Optimizing Agricultural Insights: Semantic Clustering and Topic Modelling for Farmer Queries
Lalith Abhiram Dasari, Jetty Sowmith, Mamidi Naveen Krishna, Ch. Saketh, Manju Venugopalan · 2025
The processing of complex semantic queries stands as a critical factor for agricultural support systems because farmers frequently submit such queries. The proposed solution utilizes Bidirectional Encoder Representations from Transformers (BERT) for innovative clustering and classification of complex queries. The queries underwent a transformation process where they received high-dimensional status to represent semantic content before undergoing clustering by algorithms. The model received assessments through Silhouette Score and Davies-Bouldin Index and Calinski-Harabasz Index to determine cluster unity. The system received an improvement through the implementation of Latent Dirichlet Allocation (LDA) for topic modeling inside clusters. The unified system enhances query groupings and classification and information retrieval functionalities which results in precise agricultural advisory recommendations. The K-means Clustering method with Agricultural-BERT produced the optimal results which delivered Silhouette Score of 0.1231 along with Davies-Bouldin Index of 2.7252 and Calinski Harabasz Index of 891.0811.