Multi-Perspective Question Answering

3/25/2003


Click here to start


Table of Contents

Multi-Perspective Question Answering

Participants

Part I: Overview

Problem

Our Work will Support

Task: Conceptualization

Task: Annotation

Task: Summary Representation of Opinions

Task: Automate Manual Annotations

Task: Organizing Perspective Segments

Evaluation

Part II: Conceptualization

Introduction

Influences on Perspective

Lexical Clues to Subjectivity

Attitude Expression

Attitude Typologies

Attitude Types

Types of AE Combinations

Subjectivity: Relation to Ideology

Solution Architecture

Part III: Annotation

Annotation

Two Ways of Expressing Private States

Two Ways of Expressing Private States

Nested Sources

OnlyFactive

OnlyFactive

Example

Example

Other attributes

Documents and Annotators

PPT Slide

Pilot Agreement Evaluation: ons and subjective elements

Overlapping Spans: Ons

Overlapping Spans: Expressive Subjective Elements

Untrained Annotators

Spans, Untrained Annotators

Overlapping Spans: Expressive Subjective Elements

PPT Slide

Spans, All Annotators

Pilot Agreement Evaluation: OnlyFactive

OnlyFactive Kappa Values

Future work

Part IV: Summary Representations of Opinions

Example

Sentence 1

Sentence 1: Low-level

Sentence 2

Sentence 2: Low-level

Sentence 3

Sentence 3: Low-level

Sentence 4

Sentence 4: Low-level

Sentence 5

Sentence 5: Low-level

Sentence 6

Sentence 6: Low-level

Sentence 7

Sentence 7: Low-level

Sentence 8

Sentence 8: Low-level

Sentence 9

Sentence 9: Low-level

Sentence 10

Sentence 10: Low-level

All Low-level Annotations

Creating Summary Representations of Opinions with Writer As Sole Source

Creating Summary Representations of Opinions with Writer As Sole Source

Creating Summary Representations of Opinions for Other Sources

Creating Summary Representations of Opinions

Creating Summary Representations of Opinions for Other Sources

Creating Summary Representations of Opinions

Summaries

Part V: Automating Manual Annotations

Solution Architecture

Learning Architecture

Learning problems

Pilot experiments

Future Work

Part VI: Clustering

Solution Architecture

Application Architecture

End-User Evaluation: Goal

Manual Clustering Exploration

Manual Clustering: Analysis

Manual Clustering: Issues

Manual Clusters: “Agreement”

Manual Clustering: Overall

User Task: Topic

User Task: Question

User Task: Question Response

User Task: Cluster Feature

User Task: Results

FBIS Document Collection

FBIS Document Collection Features

WNC Collection (Docs,Topics)

WNC Distribution

Topics

1. Simple evaluation

Simple Evaluation: Details

Simple Evaluation: Comparison

Topic 1: Zimbabwe - Base

Topic 1: Zimbabwe - Opinion

Topic 3: Kyoto – Base/Opinion

2. Retrospective Evaluation

Retrospective

Retrospective: Zimbabwe

Retrospective: Kyoto

End-User Summary

Future work

Future Work 2.

Summary of Future Work

Author: Janyce M. Wiebe

Email: bnunan@mitre.org

Home Page: http://nrrc.mitre.org/

Download presentation source