige, who made his test debut, failed
to show fast reflexives when it mattered.
05/24/2011
Summer School, IIIT HYderabad
18
Non Anaphoric Usage of Pronouns.
Pleonastic It
Cognative
a. It is believed that…..
b. It appears that…..
Modal Adjectives
c. It is dangerous……
d. It is important…..
Temporal
e. It is five o’clock
f. It is winter
Weather verbs
g. It is raining
f. It is snowing
Distance
h. How far it is to Chennai?
05/24/2011
Summer School, IIIT HYderabad
19
Non-anaphoric uses of pronouns
He that plants thorns must never expect to
gather roses.
He who dares wins.
Deictic
He seems remarkably bright for a child of his
age.
05/24/2011
Summer School, IIIT HYderabad
20
Noun Phrase Anaphora
Definite descriptions and Proper names
Roy Kaene has warned Manchester United he may
snub their pay deal. United’s skipper is even
hinting that unless the future Old Trafford Package
meets his demands, he could quit the club in June
2000. Irishman Keane, 27, still has 17 months to
run on his current 23,000 pound a week contract
and wants to commit himself to United for life. Alex
Ferguson’s No 1 player confirmed: If it’s not the
contract I want, I won’t sign”.
05/24/2011
Summer School, IIIT HYderabad
21
Coreference
Computational Linguists from many
different countries attended the tutorial.
The participants found it hard to cope
with the speed of the presentation,
nevertheless they manages to take
extensive notes.
05/24/2011
Summer School, IIIT HYderabad
22
Coreference
Sophia Loren says she will always be grateful
to Bono. The actress revealed that the U2
singer helped her calm down when she
became scared by a thunderstorm while
travelling by a plane.
She=> Sophia Loren
The actress=> Sophia Loren
The U2 Singer=> Bono
Her=>Sophia Loren
She=>Sophia Loren
05/24/2011
Summer School, IIIT HYderabad
23
Coreference chain
Sophia Loren says she will always be grateful to
Bono. The actress revealed that the U2 singer
helped her calm down when she became scared by
a thunderstorm while travelling by a plane.
Coreference chains
 {Sophia Loren, she, the actress, her, she}
 {Bono, the U2 singer}
05/24/2011
Summer School, IIIT HYderabad
24
Chains of object mentions in
text
Toni Johnson pulls a tape measure across the front of
what was once a stately Victorian home. A deep trench
now runs along its north wall, exposed when the house
lurched two feet off its foundation during last week's
earthquake.
Once inside, she spends nearly four hours measuring and
diagramming each room in the 80-year-old house,
gathering enough information to estimate what it would
cost to rebuild it.
While she works inside, a tenant returns with several
friends to collect furniture and clothing.
One of the friends sweeps broken dishes and shattered
glass from a countertop and starts to pack what can be
salvaged from the kitchen.
(WSJ section
of Penn Treebank corpus)
05/24/2011
Summer School, IIIT HYderabad
25
What is Anaphora Resolution

The Process of finding the antecedent for an
Anaphor is Anaphora resolution


05/24/2011
Anaphor-The reference that point to the previous
item.
Antecedent-The entity to which the anaphor
refers
Summer School, IIIT HYderabad
26
RESEARCH ON ANAPHORA
RESOLUTION: A QUICK SUMMARY
1970-1995
Primarily theoretical
Emphasis: commonsense knowledge, salience
Exception: Hobbs 1977, Shalom Lappin
1995-2005
First annotated corpora to be used to develop, evaluate and
compare systems (MUC, Geand Charniak, ACE)
First robust systems
Heuristic-based: Mitkov
ML: Vieira & Poesio 1998, 2000; Soon et al 2001, Ng and
Cardie2002
Emphasis: surface features
Exceptions: Poesio & Vieira, Harabagiu, Markert
2005-present
More sophisticated ML techniques (global models, kernels)
Richer
features –especiallySummer
semantic
information
05/24/2011
School, IIIT HYderabad
First tools
27
Application of Anaphora
Resolution
Tasks that require determining the coherence of (segments of) text
Segmentation
Post-hoc coherence check in summarization (Steinberger et al, 2007)
Tasks that require identifying the most important information ina text
Sentence selection in summarization (Steinberger et al 2005, 2007)
Indexing
Information extraction: recognize which expressions refer to objects in
the domain
Relation extraction from biomedical text (Sanchez-Grailletand Poesio,
2006, 2007)
Multimodal interfaces: recognize which objects in the visual scene are
being referred to
05/24/2011
Summer School, IIIT HYderabad
28
Different Approaches In Anaphora
Resolution



