Recommendation systems have grown considerably thanks to the inclusion of contextual data to tailor suggestions to user preferences. Recommendation systems have advanced significantly due to the integration of contextual data, enabling them to tailor suggestions to individual user preferences. However, the emotional dimension of user interactions remains largely underexplored. This article proposes a new conceptual framework that combines sentiment analysis with context-aware recommendation systems. Inspired by the Improved Word Vector for Sentiment Analysis (IWVS) method, the proposed approach integrates improved word embeddings for sentiment extraction. It maps them to a feature space enriched with contextual information, such as time, location and user behavior. This work presents a theoretical basis and methodological framework for integrating emotional intelligence into recommender systems. This work presents a theoretical foundation and methodological framework for integrating emotional intelligence into recommender systems. The system in question aims to utilize XGBoost to classify emotional states and incorporate this information into the recommendation process. This proof of concept is intended to illustrate future developments in recommender systems.