Master thesis defence by Kei Man Cheng


Predicting MBTI Personality Types with Twitter Using Doc2vec


With the emergence of various social media platforms like Facebook, Twitter, and LinkedIn, it has become increasingly convenient for people to express and share their thoughts, ideas, and emotions through text based means on online platforms. These text based social media entries can thus then be leveraged as a data that can be modeled to gain insights into individual personas and personalities. Utilizing this data has allowed innovation leaders to create better recommendation systems, provide tailored user experiences, find compatible candidates for hiring positions, and increase job satisfaction in the workplace among other things. This work will explore prior research in Natural Language Processing (NLP) and its applications in computational social science, specifically related to the idea of modeling and classifying personality from text. Previous personality modeling studies like Facebook’s myPersonality Project (2007) will be discussed and this work will present a leaner methodological framework which produces comparable results. This work will, furthermore, explore the application of Doc2Vec for personality modeling, considering its successful application in other language modeling tasks in computational social science.


Dirk Hovy

External Examiner

Ira Assent, Aarhus