Rule Based
Statistical Based
Machine Learning Based
05/24/2011
Summer School, IIIT HYderabad
29
Rule Based

Hobbs system
05/24/2011
Summer School, IIIT HYderabad
30
Hard Constraints on
Coreference





Number agreement
Person and case
Gender Agreement
Syntactic Agreement
Selectional Restrictions
05/24/2011
Summer School, IIIT HYderabad
31
Number agreement
Singular
Plural
Unspecifie
d
She,her, he,
him, his, it
We, us,
they,
them
you
John and Mary loaned Sue a cup of
coffee. Little did they know the
magnitude of her addiction.
05/24/2011
Summer School, IIIT HYderabad
32
Person and Case Agreement
First
Secon Third
d
Nominative
I,we
you
he,she,they
Accusative
me,us you
Him,her,the
m
Genitive
my,ou your
r
His,her, their
05/24/2011
Summer School, IIIT HYderabad
33
Gender Agreement
*John has a coffee machine. She loves
it.
05/24/2011
Summer School, IIIT HYderabad
34
Syntactic Agreement

Reflexives (himself, herself…) have
strong constraints on what syntactic
positions they can appear in
John bought himself a cup of coffee.
*John bought him a cup of coffee.
05/24/2011
Summer School, IIIT HYderabad
35
Selectional Constraints
Jim bought a coffee from the store. He
drank it quickly.
05/24/2011
Summer School, IIIT HYderabad
36
Also : Preferences





Recency
Grammatical Role
Repeated Mention
Parallelism
Verb Semantics

05/24/2011
Based on Salience
Summer School, IIIT HYderabad
37
Recency
John had a pop-tart. Bill had a jelly
donut. Mary wanted it.
Recent Entities are more salient
05/24/2011
Summer School, IIIT HYderabad
38
Grammatical Role
“Sue bought a cup of coffee and a
donut from Jane. She met John as
she left.”

Entities in subject position are more
salient
05/24/2011
Summer School, IIIT HYderabad
39
Repeated Mention
John went to the store to buy coffee.
He loves coffee. He drinks 5 cups a
day. At the store, Bill sold him a cup.
He was delighed.

Entities mentioned more frequently are
more salient
05/24/2011
Summer School, IIIT HYderabad
40
Parallelism
John bought coffee from Jim in the
morning. Sue bought coffee from him
in the evening.

Even with preferences to the contrary
(grammatical role) the syntactic
parallelism strongly prefers [him = Jim]
05/24/2011
Summer School, IIIT HYderabad
41
Verb Semantics
John telephoned Bill. He was jonesing
for coffee.
John criticized Bill. He was jonesing
for coffee.

Perhaps salience of different elements
in the sentence changes with respect
to the verb used.
05/24/2011
Summer School, IIIT HYderabad
42
Algorithms --- How to integrate
these preferences?

Constraints are easy to use : reject all
hypothesis which violate the hard
constraints
(if you can accurately detect the constraints!)

