Opinion Mining of Customer Feedback Data on the Web Presented By Dongjoo Lee, Intelligent Databases...

Opinion Mining of Opinion Mining of Customer Feedback Data on the Customer Feedback Data on the WebWeb

Presented By Dongjoo Lee,

Intelligent Databases Systems Lab.

1

Copyright 2008 by CEBT

Customer Feedback DataCustomer Feedback Data

2

This camera is my first digital one and was super easy to learn to use. The picture looks great and it’s simple to get the correct exposure. The memory card that comes with the camera has a very small capacity though, (it holds about 4 photos) so a separate memory card is a necessity. I’m not very happy with the memory card.”

This camera is my first digital one and was super easy to learn to use. The picture looks great and it’s simple to get the correct exposure. The memory card that comes with the camera has a very small capacity though, (it holds about 4 photos) so a separate memory card is a necessity. I’m not very happy with the memory card.”

Marketing planner

Potential customer


Goal of OMGoal of OM

3

From reliable amounts of feedback data

Automatic or semi-automatic

Effective presentation

Serves the chosen objectives

pos neg

123 20

lens

exposure

………

memory ………

Ratio of pos and neg opinions

Opinion summary

…

…


Tasks for Opinion MiningTasks for Opinion Mining

4


Development of Linguistic Resource Development of Linguistic Resource (1)(1)

Linguistic resources

Used to extract opinion and to classify the sentiment of text

Appraisal theory

A framework of linguistic resources which describes how writers and speakers express inter-subjective and ideological position

Sentiment related properties

Subjectivity, Orientation, Strength

5

vertical yellow liquid

good excellentbadterrible

positivenegative

subjectivesubjective

objectiveobjective

more intensive


Development of Linguistic Resource Development of Linguistic Resource (2)(2)

Conjunction method Hatzivassiloglou and McKeown (1997)

– adjectives in ‘and’ conjunctions usually have similar orientation, while ‘but’ is used with opposite orientation.

PMI (Pointwise Mutual Information) method Turney and Littman (2003), Baroni and Vegnaduzzo (2004)

– terms with similar orientation tend to co-occur in documents

– subjective adjectives tend to occur in the near of other subjective adjectives

WordNet Exploring method Hu et al. (2004)

– adjectives usually share the same orientation as their synonyms and opposite orientation as their antonyms

Gloss Classification method Esuli et al. (2005, 2006)

– terms with similar orientation have similar glosses

– terms without orientation have non-oriented glosses

SentiWordNet

6

andbut

positivenegative

corpuscorpus

seed termsseed terms


Sentiment ClassificationSentiment Classification

The process of identifying the sentiment – or polarity – of a piece of text or a document.

PMI method Turney et al. (2002)

– SO(phrase) = PMI(phrase, “excellent”) – PMI(phrase, “poor”)

Machine Learning method A special case of text categorization with sentiment- rather than topic-based

categories

Pang and Lee (2002)

– Default Classifier - Naïve Bayes, MaxEnt, SVM, PrTFIDF

Pang and Lee (2004)

– Use only subjective parts

NLP Combined method Whitelaw et al. (2005)

– Applied the appraisal theory

Wilson et al. (2005)

– Employ machine learning and 28 linguistic features

7


Systems for Opinion SummarizationSystems for Opinion Summarization

8

SystemSentiment Resource

Syntactic

Analysis

Extracting Opinion Expression

PresentationFeatureExtraction

SentimentAssignment

ReviewSeer(2003)

Thumbs up/down

No

Probabilistic modelNaïve Bayes Classifier

List sentences contain the feature term

RedOpal (2007)

Star rating Frequent noun and noun phrase Average star rating Order products by score of each feature

CBA minerInfrequent feature selection

WordNet exploringDominant polarity of each phrase

Bar graphOpinion

Observer (2004)

LinguisticResource

Kanayama’sSystem (2004)

Yes

Sentiment unitModifying the machine translation framework

N/A

bBNP heuristicSentiment lexiconSentiment pattern database

List sentences which bear sentiment of a product

WebFountain (2005)

OPINE (2005)

Web PMI Relaxation labeling N/A

higher precision and lower recall


DiscussionDiscussion

Provide an overall picture of the tasks and techniques for opinion mining system.

Focused on surveying and analyzing the methods for development of linguistic resources, sentiment classification, and opinion summarization.

Opinion mining has become important for all types of organizations, including for-profit corporations, government agencies, educational institutions, non-profit organizations, and the military in gauging the opinions, likes and dislikes, and the intensity of the likes and dislikes, of the products, services, and policies they offer and plan to offer.

An understanding of the overall picture of the tasks and techniques involved in opinion mining is of significant importance.

9

Thank You.Thank You.

[email protected]

Opinion Mining of Customer Feedback Data on the Web Presented By Dongjoo Lee, Intelligent Databases...

Documents

Transcript of Opinion Mining of Customer Feedback Data on the Web Presented By Dongjoo Lee, Intelligent Databases...