Personality recognition from user-generated text has emerged as a prominent research focus in natural language processing. Although there have been attempt in the literature to classify personality traits from text using random word embedding with a single-size filter, these methods suffer from low accuracy, limited iterations, and information loss. Deep learning-based methods for personality trait recognition have risen in popularity, utilizing word embedding techniques to convert words and sentences into vector representations. Despite their importance, the impact of different word-embedding techniques on personality recognition have not been explored. In this paper, a multi-size filters Convolutional Neural Network (CNN) and Long Short Term Memory (LSTM) network along with different word representation techniques is proposed to get promising results for the personality traits classification. Two word-embedding techniques: Word to Vector (Word2Vec) and Global Vector (GloVe) are distinctly applied to observe the impact of the word embedding result on the proposed model. The performance of the proposed model is tested on personality-related benchmark dataset Myers-Briggs Type Indicator (MBTI) considering accuracy measure. The experiments show that the proposed model with GloVe representation achieved higher accuracies, with a 7.24% improvement compared to Word2Vec representation across four personality traits. The proposed model with GloVe on the MBTI dataset outperformed the other models with an approximately 5% higher average accuracy.