Personality
Download as PPTX, PDF
1 like215 views
Personality Recognition" includes automatic classification of authors' personality traits, that can be compared against gold standard annotation obtained by means of the big5 personality test
Personality
- 1. Personality Recognition By: Dhwanit Gupta(201001118) Arpit Sharma(201101020) Y.Sindhusha(201305518) Charudatt Pachorkar(201102071) Mentor: Santosh K
- 2. PROBLEM STATEMENT Personality Recognition includes automatic classification of authors personality traits, that can be compared against gold standard annotation obtained by means of the big5 personality test.
- 3. CONTENTS Introduction Dataset Approach and Architecture Evaluation and Results Conclusion
- 4. INTRODUCTION Why personality recognition? Recommender systems Personalized Advertising Opinion Marketing Deception Detection Social Network Analysis
- 5. INTRODUCTION Mapping personality of person to big-5 personality traits which includes: Extraversion (sociable vs shy) Neuroticism (neurotic vs calm) Agreeableness - (friendly vs uncooperative) Conscientiousness - (organized vs careless) Openness - (insightful vs unimaginative)
- 6. DATA SET Facebook dataset of 250 users of about 10000 status. Essay dataset of about 2400 essays
- 7. APPROACHES Approach-1 Feature based approach Approach-2 Trigram approach
- 8. FEATURE BASED APPROACH Feature Extraction Feature Vector Representation Feature Vector Dimension Reduction Classification(Bayesian)
- 9. FEATURE EXTRACTION Style based features Sentimental Analysis Total number of posts of author Concept Extraction Social networking features
- 11. Extroverts tend to use Dictionary words 2nd person,3rd person singular Past tense verbs Neurotic users tend to Update their status with anger words and Less likely to use social interaction words
- 12. Feature Vector representation
- 13. FEATURE VECTOR DIMENSION REDUCTION Why? And How? using Correlation Coefficient Clustering
- 14. APPROACH-2 Trigram based approach It is based on generating two features for each status say F1 and F2 Where F1 represents normalized frequency of trigrams w.r.t to current personality trait And F2 represents normalized frequency of trigrams w.r.t remaining classes Finally train the individual classifier using SVM for feature vector (F1,F2)
- 15. EVALUATION Classifiers used SVM Bayesian With Dataset division as: 70% - training and 30% - testing
- 16. EVALUATION MEASURES Precision Recall F-Score
- 17. RESULTS Personality Trait Accurac y True Positiv e Rate(T P) Recall False Positive Rate(FP ) True Negative Rate(TN) False Negativ e Rate(FN ) Precisi on F- score Trigram Accurac y Extroversion 74% 0.28 0 1 0.72 1 0.44 41.17% Openness 70% 1 1 0 0 0.695 0.82 70.58% Neuroticism 62% 0.577 0.348 0.652 0.407 0.652 0.613 43.13% Agreeableness 60% 0.833 0.64 0.36 0.166 0.5555 0.667 58.82% Conscientiousne ss 56% 1 0.9166 0.0833 0 0.532 0.695 50.98%
- 18. CONCLUSION Results shows that style based features gives better results over trigram approach.
- 19. REFERENCES http://clic.cimec.unitn.it/fabio/wcpr13/verhoeven_wcpr13.pdf http://clic.cimec.unitn.it/fabio/wcpr13/celliwcpr13.pdf http://clic.cimec.unitn.it/fabio/wcpr13/farnadi_wcpr13.pdf http://clic.cimec.unitn.it/fabio/wcpr13/tomlinson_wcpr13.pdf http://clic.cimec.unitn.it/fabio/wcpr13/markovikj_wcpr13.pdf6 http://clic.cimec.unitn.it/fabio/wcpr13/alam http://clic.cimec.unitn.it/fabio/wcpr13/mohammad_wcpr13.pdf http://clic.cimec.unitn.it/fabio/wcpr13/appling_wcpr13.pdf
- 20. THANK YOU