*Erasmus University Rotterdam P.O. Box 1738, NL-3000 DR Rotterdam, the Netherlands † Teezir BV Wilhelminapark 46, NL-3581 NL, Utrecht, the Netherlands.

Slides:



Advertisements
Similar presentations
Victorian Curriculum and Assessment Authority
Advertisements

1 Evaluation Rong Jin. 2 Evaluation  Evaluation is key to building effective and efficient search engines usually carried out in controlled experiments.
RCQ-ACS: RDF Chain Query Optimization Using an Ant Colony System WI 2012 Alexander Hogenboom Erasmus University Rotterdam Ewout Niewenhuijse.
Polarity Analysis of Texts using Discourse Structure CIKM 2011 Bas Heerschop Erasmus University Rotterdam Frank Goossen Erasmus.
Learning Semantic Information Extraction Rules from News The Dutch-Belgian Database Day 2013 (DBDBD 2013) Frederik Hogenboom Erasmus.
Semantic News Recommendation Using WordNet and Bing Similarities 28th Symposium On Applied Computing 2013 (SAC 2013) March 21, 2013 Michel Capelle
What is the purpose of your essay? To argue/contend. Your essay will always work best if you have a strong contention and you argue it enthusiastically.
A Linguistic Approach for Semantic Web Service Discovery International Symposium on Management Intelligent Systems 2012 (IS-MiS 2012) July 13, 2012 Jordy.
Exploiting Discourse Structure for Sentiment Analysis of Text OR 2013 Alexander Hogenboom In collaboration with Flavius Frasincar, Uzay Kaymak, and Franciska.
Connecting Customer Relationship Management Systems to Social Networks 7th International Conference on Knowledge Management, Services, and Cloud Computing.
Determining Negation Scope and Strength in Sentiment Analysis SMC 2011 Paul van Iterson Erasmus School of Economics Erasmus University Rotterdam
Exploiting Emoticons in Sentiment Analysis SAC 2013 Daniella Bal Erasmus University Rotterdam Flavius Frasincar Erasmus University.
A Framework for Automated Corpus Generation for Semantic Sentiment Analysis Amna Asmi and Tanko Ishaya, Member, IAENG Proceedings of the World Congress.
IVITA Workshop Summary Session 1: interactive text analytics (Session chair: Professor Huamin Qu) a) HARVEST: An Intelligent Visual Analytic Tool for the.
Predicting Text Quality for Scientific Articles Annie Louis University of Pennsylvania Advisor: Ani Nenkova.
Predicting Text Quality for Scientific Articles AAAI/SIGART-11 Doctoral Consortium Annie Louis : Louis A. and Nenkova A Automatically.
Erasmus University Rotterdam Frederik HogenboomEconometric Institute School of Economics Flavius Frasincar.
WEBQUEST Let’s Begin TITLE AUTHOR:. Let’s continue Return Home Introduction Task Process Conclusion Evaluation Teacher Page Credits This document should.
RCQ-GA: RDF Chain Query Optimization using Genetic Algorithms BNAIC 2009 Alexander Hogenboom, Viorel Milea, Flavius Frasincar, and Uzay Kaymak Erasmus.
Sentiment Lexicon Creation from Lexical Resources BIS 2011 Bas Heerschop Erasmus School of Economics Erasmus University Rotterdam
Chapter 1 INTRODUCTION TO ACCOUNTING INFORMATION SYSTEMS
Automatically Annotating Web Pages Using Google Rich Snippets 11th Dutch-Belgian Information Retrieval Workshop (DIR 2011) February 4, 2011 Frederik Hogenboom.
Detecting Economic Events Using a Semantics-Based Pipeline 22nd International Conference on Database and Expert Systems Applications (DEXA 2011) September.
Analyzing Sentiment in a Large Set of Web Data while Accounting for Negation AWIC 2011 Bas Heerschop Erasmus School of Economics Erasmus University Rotterdam.
Types of Essays... and why we write them.. Why do we write essays? Hint: The answer is NOT ‘because sir/miss told me to’
Word Sense Disambiguation for Automatic Taxonomy Construction from Text-Based Web Corpora 12th International Conference on Web Information System Engineering.
Sentiment Analysis with a Multilingual Pipeline 12th International Conference on Web Information System Engineering (WISE 2011) October 13, 2011 Daniëlla.
Opinion mining in social networks Student: Aleksandar Ponjavić 3244/2014 Mentor: Profesor dr Veljko Milutinović.
Mining and Summarizing Customer Reviews Minqing Hu and Bing Liu University of Illinois SIGKDD 2004.
Erasmus University Rotterdam Introduction Nowadays, emerging news on economic events such as acquisitions has a substantial impact on the financial markets.
Dr. MaLinda Hill Advanced English C1-A Designing Essays, Research Papers, Business Reports and Reflective Statements.
Erasmus University Rotterdam Introduction With the vast amount of information available on the Web, there is an increasing need to structure Web data in.
A News-Based Approach for Computing Historical Value-at-Risk International Symposium on Management Intelligent Systems 2012 (IS-MiS 2012) Frederik Hogenboom.
