Emotion detection from textual has become necessary in fields like mental health care, social media monitoring, and customer support. Recent developments in deep learning (DL) have delegated advanced approaches for recognizing emotions, yet challenges remain due to the subtle distinctions of languages, such as cultural variability and context. This paper comprehensively reviews progress in textual-based emotion detection, aiming at DL methods, commonly used datasets, and data pre-processing techniques. Attending to interdisciplinary proposals, data processing transparency, and multilingual abilities, this review pursues to answer three primary queries: the unification of interdisciplinary perception, the efficacy of DL models across languages, and the simplicity of pre-processing techniques used. Disseminating a structured overview of dominant techniques by packaging these findings in the relative tables, focusing on barriers to progress. This study provides a structure that shows the demand for dynamic and precise models to control language complexity and future explorations in the emotional intelligence.