The following definitions are given for the word bark:

(i) the hard outer covering of a tree

(ii) the short, loud noise that a dog makes
(iii) to make a loud, short noise or noises
(iv) to give orders, ask questions, etc. in a loud unfriendly way
(a) Apply Lesk algorithm to the following sentence to
disambiguate the word bark (marks will be given only if the
details are given).
The dog barked hard in front of the tree
(b) Apply Simplified algorithm to the above sentence to
disambiguate the word bark (marks will be given only if the
details are given).
Note: (iii) definition is also related to Dog’s noise.
(c) Compare the results with (a) & (b) and give the analysis.
Discourse Analysis –

Pronoun resolution
Anaphora resolution,
Constraints &
Algorithm
 What is Anaphor Resolution
Anaphora is an instance of an expression referring to
another and the task of resolving the antecedent and
anaphor pair is called anaphora resolution.
Ex: anaphor

John went home because he felt sick.

antecedent

Here ‘he’ is referring to ‘John’

Anaphoric Reference vs Cataphoric Reference

Cataphoric Reference:

A cataphoric reference refers to something within a text

that has not yet been identified.
Ex:
He is quick as panther. [More Text……]. His name is
Jason Bourne. [More Text…]

Frequently seen in Novel.

Hard to Resolve.
 Approaches for Anaphor Resolution

Rule Based system:

Lappin & Leass Algorithm

Hobbs Algorithm
 Necessary Tools Required

1. Pos Tagger – To extract the Markables

2. Name Entity Recognition - To identify Name, Place, Organization

3. Wordnet – For Synonyms

4. Gender Database – For Male , Female or Unknown

 Lappin & Leass Algorithm

1.Personal Pronoun: I, You, He, She etc.

2.Reflexive Pronoun: Myself, Yourself etc.

3.Possessive Pronoun: Mine, Yours, His etc

4.Reciprocal Pronoun: Each other, One another

5.Pleonastic Pronoun : It

6. Demonstrative Pronoun: This, That etc.

7. Relative Pronoun: Whose, Which etc

8. Interrogative Pronoun: What, Were etc

9. Negative Pronoun: Nothing, Nobody etc

Syntactic and Semantic constraints on
coreference
• Number agreement
John has an Acura. It is red.

• Person agreement
*John and Mary have Acuras. We love
them (where We=John and Mary)
• Gender agreement
John has an Acura. He/it/she is attractive.
 Contd..
• Syntactic constraints
John bought himself a new Acura
(himself=John)

John bought him a new Acura

(him = not John)
Number agreement in English pronominal system

Singular Plural Unspecified

She, her, he, We, us, they, you

him, his, it them

John has a new Acura. It is red.

John has three new Acuras. They are red.
 Person and Case Agreement
First Second Third
Nominative I, We you He, she, they
(subject)

Accusative Me, Us you Him, her, them

(object)

Genitive My, Our your His, her, their

You and I have Acuras. We love them.

John and Mary have Acuras. They love them.

* John and Mary have Acuras. We love them.

Masculine Feminine nonpersonal

He, him, his She, her it

John has an Acura. He is attractive. (he=John, not

the Acura)

John has an Acura. It is attractive. (it=the Acura,

not John)
 Selectional Restrictions
• The selectional restrictions that a verb
places on its arguments may be responsible for
eliminating referents.

John parked his Acura in the garage. He

had driven it around for hours.

* There are two possible referents for it, the

Acura and the garage. The verb drive
eliminates the garage.
 Preferences in Pronoun Interpretation
• Recency: Entities introduced in the recent utterances are
more salient than those introduced from utterances
further back.
John has an Integra. Bill has a Legend. Mary likes to
drive it.
it is more likely to refer to Legend than Integra.

Grammatical Role: Subject preference

– John went to the Acura dealership with Bill. He bought an
Integra.
(he = John)
• Bill went to the Acura dealership with John. He bought
an Integra
(he =Bill)

• John and Bill went to the Acura dealership. He bought

an Integra.
(he =??)
 Repeated Mention preference

Ex: John needed a car to get to his new job.

He decided that he wanted something sporty.
Bill went to the Acura dealership with him.
He bought an Integra.
 Parallelism Preference
Preferences that appear to be induced by
parallelism effects:

• Mary went with Sue to the Acura dealership.

Sally went with her to the Mazda dealership.
(her = Sue)

• Marry went with Sue to the Acura dealership.

Sally told her not to buy anything.
(her =Marry)
 Verb Semantics Preferences

John telephoned Bill. He lost the pamphlet on Acuras.

John criticized Bill. He lost the pamphlet on Acuras.

* Implicit causality
• Implicit cause of telephoning is subject.
• Implicit cause of criticizing is object.
Lappin and Leass (1994) Algorithm
• A straightforward algorithm for pronoun
interpretation that takes many of these
consideration.