1 Academic Skills Tips for Essay Writing. 2 Outline of today’s lecture Academic skills Essay writing Paraphrasing Summarizing.
Put the Title of the WebQuest Here A WebQuest for xth Grade (Put Subject Here) Designed by (Put Your Name Here) Put Your Address Here Put some interesting.
Ontology Updating Driven by Events Dutch-Belgian Database Day 2012 (DBDBD 2012) November 21, 2012 Frederik Hogenboom Jordy Sangers.
Evaluating a Research Report
Designing Ranking Systems for Consumer Reviews: The Economic Impact of Customer Sentiment in Electronic Markets Anindya Ghose Panagiotis Ipeirotis Stern.
Going Deeper with Mark Twain A WebQuest for 10th Grade Composition Designed by Sandy Schaufelberger Wes-Del High School, Gaston IN
How Useful are Your Comments? Analyzing and Predicting YouTube Comments and Comment Ratings Stefan Siersdorfer, Sergiu Chelaru, Wolfgang Nejdl, Jose San.
Semantics-Based News Recommendation with SF-IDF+ International Conference on Web Intelligence, Mining, and Semantics (WIMS 2013) June 13, 2013 Marnix Moerland.
Erasmus University Rotterdam Introduction Content-based news recommendation is traditionally performed using the cosine similarity and TF-IDF weighting.
Towards Cross-Language Sentiment Analysis through Universal Star Ratings KMO 2012 Malissa Bal Erasmus University Rotterdam Flavius.
What is an Annotated Bibliography? First, what is an annotation?  More than just a brief summary of an article, book, Web site etc.  It combines summary.
MOTIVATION AND CHALLENGE Big data Volume Velocity Variety Veracity Contributor Content Context Value 5 Vs of Big Data 3 Cs of Veracity.
Poorva Potdar Sentiment and Textual analysis of Create-Debate data EECS 595 – End Term Project.
Software Quality in Use Characteristic Mining from Customer Reviews Warit Leopairote, Athasit Surarerks, Nakornthip Prompoon Department of Computer Engineering,
English Language Services
Lexico-semantic Patterns for Information Extraction from Text The International Conference on Operations Research 2013 (OR 2013) Frederik Hogenboom
Blog Summarization We have built a blog summarization system to assist people in getting opinions from the blogs. After identifying topic-relevant sentences,
Creating Subjective and Objective Sentence Classifier from Unannotated Texts Janyce Wiebe and Ellen Riloff Department of Computer Science University of.
Welcohhhto week 8 No reading or discussion today Portfolio and oral presentation feedback Assignment 3 Discussion of assignment dates 3 facets of critiquing.
Critical Essays National 5. Purpose of the Critical Essay A DISCURSIVE essay on a text Presenting an ARGUMENT – clear line of thought which is linked.
1 Generating Comparative Summaries of Contradictory Opinions in Text (CIKM09’)Hyun Duk Kim, ChengXiang Zhai 2010/05/24 Yu-wen,Hsu.
Advantages of Query Biased Summaries in Information Retrieval by A. Tombros and M. Sanderson Presenters: Omer Erdil Albayrak Bilge Koroglu.
1 Adaptive Subjective Triggers for Opinionated Document Retrieval (WSDM 09’) Kazuhiro Seki, Kuniaki Uehara Date: 11/02/09 Speaker: Hsu, Yu-Wen Advisor:
Barcelona Declaration of Measurement Principles Presented June 17, 2010 Revised June 20, 2010 Final July 19, 2010 Global Alliance ICCO Institute for Public.
Semantics-Based News Recommendation International Conference on Web Intelligence, Mining, and Semantics (WIMS 2012) June 14, 2012 Michel Capelle
Proposal Daniel Michlits h Research Seminar System Analyses.
Argumentative Writing Grades College and Career Readiness Standards for Writing Text Types and Purposes arguments 1.Write arguments to support a.
Date: 2013/9/25 Author: Mikhail Ageev, Dmitry Lagun, Eugene Agichtein Source: SIGIR’13 Advisor: Jia-ling Koh Speaker: Chen-Yu Huang Improving Search Result.
Research Methodology II Term review. Theoretical framework  What is meant by a theory? It is a set of interrelated constructs, definitions and propositions.
1. 1.To examine the information included in business reports. 2.To understand how to organize documents in order to ensure clear communication. 3.To analyze.
Web Analytics & Social Media Monitoring Assignment Briefing June and September 2013 Clive Whysall CAM Examiner.
RESEARCH MOTHODOLOGY SZRZ6014 Dr. Farzana Kabir Ahmad Taqiyah Khadijah Ghazali (814537) SENTIMENT ANALYSIS FOR VOICE OF THE CUSTOMER.
RI.6.2 Determine a central idea of a text and how it is conveyed through particular details; provide a summary of the text distinct from personal opinions.
Introductions and Conclusions
National 5 Critical Essays.
E-Commerce Theories & Practices
National 5 Critical Essays.
Presentation transcript:

