—or— how to get 1,500 personality tests in a week...mbti on twitter personality traits on...

1
MBTI on Twitter Personality Traits on Twitter —Or— How to Get 1,500 Personality Tests in a Week Barbara Plank and Dirk Hovy University of Copenhagen, Denmark [email protected],[email protected] INTJ INFP INFJ ENFP INTP ISFJ ENTP ISFP ISTJ ENTJ ENFJ ESTP ESTJ ESFJ ESFP ISTP 0% 3% 6% 9% 12% 15% 18% corpus expected Contributions Corpus collection Statistical Analysis http://www.capt.org/mbti-assessment/ Introduction & Motivation Results •most work: small samples, closed vocabularies •here: large-scale, open vocabulary approach to personality prediction How many personality tests can we get in a week? manually checked 1,500 users annotated with MBTI and gender >100 tweets/user, in total 1.2m tweets Twitter API: “Briggs” + one of 16 MBTI Twitter corpus E vs I N vs S F vs T J vs P 0 25 50 75 100 E vs I N vs S F vs T J vs P 0 25 50 75 100 0 250 500 750 1000 Female Male 63% 37% Twitter corpus General US population using social media data for personality prediction analyze predictive features for various dimensions novel corpus of 1.2m tweets / 1,500 authors with Myers-Briggs type indicators (MBTI) & gender Myers-Briggs raw gender-controlled accuracy

Upload: others

Post on 07-Apr-2020

6 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: —Or— How to Get 1,500 Personality Tests in a Week...MBTI on Twitter Personality Traits on Twitter —Or— How to Get 1,500 Personality Tests in a Week Barbara Plank and Dirk Hovy

MBTI on Twitter

Personality Traits on Twitter —Or—

How to Get 1,500 Personality Tests in a WeekBarbara Plank and Dirk Hovy

University of Copenhagen, Denmark [email protected],[email protected] !

INTJINFPINFJENFPINTPISFJENTPISFPISTJENTJENFJESTPESTJESFJESFPISTP

0% 3% 6% 9% 12% 15% 18%

corpus expected

Contributions

Corpus collection

Statistical Analysis

http://www.capt.org/mbti-assessment/

Introduction & Motivation

Results

•most work: small samples, closed vocabularies •here: large-scale, open vocabulary approach to

personality prediction How many personality tests can we get in a week?

‣manually checked 1,500 users annotated withMBTI and gender

‣>100 tweets/user, in total 1.2m tweets

‣Twitter API: “Briggs” + one of 16 MBTI

Twitter corpus

E vs I

N vs S

F vs T

J vs P

0 25 50 75 100

E vs I

N vs S

F vs T

J vs P

0 25 50 75 100

0

250

500

750

1000

Female Male

63% 37%

Twitter corpus General US population

‣using social media data for personality prediction

‣analyze predictive features for various dimensions

‣novel corpus of 1.2m tweets / 1,500 authors with Myers-Briggs type indicators (MBTI) & gender

Myers-Briggs

raw gender-controlled

accu

racy