Preferences more difficult – how can
one integrate these different
preferences?
05/24/2011
Summer School, IIIT HYderabad
43
Hobbs Tree Search Algorithm

Given parse trees, search them in a
specific order to find the most likely
referent
05/24/2011
Summer School, IIIT HYderabad
44
Hobbs in Detail
1.
Begin at NP
2.
Go up tree to first NP or S. Call this X, and the path p.
3.
Traverse all branches below X to the left of p. Propose as
antecedent any NP that has a NP or S between it and X
4.
If X is the highest S in the sentence, traverse the parse
trees of the previous sentences in the order of recency.
Traverse left-to-right, breadth first. When a NP is
encountered, propose as antecedent. If not the highest
node, go to step 5.
05/24/2011
Summer School, IIIT HYderabad
45
Hobbs cont.
5.
6.
7.
8.
From node X, go up the tree to the first NP or S.
Call it X, and the path p.
If X is an NP and the path to X did not pass through
the nominal that X dominates, propose X as
antecedent
Traverse all branches below X to the right of the
path, in a left-to-right, breadth first manner.
Propose any NP encountered as the antecdent
If X is an S node, traverse all brnaches of X to the
right of the path but do not go below any NP or S
encountered. Propose any NP as the antecedent.
05/24/2011
Summer School, IIIT HYderabad
46
Lappin and Leass (1994) Anaphora
Resolution Algorithm
 The Lappin and Leass(1994) anaphora resolution algorithm uses
salience weight in determining the antecedent to the pronominals.
 It requires as input a fully parsed sentence structure and
uses hierarchy in identifying the subject, object etc.
 This algorithm uses syntactic criteria to rule out noun
phrases that cannot possibly corefer with it.
 The antecedent is then chosen according to a ranking based
on salience weights.
05/24/2011
Summer School, IIIT HYderabad
47
The salience Factors and Weights
A pronoun P is non-coreferential with a (non-reflexive or nonreciprocal) noun phrase N if any of the following conditions
hold:





P and N have incompatible agreement features.
P is in the argument domain of N.
P is in the adjunct domain of N. P is an argument of a head
H, N is not a pronoun, and N is contained in H.
P is in the NP domain of N.
P is a determiner of a noun Q, and N is contained in Q.
05/24/2011
Summer School, IIIT HYderabad
48
Examples
Condition 1:
The woman said that he is funny.
Condition 2:
She likes her. John seems to want to see him.
Condition 3:
She sat near her.
Condition 4:
He believes that the man is amusing.
This is the man he said John wrote about.
Condition 5:
John’s portrait Summer
of him
is interesting.
05/24/2011
School, IIIT HYderabad
49
Salience Factors and Weights
Salience factor types with initial weights
Factor type
Initial weight
Sentence recency
100
Subject emphasis
80
Existential emphasis
70
Accusative emphasis
50
Indirect object and oblique
complement emphasis
40
Head noun emphasis
80
Non-adverbial emphasis
50
05/24/2011
Summer School, IIIT HYderabad
50
Kennedy 1996
The linguistic analysis for anaphora resolution includes
The output of a part of speech tagger,
Augmented with syntactic function annotations for each
input token;
Using LINGSOFT
05/24/2011
Summer School, IIIT HYderabad
51
A set of patterns are used for identifying



The NP Chunking with position of the NP in the text:
Nominal Sequencing in two subordinate syntactic
environments:
a. in an adverbial adjunct
b. in an NP (i.e. containment in a prepositional
or clausal complement of a noun, or
containment in a relative clause)
Expletive “it”:
05/24/2011
Summer School, IIIT HYderabad
52
Anaphora Resolution

Uses Lappin and Lease algorithm
SENT-S:
CNTX-S:
SUBJ-S:
EXST-S:
POSS-S:
ACC-S:
DAT-S:
OBLQ-S:
HEAD-S:
ARG-S:
05/24/2011
100 iff in the current sentence
50 iff in the current context
80 iff GFUN = subject
70 iff in an existential construction
65 iff GFUN = possessive
50 iff GFUN = direct object
40 iff GFUN = indirect object
30 iff the complement of a preposition
80 iff EMBED = NIL
50 iff ADJUNCT
= NIL
Summer School, IIIT HYderabad
53
Mitkov 1997

No Parsing of the Input Sentence

Boosting indicators

First Noun Phrases: A score of +1 is assigned to the first
NP in a sentence.