• Two types of operation performed:

 Discourse model update and Pronoun resolution
•  How does it work?
- When a NP that evokes a new entity is encountered, a
representation for it must be added to the discourse model

- Compute the degree of salience (i.e.; salience value) for

the NP. It is calculated as the sum of weights assigned by a
set of salience factors

- The weights that each factor assigns to an entity in the

discourse model are cut half each time a new sentence is
processed. Capture the Recency preference.
 Salience factor types with initial weights
Factor type Initial weight
Sentence recency 100

Subject emphasis 80

Existential emphasis 70

Accusative (direct object) emphasis 50

Indirect object and oblique complement 40

emphasis
Head noun emphasis 80

Non-adverbial emphasis 50
•  These five positions are exemplified by the position
of the italicized phrases in the following respectively.

- An Acura Integra is parked in the lot. (subject)

- There is an Acura Integra parked in the lot.

(existential predicate nominal)

- John parked an Acura Integra in the lot. (object)

- John gave his Acura Integra a bath. (indirect object)

- Inside his Acura Integra, John showed Susan his new

CD player. (demarcated adverbial PP; those separated
by punctuation, as with comma in example)

- The Acura Integra gets 80 points for being denoted

by a head noun.
Not true for following case:
* The owner’s manual for an Acura Integra is on John’s
desk.
• Pronoun Resolution:
• Factor in two more salience weights
- Role Parallelism +35
- Cataphora -175
• These two can not be calculated
independently of the pronoun, and
thus cannot be calculated during
discourse model update
•  before this update ←initial Salience
value
•  After the update ← Final salience
value
An Example Lappin and Leass
algorithm(1994)
1. Bob wanted an Acura Integra. 2. John saw it at the dealership.
3. He showed it to him. 4. He bought it.

Rec Subj. Exist Obj Ind-obj Non- Head Total

Adv N
B 100 80 50 80 310
o
b

I 100 50 50 80 280
n
t
e
g
r
a
 1. Bob wanted an Acura Integra. 2. John saw it at the dealership.
3. He showed it to him. 4. He bought it.
• There are no pronouns to be resolved in this sentence, so we move on
to next, halving the above factors.
Referent Phrases Value
Bob {Bob} 155
Integra {An Acura Integra} 140

The first noun phrase in second sentence is John that

introduces new discourse referent. It is in the subject
position. So, it will get (Rec: 100+ subj: 80+ Non adv:
50+ HN:80 =310)

Referent Phrases Value

Bob {Bob} 155
Integra {An Acura Integra, it} 140 + 280 = 420
John {John} 310
1. Bob wanted an Acura Integra. 2. John saw it at the dealership.
3. He showed it to him. 4. He bought it.
• Dealership will get (rec:100+non-adv 50+ HN:80)=230

Referent Phrases Value

Bob {Bob} 155

Integra {An Acura Integra, it} 140 + 280 = 420

John {John} 310

Dealership {the dealership} 230

1. Bob wanted an Acura Integra. 2. John saw it at the dealership.
3. He showed it to him. 4. He bought it.
• Next sentence: Divide the previous salience factors by two

Referent Phrases Value

Bob {Bob} 155/2= 77.5
Integra {An Acura Integra, 420/2=210
it}
John {John} 310/2=155
Dealership {the dealership} 230/2=115
 1. Bob wanted an Acura Integra. 2. John saw it at the dealership.
3. He showed it to him. 4. He bought it.

• He will be resolve to John due to high

salience factors. Add Salience
factor(recency:100+ subpos: 80+
nonadv:50 + HN:80)=310

•  It to Integra: It will get 280 (rec:100+

obj:50+non-adv:50+ HN: 80)

•  him can not be John due to syntactic

constraints. So, him will be resolved to
Bob . (rec:100+indObj:40+non-adv 50+
HN:80)=270
 1. Bob wanted an Acura Integra. 2. John saw it at the dealership.
3. He showed it to him. 4. He bought it.

Referent Phrases Value

Bob {Bob, him} 77.5+270=347.5

Integra {an Acura Integra, it 1, 210+280=490

it2 }

John {John, he1} 155+310=465

Dealership {the dealership} 115

1.Bob wanted an Acura Integra. 2. John saw it at the dealership.
3. He showed it to him. 4. He bought it.
Sentence:4 Divide the previous salience factors by two
Referent Phrases Value

Bob {Bob, him} 347.5/2= 173.75

Integra {an Acura Integra, it 1, 490/2=245

it2 }

John {John, he 1} 465/2=232.5

Dealership {the dealership} 115/2=57.5

1.Bob wanted an Acura Integra. 2. John saw it at the dealership.
3. He showed it to him. 4. He bought it.

Looks like he will be resolve to John

and it to Acura!
Resolve the pronouns from the following paragraph
using Lappin &Leass algorithm:
•The Salience factors and weights in Lappin and Leass’s system are as
follows:
•Sentence Recency 100
•Subject emphasis 80
•Head noun emphasis 80
•Existential emphasis 70
•Accusative emphasis 50
•Non-adverbial emphasis 50
•Indirect object and 40
•oblique complement emphasis
•
•
 Example:
• 1.George (310) needed a new car(280).
2.He (310) did not find previous car
(280).
3. Jerry (310) went with him (280) to the
car dealers (230).
4.He (310) also needs a car (280) for
himself(280).
5.They (310) bought those (230).