Enhancing Film Discovery through Social Media Traces Using Sentiment Analysis and Machine Learning Combined Approach

Films have historically played a significant role in cultural expression and entertainment, influencing global people and affecting society on various levels. In the contemporary era of digital technology, characterized by vast amounts of data, the accessibility and discovery of films have become crucial. Nevertheless, efficiently searching for relevant movies within this vast data landscape poses a significant challenge. Current search systems primarily rely on textual and visual analysis and often fail to incorporate users' social interactions as social media traces valuable on various social platforms. To address this challenge, we introduce Social media-Enhanced Ranking Framework (SERF) that aims to enhance the research process and gain a deeper understanding of users' sentiments and preferences. Our research aligns with the United Nations Sustainable Development Goal 9, which emphasizes innovation in industry, technology and infrastructure. The objective is to redefine how we discover films combining between traditional information retrieval methods and the dynamic feedback loop offered by social media. In the proposed approach, film recommendations are determined through a combination of textual scores and a social search model. This last relies on the frequency of generated social media traces and sentiment analysis conducted by machine learning algorithms. Our empirical experiments reveal the pivotal impact of the exploitation of the reactive and the descriptive social traces on the repositioning of cinematic materials into the final re-ranked list. Additionally, our findings highlight the superior performance of the deep RNN-LSTM model with embedding, outperforming traditional machine learning algorithms with an AUC score of 89%, affirming the effectiveness of deep learning techniques in enhancing predictive accuracy of relevant documents.