Indicating Verbs: A score of +1 is assigned to those NPs
immediately following a verb which is a member of a
predefined set (including verbs such as discuss, present,
illustrate, identify, summarise, examine, describe, define,
show, check, develop, review,
05/24/2011
Summer School, IIIT HYderabad
54
MARS Cont….


05/24/2011
Lexical Reiteration: A score of +2 is
assigned to those NPs repeated twice or
more in the paragraph in which the
pronoun appears, a score of +1 is
assigned to those NPs repeated once in
that paragraph.
Section Heading Preference: A score of
+1 is assigned to those NPs that also
occur in the heading of the section in
which the pronoun appears.
Summer School, IIIT HYderabad
55
Boosting indicators contd..
Collocation Match: A score of +2 is assigned to those
NPs that have an identical collocation pattern to the
pronoun.

Immediate Reference: A score of +2 is assigned to those
NPs appearing in constructions of the form
“… (You) V1 NP … con (you) V2 it (con (you) V3 it)”, where
con Є {and/or/before/after…}.

Sequential Instructions: A score of +2 is applied to NPs in
the NP1 position of constructions of the form: “To V1 NP1
V2 NP2. (Sentence). To V3 it, V4 NP4“ the noun phrase
NP1 is the likely antecedent of the anaphor it (NP1 is
assigned a score of 2).

Term Preference: A score of +1 is applied to those NPs
identified as representing terms in the genre of the text.

05/24/2011
Summer School, IIIT HYderabad
56
Impeding indicators

Indefiniteness: Indefinite NPs are
assigned a score of -1.

Prepositional Noun Phrases: NPs
appearing in prepositional phrases are
assigned a score of -1.
05/24/2011
Summer School, IIIT HYderabad
57
Indian Language

Types of anaphors

Dravidian type with gender marked
pronouns


Aryan Language Type without gender
marking in the pronouns


05/24/2011
avan, aval, atu (example) Telugu does not
have all other Dravidian languages have.
“us” as in Hindi (example)
FROM CHAPTER
Summer School, IIIT HYderabad
58
<ANT 1,2,3> uutti </ANT> oru alakiya
malai nakaram.
Ooty(N)
one(Q) beautiful(ADJ) hill(N) town(N).
(Ooty is a beautiful hill station.)
<ANA 1> ithu </ANA> malaikalin raani.
This
hill(N)+GEN queen(N).
(This is Queen of hills.)
<ANA 2> anku </ANA> puungkaa, pataku cavaari, malai rayil untu.
There
park(N), boating(VBN) hill train(N)
present(V+PRESENT+3S)
(There are park, boating, hill train )
<ANA 3> athu </ANA> oru cirantha currulaath thalam.
it(PN)
one(Q) best(ADJ) tourist(N) place(N).
(It is a best tourist place.)
05/24/2011
Summer School, IIIT HYderabad
59
“Vasisth” a Rule Based Anaphora
Resolution System
1. mo:han(i)
avanRe(i) kuttiye
mohan
he-poss
child-acc
(Mohan saw his child.)
2. mo:han(i) avanRe(i) kuttiye
kantu
mohan
he-poss
kantu.
see-pst
ennu kRisnan paRannu.
child-acc see-pst compl krishnan say-pst
(Krishnan said that Mohan saw his child.)
3. *mo:han(i) avane(i) aticcu.
mohan
he-acc beat-pst
(Mohan beat him.)
4. mo:han avane(i) aticcu
ennu
kRisnan(i) paRannu.
mohan
he-acc beat-pst compl krishnan say-pst
(Krishnan said that Mohan beat him.)
05/24/2011
Summer School, IIIT HYderabad
60
The Algorithm for Intra-sentential
Anaphora
A pronoun P is coreferential with an NP iff the following
conditions hold:
a. P and NP have compatible P, N, G features.
b. P does not precede NP.
c. If P is possessive, then NP is the subject of
the clause which contains P.
d. If P is non-possessive, then NP is the subject
of the immediate clause which does not
contain P.
05/24/2011
Summer School, IIIT HYderabad
61




