A recommender system is a type of information filtering system that predicts the preferences or interests of users for items such as movies, music, books, or products, and provides personalised recommendations accordingly. It analyses past interactions, user behaviour, and item attributes to suggest items that users are likely to find appealing or relevant. Aspect-Based Sentiment Analysis (ABSA) is a subsection of sentiment analysis aimed at extracting the sentiments regarding specific aspects within a document rather than overall sentiment of said document. For a recommender system to recommend better, it needs as much information about each user and item (restaurants in this case) as possible. By extracting aspect-wise sentiments from reviews through ABSA models, the goal is to enhance recommender systems using that additional information. The recommender system is implemented using a regression-based neural network, which takes in user and business IDs as input, embeds them as a vector, and outputs a rating value for each aspect. The rating values (obtained through ABSA) are used to train the neural network for each aspect and then fed into another neural network, which predicts the final rating, which is used to rank the items for each user. Thus utilising state-of-the-art models for ABSA as well as the recommender system, this paper aims to produce a recommender system that can generate better and more personalised recommendations. By combining these two approaches, the study leverages ABSA’s ability to capture fine-grained sentiment insights at the aspect level with the predictive power of neural collaborative filtering in modeling user preferences. RMSE and MAE are used as evaluation metrics as the output comes from a regression model. While the proposed model does not outperform individual baseline models in terms of predictive accuracy, it offers valuable insights into integrating ABSA with a regression model, handling of neutral sentiments and tackling imbalances in the dataset. This study highlights the potential of integrating these methodologies and paves the way for further refinement and exploration of hybrid approaches in recommender systems.