A hybrid approach for anaphora resolution in hindi

Post on 08-Jul-2015

130 views 0 download

Tags:

description

Anaphora Resolution

Transcript of A hybrid approach for anaphora resolution in hindi

A Hybrid Approach for

Anaphora Resolution in Hindi

By :-

Mujadia Vandan (201323602)

Krishnakant Vishwakarma (201321582)

Paper courtesy (Praveen et.al)

Contents

What is Anaphora/Cataphora?

Types of Anaphora

Dataset

◦ Annotation scheme

Approaches

Results

Applications

What Anaphora/ Cataphora

means?

Cataphora

Anaphora (Act of carrying back)

Types of Anaphora

Reflexive Pronoun

{"अपने","अपना","अपनी","खदु"}Spatial/Locative Pronoun

{"वहाां","यहाां","जहाां","वहााँ"}Relative Pronoun

{"जजसका","जजससे","जो","जजसे"}Personal Pronoun

First, Second, Third person.

Event

Data set(Hindi/Urdu Treebank)

Annotation Scheme

Extends the SSF format by introducing the

attribute value pair ‘ref’.

ref=‘..%Sentence-id%NP(Chunk-id)’.

Ex: ref=‘..%3%NP17’

ref=‘NP5’.

Marked on PRP in SSF format.

Approach

Hybrid Approach

◦ Rule Base (Base line)

◦ Learning Base(Decision tree, SVM)

Hybrid System

Rule base(Reflexive)

Rule base(Locative)

Like this they have used different dependency(Karaka relation) to resolve relative and personal anaphora in order of (k1 >k2 >r6 >k4 >k3 >others) for personal anaphora resolution.

Features(for Learning)

Number : singular, plural, honorific

Named Entity categories: ‘Person’, ’

Organization’ ,’Location ’ ,’ Number/Time’

Distance feature: #NP chunks and

#sentences between the pronoun and the

candidate NP.

Animacy : ‘human’, ‘animate’, ‘rest’.

Learning Instances

Result

Applications

Machine Translation

Information Extraction

Summarization

And in……….almost all NLP applications

Thank You