Vasisth is a multilingual Anaphora
Resolution system
Rule based
With minimum Parsing
Exploit the Morphology of Indian
Languages
05/24/2011
Summer School, IIIT HYderabad
62
“VASISTH” Using Salience Measure
for Indian Languages
No In-depth Parsing
Exploit the Rich Morphology of the
Language
The analysis depends on the salience
weight of the candidate (NP) for the
antecedent-hood of an anaphor from a
list of probable candidates.
05/24/2011
Summer School, IIIT HYderabad
63
The salience weight assignment
a) The current sentence gets a score of 50 and it reduces by 10
for each preceding sentence till it reaches the fifth sentence.
The system considers five sentences for identifying the
antecedent.
b) The current clause gets a score of 75 if the pronoun present
in the clause is a possessive pronoun and if it is a nonpossessive pronoun it gets zero score.
c) The immediate clause gets the score 70 in the case of
Possessive pronoun and gets a score of 75 for nonpossessive pronouns.
d) For non-immediate clause, the possessive pronoun gets a
score of 30 and non-possessive pronoun gets a score of 65.
05/24/2011
Summer School, IIIT HYderabad
64
e)The analysis showed that the subject could be the
most probable antecedent for the pronoun. The
case markings the subject of a sentence could
take are nominative and dative.
A Nominative, a Dative and a Possessive NP
with a nominative/Dative head could become a
subject of a sentence.
05/24/2011
Summer School, IIIT HYderabad
65
f) The direct object of a sentence could be identified by the
case markings and all the case markings other than the
subject are considered for object. The next most probable
NP for antecedent-hood is the direct object and hence it
gets a score of 40.
g) The third NP in a clause, which is not identified as the
subject or object, is considered as the indirect object and
gets a low score of 30.
05/24/2011
Summer School, IIIT HYderabad
66
Salience factor weights for Indian Languages
Salience Factors
Current sentence
Weights
50- Reduced by 10 for preceding
sentences upto 5th sentence
Possessive
Current clause
Immediate clause
Non-immediate clause
75
70
30
Non-Possessive
Current clause
Immediate clause
Non-immediate clause
Possessive and Non-Possessive
N.Nom
N.Poss
N.Dat
05/24/2011
N.Acc, Loc,
Instr…
N.others(3rd NP)
0
75
65
80
50
50
Summer School, IIIT
HYderabad
40
30
67
How it works
The salience weight to an NP is assigned in the following
way
Identify the Pronoun
 Consider Four sentences above the sentence containing
the Pronoun
 Consider all the NPs preceding the Pronoun ( This is
the general rule)
05/24/2011
Summer School, IIIT HYderabad
68
Here we take some NPs which follow the the
Pronoun since Tamil
All Indian languages are relatively free word
Order
Assign Salience Weights.
The NP which gets the maximum salience weight
and agrees in png with the anaphor is
considered as the antecedent to the anaphor
05/24/2011
Summer School, IIIT HYderabad
69
Machine Learning

CONLL Task and its results
05/24/2011
Summer School, IIIT HYderabad
70
Introduction

Goal of the task


Automatically identify coreference chains in a document
The coreference chains can include





Names
Nominal mentions
Pronouns
verbs that are coreferenced with a noun phrases.
The data used is the Ontonotes


English documents from Ontonotes
This consists of five different types of genres

05/24/2011
Newswire (NW), Broadcast News (BN), Telephonic
Conversation (TC), Broadcast conversation (BC), Web blogs
(WB)
Summer School, IIIT HYderabad
71
Introduction

