Paper Title
Emotion Recognition Using Text with Emojis and Speech
Abstract
Emotion Recognition can enhance the competencies of social media platforms and other tech companies, as the
need for grasping the user's attention towards them has increased tremendously. This paper concentrates on recognising user
emotion through text-only, text with emojis, and speech. A hybrid model is designed with the combination of RoBERTa-
LSTM-CNN models, which outperforms the individual models. Speech Emotion Recognition involves extracting the
features of an audio file irrespective of the semantic contents. MFCC-CNN is used to analyse the audio features based on
various augmented audio segments, and speech-to-text transcription is used to increase the accuracy of the prediction. Based
on the detected emotion, content is recommended through music and videos to enhance the user's mood. Feedback taken
from users is stored in the database to improve future predictions.
Keywords - MFCC, CNN, Bi-LSTM, BERT, RoBERTa, Word Embedding.