Automated Personality Classification - PowerPoint Presentation

Slide1

Automated Personality Classification

A. KARTELJ and V. FILIPOVIC

School of Mathematics, University of Belgrade, Serbia

and

V. MILUTINOVIC

School of Electrical Engineering, University of Belgrade, SerbiaSlide2

Agenda

Problem overview

Classification of the existing solutions

Presentation of the existing solutionsComparison of the solutionsWork in progress:Bayesian Structure Learning for the APCFuture work: Video Based APCConclusions

MULTI 2012

2

3.10.2012Slide3

Problem Overview

MULTI 2012

3

3.10.2012Slide4

The Big 5 Model

MULTI 2012

4

3.10.2012Slide5

The Steps in Our Research

Survey paper

(under review at ACM CSUR)

Research paper:A new APC model based on Bayesian structure learning (in progress)Real-purpose applicationof the APC model from step 2Go to step 3 MULTI 2012

5

3.10.2012Slide6

Elements of APCCorpus:

Essay, weblog, email, news group,

Twitter counts...

Personality measurement:Questionnaire (internet and written). We are searching for an alternative!Model:Stylistic analysis, linguistic features, machine learning techniquesMULTI 201263.10.2012Slide7

Applications

MULTI 2012

7

3.10.2012Slide8

Mining People’s Characteristics

MULTI 2012

8

3.10.2012Slide9

Classification of Solutions

MULTI 2012

9

3.10.2012

C1 criterion separates solutions by type of conversation (1 = self-reflexive, N = continuous)

C2 criterion separates solutions by approach (TD = top-down, DD = data-driven, or HY = hybrid) Slide10

Linguistic Styles: Language Use as an Individual Difference

Pennebaker

and King [1999]MULTI 2012103.10.2012Slide11

LIWC and MRC Features

Feature

Type

ExampleAnger wordsLIWCHate, killMetaphysical issuesLIWCGod, heaven, coffin

Physical state / function

LIWC

Ache, breast, sleep

Inclusive

words

LIWC

With, and, include

Social processes

LIWC

Talk, us, friend

Family members

LIWC

Mom, brother, cousin

Past

tense verbs

LIWC

Walked, were, had

References

to friends

LIWC

Pal, buddy, coworker

Imagery of

words

MRC

Low:

future, peace – High: table, car

Syllables per word

MRC

Low:

a – High: uncompromisingly

Concreteness

MRC

Low: patience, candor

– High: ship

Frequency of use

MRC

Low: duly, nudity –

High: he, the

MULTI 2012

11

3.10.2012Slide12

What Are They Blogging About? Personality, Topic and Motivation in

Blogs

Gill

et al. [2009]MULTI 2012123.10.2012Slide13

Taking Care of the Linguistic Features of Extraversion

Gill

and

Oberlander [2002]MULTI 2012133.10.2012Slide14

Personality Based Latent Friendship Mining Wang et al. [2009]

MULTI 2012

14

3.10.2012Slide15

A Comparative Evaluation of Personality Estimation Algorithms for the TWIN

Recommender

System

Roshchina et al. [2011]MULTI 2012153.10.2012Slide16

Predicting Personality with Social MediaGolbeck

et al. [

2011]MULTI 2012163.10.2012Slide17

Our Twitter Profiles, Our Selves: Predicting Personality with TwitterQuercia et al.

[

2011

]MULTI 2012173.10.2012Slide18

Paper

Input

Corpus

Features Algorithm

Soft.

Cit.

I

S

A

R

[Pennebaker and King 1999]

text

essays

LIWC

correlations

n/a

455

H

H

H

M

[Mairesse et al. 2007]

text,

speech

essays

LIWC,

MRC

C4.5,

NB

,

SMO, M5’

Weka

99

M

M

H

M

[Gill et al. 2009]

text

weblogs

(14.8words)

LIWC

linear

regression

n/a

26

H

H

M

M

[Yarkoni 2010]

text

w

eblogs

(100K words)

LIWC

correlations

n/a

21

H

M

M

M

[Gill and Oberlander 2002]

text

emails

(105

students)

bigrams

bigram

analysis

n/a

49

L

M

M

L

[Nowson et al. 2005]

text

weblogs

(410K

words)

word list correlations n/a 48LHHL[Oberlander 2006]text weblogs (410K words)N-grams NB, SMO Weka 53HMHM[Wang et al. 2009]text, weblogs (200 pairs)lexical freq. ,TFIDFlogistic regressionMinitab 1HMMM[Iacobelli et al. 2011]text weblogs (3000)LIWC, bigrams,SVM, SMO, NB..Weka 1HHMH[Argamon et al. 2005]text essaysword list, conj.SMO Weka 38HMMM[Argamon et al. 2007]text essaysword list, conj.SMO Weka, ATMan45HMMM[Mairesse and Walker 2006]text , conv. extracts96 persons (≈100Kwords)LIWC, MRC, utterance…RankBoost n/a 22MMHM[Rigby and Hassan 2007]text mail. lists (140K emails)LIWC C4.5 Weka, SPSS30MHML[Roshchina et al. 2011]text TripAdvisor reviews LIWC, MRCLinear, M5, SVMWeka 2HMLM[Quercia et al. 2011] meta 335 Twitter usersTwitter counts M5’ rules Weka 5MHMM[Golbeck et al. 2011]text, meta279 FB users 5 classes (161 in total)M5’ rules, Gaussian processesWeka 12HMMM[Celli 2012] text 1065 posts22 ling. Featuresmajority-based classificationn/a 1MMMM

MULTI 2012

18

3.10.2012Slide19

Naive Bayes Classifier

MULTI 2012

19

3.10.2012Slide20

Naive Bayes and Bayesian Network

MULTI 2012

20

3.10.2012Slide21

Bayesian Network for the APC

MULTI 2012

21

3.10.2012Slide22

Bayesian Network Structure Learning

Obtain corpus (training set T)

Fit T to appropriate network structure by:

ILP formulation + solver (CPLEX, Gurobi…) on smaller instancesApply metaheuristic on larger instancesValidate quality of metaheuristic approachCompare obtained APC accuracy with other approachesMULTI 201222

3.10.2012Slide23

Other Ideas

MULTI 2012

23

Games with a purpose

(GWAP)

Clustering personality characteristics

3.10.2012Slide24

Packing everything together: Video Based APC

MULTI 2012

24

3.10.2012Slide25

ConclusionsClassification of the existing solutions (Survey paper)Filling the gaps inside classification tree

Introducing Bayesian Structure Learning for the APC

Utilizing metaheuristics in dealing

with high dimensionalityAPC potential: social networks, recommender, and expert systemsMULTI 2012253.10.2012Slide26

THANK YOU!

Aleksandar

Kartelj kartelj@matf.bg.ac.rsVladimir Filipovic vladaf@matf.bg.ac.rsVeljko Milutinovic vm@etf.bg.ac.rs