Coreferencing



Coreferents are classified into two types




Two referential entities when they exist in the
real world
Coreference analysis determines whether or not
two entities refer to the same entity
Pronominal referents
Non-pronominal referents
Pronominal referents are pronouns, which
refer to other nouns in the text
Non-pronominal referents are names,
nominal mentions and other noun phrases
05/24/2011
Summer School, IIIT HYderabad
72
A Sample Text
Eagle Clothes Inc. , which is operating under Chapter 11 of the
Federal Bankruptcy Code , said it reached an agreement with its
creditors .
Under the accord , Albert Roth , chairman and chief executive
officer , and Arthur Chase , Sam Beigel , and Louis Polsky will
resign as officers and directors of the menswear retailer .
Mr. Roth , who has been on leave from his posts , will be
succeeded by Geoffrie D. Lurie of GDL Management Inc. , which is
Eagle 's crisis manager .
Mr. Lurie is currently co-chief executive.
05/24/2011
Summer School, IIIT HYderabad
73
Coreference Tagged Text
<Coref_chain Id="1">Eagle Clothes Inc.</Coref_chainId="1"> , which
is operating under Chapter 11 of the federal Bankruptcy Code , said
<Coref_chain Id="1">it</Coref_chain Id="1"> reached an agreement
with <Coref_chain Id="1">its</Coref_chain Id="1"> creditors .
Under the accord , <Coref_chain Id="2">Albert Roth , chairman and
chief executive officer</Coref_chain Id=“2"> , and Arthur Chase , Sam
Beigel , and Louis Polsky will resign as officers and directors of the
<Coref_chain Id="1">menswear retailer</Coref_chain Id="1"> .
<Coref_chain Id="2">Mr. Roth , who has been on leave from his posts
,</Coref_chain Id="2"> will be succeeded by <Coref_chain
Id="3">Geoffrie D. Lurie of GDL Management Inc. , which is
<Coref_chain Id="1">Eagle 's</Coref_chain Id="1"> crisis manager
</Coref_chain Id="3">.
<Coref_chain Id="3">Mr. Lurie </Coref_chain Id="3"> is currently cochief executive
.
05/24/2011
Summer School, IIIT HYderabad
74
Our Approach


Two different approaches are used for the two
types of coreferents
Mainly two modules




Pronominal resolution module
Non-pronominal resolution module
Pronominal resolution refers to identification of
a Noun phrase (NP) that is referred by a
pronominal
Non-Pronominal resolution refers to
identification of a NP referring to other NP.
05/24/2011
Summer School, IIIT HYderabad
75
Pronominal Resolution Module


Pronominal Resolution is done using refined salience
measure.
All pronouns do not refer back to an entity
 For example in the sentence “It will rain today”
 The pronoun “It” does not refer to any entity
 Such an instance of “It” is called as pleonastic “it”
 Before we do pronominal resolution, we need to
identify such non-anaphoric pronouns
05/24/2011
Summer School, IIIT HYderabad
76
Pronominal Resolution Module (Contd…)

The first step in the pronominal
resolution module is



identification of non-anaphoric pronouns
and filter the non-anaphoric pronouns
from text
The identification of non-anaphoric
pronouns is done using CRFs, a
machine learning approach
05/24/2011
Summer School, IIIT HYderabad
77
Pronominal Resolution Module (Contd…)

Using CRFs we build a language
model

The features used for the training are




Word
Part-of-speech (POS) tag
Window of five words
Non-anaphoric pronoun filtering has
been observed to improve the results
05/24/2011
Summer School, IIIT HYderabad
78
Pronominal Resolution Module (Contd…)

The second step - core task of the
pronoun resolution module



the identification of antecedent for a pronoun
Use a refined salience measure based
approach
WordNet and NE tag information are used
to match the category of pronoun and the
antecedent
05/24/2011
Summer School, IIIT HYderabad
79
Pronominal Resolution Module (Contd…)

