The Future of Agriculture: Analysing User Sentiment on Smart Farming with Explainable Artificial Intelligence

Smart farming is an innovative new approach to traditional agricultural practices that leverages disruptive technologies (DTs) and information and communication technologies (LCT's) to improve efficiency, lower costs, and reduce the wastage of crops and resources. A significant challenge to the widespread implementation of smart farming projects is the lack of knowledge and perceived disadvantages. In this study, sentiment analysis has been performed on YouTube comments to understand user sentiment towards new smart farming technologies. Three text representation techniques, count vectorizer, term frequency-inverse document frequency (TF -IDF), and fastText embeddings have been used on a smart farming corpus to analyse user sentiments. Different parametric and non-parametric machine learning algorithms have been used as classifiers on these feature vectors. The results suggest that TF-IDF of unigrams give the best macro-fl score of 0.6616 using a support vector machine-radial basis function (SVM-R) classifier. Visualisations have also been generated using Shapley Additive explanations (SHAP) to provide insight into predictions.