SYMPOSIUM ON BIAS AND DIVERSITY IN IR(aka LIVINGKNOWLEDGE SUMMER SCHOOL)

LivingKnowledge Consortium

ESSIR Summer School 2011August 31, 2011, Koblenz, Germany

Bias and Diversity in IR, August 31, 2011 2

What are the different

perspectives?

Are some sources biased?

What types of images are used

VISION

Diversity and bias in the Web today:• High diversity of content and content

creators • But: no systematic support to explore

the diversity

Vision of Diversity-related research: • Make diversity, bias and evolution traceable,

understandable and exploitableAim and scope of this symposium: • Give an overview of leading-edge technology for

dealing with diversity and bias in IR3Bias and Diversity in IR, August 31, 2011

CONTEXT: LIVING KNOWLEDGE PROJECT

• Part of the presented material are relevant results from the LivingKnowledge project

• EU FET Project since February 2009• Project Goal:

Improve navigation and search in large cross-media datasets and the Web by considering diversity, time, opinions and bias

• Relevant project results:– Testbed with more than 40 components (will be shown and

used today)– Innovative methods and algorithms, e.g sentiment extraction

from images, image forensics, contextualization of facts– First applications of diversity-aware technology,

e.g. Time explorer

Bias and Diversity in IR, August 31, 2011 4

CONTEXT: LIVING KNOWLEDGE PROJECT

• Relevant project results:– Testbed with more than 40 components (will be shown and

used today)– Many innovative methods and algorithms, e.g. sentiment

extraction from images, image forensics, contextualization of facts;

– First applications of diversity-aware technology, e.g. Time explorer

Bias and Diversity in IR, August 31, 2011 5

Module Status Documentation

HTML Content Analyser In testbed YesCodebook Aggregator In testbed YesURL Extractor In testbed YesDocument Layout Extractor In testbed YesPlaces Annotator In testbed YesEntity Disambiguator In testbed YesEXIF Image metadata Extractor In testbed YesImage Colourfulness In testbed YesImage Naturalness In testbed Yesimage Sentiment Annotator In testbed YesQuote Extractor In testbed YesPhrase Extractor In testbed YesTimeML Annotator In testbed YesDictionary Annotator In testbed YesFact Annotator In testbed Yes‘Best Image’ Annotator In testbed No

Contrast adjustment impacts on opinion

perceived by the users.

Impact of Contrast adjustment on opinion perceived by the users

SYMPOSIUM STRUCTURE & OVERVIEW

• Structured into 3 parts

• Part I: Discovering Facts and Opinions and their Evolution over Time

• Part II: Discovering Dias and Diversity in Multimedia Content

• Part III: A Testbed for Diversification in Search– Demonstration of selected technology– Time for further in depth discussions

Bias and Diversity in IR, August 31, 2011 6

Part I: EXTRACTION OF OPINIONS FROM THE WEB

• Introduces basic concepts in opinion extraction• Overview of standard coarse- and fine-grained

opinion extraction methods• Introduction to domain adaptation• Overview of opinion resources• Shows some recent research results in the area

7Bias and Diversity in IR, August 31, 2011

PART I: TEMPORAL FACT EXTRACTION, DISAMBIGUATION, AND EVOLUTION

• Factual knowledge evolves over time

Bias and Diversity in IR, August 31, 2011

1965-1981Bodybuilder

2002-2011Politician

1982-2002Actor

Goal: Exploring the temporality of entities Extract entities from Web documents Disambiguate them to canonical entities Aggregate facts to investigate their dynamics

8

• Why look at images?• Image representation and understanding• Information extraction techniques

PART II: A BRIEF INTRODUCTION TO EXTRACTING INFORMATION FROM IMAGES

INFORMATION EXTRACTION FROM MULTIMEDIA CONTENT ON THE SOCIAL WEB

• How to exploit combined information from visual data and meta data

• Photos on the social web: attractiveness, maps and sentiment

• Videos on the social web: tagging9Bias and Diversity in IR, August 31, 2011

FROM THEORY TO PRACTICE

Part III: A BRIEF INTRODUCTION TO THE DIVERSITY ENGINE TESTBED

Learn how to run these tools and visualize the resultsAdd your own toolsBuild diversity-aware applications

<lk-application logDir="log" appDir="devapps/example"><corpus dir="corpora/examples/smallarc" format="arc"/><image-pipeline>

<annotators></annotators>

</image-pipeline><pipeline>

<annotators><annotator exec="./opennlp"/><annotator exec="./sst"/>

</annotators></pipeline><visualize/><indexer solrHomeDir="solr/solr“

solrDataDir="solr/solr/data“converter="conf/testbed/tutorial-lk2solr.xml"/>

<searcher appTitle="LivingKnowledge - Example Application"appShortTitle="Example Application"appUrl="http://localhost:8983/solr/">

<facets><facet field="per" description="Person"/><facet field="loc" description="Location"/>

</facets></searcher>

</lk-application>

10Bias and Diversity in IR, August 31, 2011