The table below shows the refined salience
factors and the weights assigned for them
Salience Factors
Weights
Current Sentence (sentence in which pronoun occurs)
100
For the preceding sentences up to four sentences from the current sentence
Reduce sentence score by 10
Current Clause (clause in which pronoun occurs)
100 – for possessive pronoun
50 – for non-possessive pronouns
Immediate Clause (clause preceding or following the current clause)
50 – for possessive pronoun
100 – for non-possessive pronouns
Non-immediate Clause (neither the current or immediate clause)
50
Possessive NP
65
Existential NP
70
Subject
80
Direct Object
50
Indirect Object
40
Compliment of PP
05/24/2011
Summer School, IIIT HYderabad
30
80
Non-Pronominal Resolution Module

Use CRFs, a machine learning method

Preparing the Training data for CRFs learning





All positive pairs of NPs, and negative pairs of NPs
are taken for training
Positive pairs are the NPs and the anaphor
All NPs between an anaphor and antecedent are the
negative NPs
Do not consider the NPs containing the pronouns,
We consider the NP on the left side as antecedent NP
and NP on the right side as anaphor NP
05/24/2011
Summer School, IIIT HYderabad
81
Features – Non-pronominal Resolution
Module

The features used for training are

Distance feature



Definite NP


If the antecedent NP is a definite NP, has value 1 else 0
Demonstrative NP


the possible values are 0,1,2,….
Calculated based on number of sentences between NPs
If the antecedent NP is demonstrative, has value 1 else 0
String match


05/24/2011
the possible values are between 0 and 1
calculated as ratio of the number of words matched between
the NPs and the total number of words of the anaphor NP.
Summer School, IIIT HYderabad
82
Features – Non-pronominal Resolution Module
(Contd…)





Number Agreement

use the gender data file (Bergsma and Lin, 2006) provided by CoNLL

Also use POS information
Gender agreement

use the gender data file (Bergsma and Lin, 2006) provided by CoNLL
Alias feature

the possible values are 0 or 1.

this is obtained using three methods
 Comparing the head of the NPs, if both are same then scored as 1
 If both the NPs start with NNP or NNPS POS tags, and if they are
same then scored as 1
 Looks for Acronym match, if one is an acronym of other it is scored
as 1
Proper NP

If both NPs are proper NPs then value of 1 else 0
NE Tag Info

If NE tag present then value of 1 else 0
05/24/2011
Summer School, IIIT HYderabad
83
Non-Pronominal Resolution Module
(Contd…)

The semantic class information (noun
category) obtained from the WordNet
is used for the filtering purpose.

05/24/2011
The pairs which do not have semantic
feature match are filtered out.
Summer School, IIIT HYderabad
84
Complete Coreference Chain Building


The Coreferring pairs obtained from pronominal
resolution system and Non-pronominal system are
merged to generate the complete coreference chains.
The merging is done as follows:




A member of a coreference pair is compared with all the
members of the coreference pairs identified
if it occurs in anyone of the pair, then the two pairs are grouped
this process is done for all the members of the identified pairs
and the members in each group are aligned based on their
position in the document to form the chain
We show the sample outputs for building chain in the
next few slides
05/24/2011
Summer School, IIIT HYderabad
85
Evaluation



The data used for training and testing is the
English portion of the Ontonotes
The training and development data
consisted of 1876 files from different genres
viz., Newswire, Broadcast news, Broadcast
conversation, Web blogs, magzine articles
The metrics used for evaluating the
complete system are



05/24/2011
MUC
B-Cubed
CEAFE
Summer School, IIIT HYderabad
86
Evaluation


We have performed evaluation of the
Pronominal resolution module separately
using the development data
We perform the evaluation of non-anaphor
detection engine