*Erasmus University Rotterdam P.O. Box 1738, NL-3000 DR Rotterdam, the Netherlands † Teezir BV Wilhelminapark 46, NL-3581 NL, Utrecht, the Netherlands An Empirical Study for Determining Relevant Features for Sentiment Summarization of Online Conversational Documents WISE 2012 Gino Mangnoesing* Arthur van Bunningen † Alexander Hogenboom* Flavius Frasincar* Frederik Hogenboom* November 30, 2012

Introduction (1) The Web offers an overwhelming amount of textual data, containing traces of sentiment Information monitoring tools for tracking sentiment are of paramount importance for today's businesses WISE

Introduction (2) A reliable analysis of the sentiment of authors of user- generated content is crucial for reputation management There is a great need for knowing what people (dis)like about products or brands and why they feel this way A major challenge lies in the identification of text segments capturing and explaining the sentiment of a text as a whole WISE

Sentiment Analysis Sentiment analysis is typically focused on determining the polarity of natural language text: positive, negative, neutral, or anything in between Many practical applications typically aim to capture a general mood (reviews, consumer confidence, politics) As such a quantification of sentiment in a score or class alone is not sufficient anymore, methods for summarizing opinionated texts are gaining interest WISE

Sentiment Summarization (1) The goal of sentiment summarization is to distinguish relevant from irrelevant text fragments with respect to conveying the overall sentiment of an opinionated text Existing work considers using as feature whether or not a text segment: –Discusses one or more (sub)aspects of the topic –Contains an opinion about the topic –Is rather positive or rather negative (high intensity) –Is part of the introduction of the text –Is part of the conclusion of the text –Contains an adjective –Contains an adverb WISE

Sentiment Summarization (2) It may also be relevant whether or not a text segment: –Addresses an event or experience described in the text –Contains an advice or recommendation –Contains an argument supporting an opinion, vision, or statement in the text –Contains or is part of a comparison in the text –Contains non-stopwords that are present in the title of the text –Contains a list or sequence –Is relatively long –Is relatively short WISE

Feature Evaluation (1) We have performed an analysis on a collection of 60 Dutch conversational documents (forum posts) about the Dutch company Ziggo For each document, 7 candidate summary sentences were selected, such that each of our 15 considered features was as well-represented as possible Each of these 420 sentences was evaluated by 3 out of 9 human annotators: –Sentiment of text (negative, neutral, positive) –Sentiment of sentence (negative, neutral, positive) –Relevance of sentence (highly irrelevant, irrelevant, relevant, highly relevant) –Applicability of each of our 15 features (true, false) WISE

Feature Evaluation (2) WISE

Feature Evaluation (3) Scores were aggregated per sentence: –Sentiment scores were averaged over all 3 annotators –Relevance scores were mapped onto binary classes (relevant and irrelevant) and subsequently aggregated by means of majority voting over all 3 annotators –Applicability of features was aggregated per feature by means of majority voting over all 3 annotators We have identified important proxies for the relevance of a sentence in a sentiment summary by means of: –The information gain metric –Feature selection focused on subsets of features with low inter-correlation, but high correlation with relevance We have used 10-fold cross-validation in our analyses WISE

Experimental Results (1) WISE

Experimental Results (2) WISE

Conclusions Relevant text segments for sentiment summaries are: –Segments discussing (aspects of) the text’s subject –Long text segments –Segments containing opinions –Segments containing arguments supporting these opinions It is not so much the absolute position of text segments, but rather the role that sentiment-carrying text segments play that renders them useful in summaries reflecting a text’s sentiment WISE

Future Work Validate our findings in other corpora and languages Investigate how to account for structural features (e.g., argumentation structures) of content in sentiment summarization Investigate the link between the sentiment of relevant text segments and the sentiment of a text as a whole Optimize the combination of relevant sentences in a final sentiment summary Implement our findings in an automated sentiment summarization tool WISE

Questions? Alexander Hogenboom Erasmus School of Economics Erasmus University Rotterdam P.O. Box 1738, NL-3000 DR Rotterdam, the Netherlands WISE