International Journal of Scientific Research and Engineering Development

Myers Briggs Type Indicator (MBTI) is a person- ality classification system consisting of 16 types, created on the basis of four categories: Introversion–Extraversion (I/E), Sensing–Intuition (S/N), Thinking–Feeling (T/F), and Judging– Perceiving (J/P). Generally, the Myers Briggs Type Indicator test is done using questionnaires that require the user to play an active part with the potential for bias and scalability problems. However, with the increasing number of user-generated content on social networking sites, blogs, etc., text has become a novel passive source for behavioral patterns that can be tapped for personality styles [1], [2], [8]. The proposed work introduces a hybrid deep learning ap- proach which infers the 16 types of MBTI personalities from text directly by leveraging context embedding via a pre-trained transformer model and a Bi-LSTM network for modeling se- quential deep personality features. The approach employs a pre- trained model like BERT as a starting point to acquire context embeddings of sentences/documents which represent context- related features about user posts at a discourse level. The context embeddings obtained in this fashion are subsequently used as inputs to a Bi-LSTM network which identifies personality- related sequential features at a superstructural level. The final model predicts either four MBTI personality axes as four binary classification tasks or predicts directly the MBTI personality type as a 16-class classification task. The experiment on the MBTI text data mined from online communities has indicated that the transformer + Bi-LSTM architecture described above is more accurate than traditional machine learning methods using TF-IDF and shallow models in terms of both accuracy and F1-score. The analysis also indicates that individual models of either contextual embeddings or sequen- tial information are less accurate than models incorporating both factors. This has indicated that the combination of both is more informative than either of them alone for personality predictions.

Chunduri Raghavendra, G.Sai Krishna, Ch.Vasanth Eswar, K.Anjireddy, B.Vamsi, B.Chanukya Babu,"Personality Trait Prediction from Text Using Transformer Embeddings and Deep Sequence Modeling Techniques" International Journal of Scientific Research and Engineering Development, V9(2): Page(834-839) Mar-Apr 2026. ISSN: 2581-7175. www.ijsred.com. Published by Scientific and Academic Research Publishing.