In the evaluation of pronominal resolution
module we studied how the non-anaphoric
pronoun detection improves
05/24/2011
Summer School, IIIT HYderabad
87
Results – Pronominal Resolution Module
Evaluation of Non-anaphoric
pronoun detection component
Evaluation of Pronominal resolution
module
Actual
(gold
standar
d)
System
identifie
d
Correctl
y
Accur
acy
(%)
System type
Total
Anaph
oric
Prono
uns
System
identifie
d
pronoun
s
System
correctly
Resolved
Pronouns
Prec
(%)
Anaphoric
Pronouns
939
908
96.6
Without nonanaphoric
pronoun
detection
939
1099
693
63.1
Nonanaphoric
pronouns
160
81
50.6
939
987
693
70.2
Total
1099
989
89.9
With
nonanaphoric
pronoun
detection
Type
pronoun
of
05/24/2011
Summer School, IIIT HYderabad
88
Results– Complete System
Metric
Mention Detection
Coreference Resolution
Recall
Precision
F1
Recall
Precision
F1
MUC
68.1
61.5
64.6
52.1
49.9
50.9
B-CUBED
68.1
61.5
64.6
66.6
67.6
67.1
CEAFE
68.1
61.5
64.6
42.8
44.9
43.8
05/24/2011
Summer School, IIIT HYderabad
89
Results Discussion

On analysis of the output we found
mainly three types of errors. They are

Newly invented chains



05/24/2011
The system identifies new chains that are not
found in the gold standard annotation.
This reduces the precision of the system.
This is because of the string match as one of
the features.
Summer School, IIIT HYderabad
90
Results Discussion (Contd…)

Only head nouns in the chain
 system while selecting pair for identifying
coreference, the pair has only the head
noun instead of the full phrase.
 In the phrase “the letters sent in recent
days”, the system identifies “the letters”
instead of the whole phrase.
 This affects both the precision and recall
of the system.
05/24/2011
Summer School, IIIT HYderabad
91
Conclusion



Presented a coreference resolution system which
combines the pronominal resolution using refined
salience based approach with non-pronominal
resolution using CRFs, machine learning approach.
Non-anaphoric pronouns identification improves the
precision.
In non-pronominal resolution algorithm, the string match
feature is an effective feature.



But, this feature is found to introduce errors.
We need to add additional contextual and semantic feature to
reduce above said errors.
The results on the development set are encouraging.
05/24/2011
Summer School, IIIT HYderabad
92
Problems in Anaphora
Resolution

Ambiguous Sentences




We gave the bananas to the monkeys because
they were hungry.
We gave the bananas to the monkeys because
they were ripe.
We gave the bananas to the monkeys because
they were here.
Complement anaphora


05/24/2011
(1) Only a few of the children ate their ice-cream. They ate the
strawberry flavour first.
(2) Only a few of the children ate their ice-cream. They threw it
around the room instead
Summer School, IIIT HYderabad
93
The Prime Minister of New Zealand visited us yesterday.
The visit was the first time she had come to New York
since 1998.
If the second sentence is quoted by itself, it is necessary to
resolve the anaphor:
 The visit was the first time the Prime Minister of New
Zealand had come to New York since 1998.
Although of course, as The Prime Minister of New Zealand
is an office of state and she would seem to refer to the
person currently occupying that office, it could quite easily
be that the Prime Minister of New Zealand had visited New
York since 1998 and before the present day, whilst the
present incumbent she had not.



05/24/2011
Summer School, IIIT HYderabad
94
Corpus Creation for ML
Method

Annotation Guide lines




05/24/2011
PALINKA
MUC
GATE
Give examples
Summer School, IIIT HYderabad
95
Tools




GATE
Java-RAP (pronouns)
GUITAR (Poesio & Kabadjov, 2004;
Kabadjov, 2007)
BART (Versleyet al, 2008)
05/24/2011
Summer School, IIIT HYderabad
96
TASKS and Conferences



CoNLL shared task
Conference
Discourse Anaphora and Anaphora
Resolution Colloquium: DAARC
05/24/2011
Summer School, IIIT HYderabad
97
Where it is required?




Machine Translation
Information Extraction
Summarization
And in……….almost all NLU applications
05/24/2011
Summer School, IIIT HYderabad
98
References



Massimo Poesio Slides: “Anaphora
resolution for Practical task”
Ruslan Mitkov: “MARS a Knowledge
Poor anaphora resolution system”
MORE FROM OTHER SOURSES
05/24/2011
Summer School, IIIT HYderabad
99
Thank You