Rice is one of the most important cereal crops which supports food security to billions of people across the world. However, it is vulnerable to several insect-pest infestations which significantly damage the crop and reduce the overall yield. Traditional pest management practices aim to minimize damage of crops relying on manual field inspection, which are labour-intensive, time consuming, and prone to human error. With recent advances in deep learning, automated detection frameworks like YOLO networks, have emerged as promising tools for real-time insect-pest detection for crops. In this study, we proposed an improved lightweight network named YOLODWSimAM, built upon the baseline YOLOv11 architecture. In the proposed network, a simple parameter-free attention module (SimAM) and a depthwise convolution (DWConv) are integrated into a lightweight unified block, DW-SimAM, which enhances the network's feature extraction capability with reduced trainable parameters. An insect-pest dataset, consisting of 1721 images, was acquired from different rice fields and annotated with bounding boxes to locate the instances of insect-pests within the images. With only 1.93 million trainable parameters and 4.9 GFLOPs, the proposed YOLO-DWSimAM network achieved an mAP@50 of 0.851 on the test dataset, outperforming the baseline YOLOv11 architecture by 1.19%. Additionally, the proposed YOLODWSimAM network demonstrated strong performance on the edge device, especially Raspberry Pi 4, for rice insect-pest detection, achieving average inference speed of 0.67 s and FPS of 1.49 using input images of size 640 & times; 640. These findings validated that the proposed network achieved higher detection accuracy while being computationally efficient, making it suitable for real-time insect-pest management in resource-constrained agricultural environment.

Introduction Long non-coding RNAs (lncRNAs) are a major class of non-coding RNAs (ncRNAs) longer than 200 nucleotides. They play key roles in plant embryogenesis, root development, reproduction, and gene silencing. Accurate identification of lncRNAs in crop species is crucial for understanding their biological functions. However, most existing computational tools are designed for human and animal lncRNAs, limiting their applicability to crop genomes. Therefore, this study aims to develop a crop-specific computational tool for the accurate classification of lncRNAs and coding RNAs in crop species, addressing the limitations of current approaches.Methods An XGBoost classifier was trained to distinguish lncRNA and coding RNA (cRNA) sequences using sequence-intrinsic features derived from five crop species: wheat, sorghum, rice, soybean, and maize. Model performance was evaluated against benchmark tools, CPC2 and PLEKv2Results The trained XGBoost classifier achieved an accuracy of 95.30%, precision of 93.90%, recall of 98.40%, F1-score of 96.10%, and an area under the ROC curve (AUC-ROC) of 99.40%, outperforming existing tools. These results demonstrate the model's reliability in distinguishing lncRNAs from coding RNAs.Discussion The trained XGBoost classifier was deployed as CLnc-Pred, a web-based application that allows users to input or upload FASTA sequences for lncRNA prediction. This framework enables efficient and accurate identification of lncRNAs in crop species.Conclusion CLnc-Pred enhances accessibility and accuracy in crop lncRNA research and supports downstream functional and regulatory analyses. Future work will focus on expanding datasets, incorporating additional plant species, and extending the framework to support multi-class classification of diverse ncRNA types.

Accurate detection of wheat spikes and reliable yield prediction are critical for optimizing crop production and resource management. This study presents an integrated framework for spike detection and yield estimation using pseudo-RGB images derived from hyperspectral data. A YOLOv8 model was trained on 1,050 images, achieving high precision, recall, and mean average precision values. The bounding boxes and masks generated byYOLOv8 were used to quantify spike count and spike area, while six vegetation indices were extracted from hyperspectral images acquired at the booting stage. Three multiple linear regression models were developed for yield prediction: one based on spike features, another on vegetation indices, and a third combining both. The combined model achieved the highest accuracy, with a five-fold cross-validation R2of 0.902 +/- 0.007, RMSE of 1.739 +/- 0.133 g,and MAE of 1.289 +/- 0.066 g. Compared with previous approaches, the proposed framework demonstrated improved performance, highlighting the value of integrating spike morphology and spectral data for yield prediction. Overall, the study shows that hyperspectral imaging can simultaneously provide morphological and physiological traits, reducing reliance on high-resolution RGB data in wheat phenotyping

The accurate identification of crop types and estimation of crop acreage are crucial for land management, food security, and policy planning. This study explores methodologies for crop area estimation using machine learning algorithms applied to remote sensing imagery, with Hisar, Haryana, as a case study. It compares the proposed remote sensing-based methods for the years 2018-19 to 2022-23 with traditional statistical approaches using historical datasets for the years 1998-2021 for estimating mustard and wheat acreage. The enhanced spatial and temporal variations in the fused dataset improve crop identification and help overcome the limitations of traditional methods. Our proposed methodology integrates unsupervised clustering, remapping with a supervised classifier using ground truth data, and validation. The results indicate that the proposed approach outperforms traditional methods in accuracy and effectiveness. Based on experimental evidence, the study recommends using a fusion dataset, spatially diverse ground truth data for a given year, and a hybrid machine learning algorithm to improve crop acreage estimation for years beyond the current period. The proposed hybrid KRF method applied to the fusion dataset achieved superior accuracy compared to the RF classifier and statistical models, with the lowest MAPE of 2.55

Reliable crop yield estimates at fine spatial resolution are essential for precision agriculture, food security planning, and insurance schemes. However, yield statistics are reported at coarser administrative levels, limiting their applicability for field-scale analysis. This study proposes a multi-stage hybridized framework that integrates deep learning (DL) models with geostatistical residual kriging to disaggregate village-level crop yield statistics to the pixel level. The proposed methodology is demonstrated using wheat and mustard crops as case study in the semi-arid districts, Haryana, India. The study identifies suitable data combination by evaluating multiple combinations of soil, weather, Sentinel-1, and Sentinel-2 bands data for yield disaggregation. Results show that datasets combining spectral and weather information consistently outperform other data combinations. Validation results showed that the strongest numerical accuracy was observed for machine learning algorithms, e.g., random forest, with an R2 of 0.9949, but it lacks spatial realism. On the other hand, DL models had comparable numerical accuracy and also produced smoother and more realistic spatial transitions but exhibited spatially structured residuals. To mitigate these spatial biases, residual kriging was applied to DL outputs, resulting in RMSE reduction of 35-45% and generating smoother pixel-level maps that preserved fine-scale heterogeneity and aligned with reported village yields. Moran's I analysis confirmed significant residual spatial autocorrelation for DL models, justifying the use of geostatistical correction. Thus, the proposed hybridized framework emerged as best for balancing statistical accuracy with spatially realistic yield disaggregation. This study provides one of the first empirical demonstrations of village-to-pixel yield disaggregation using the identified weather and satellite band data combination.

Accurate assessment of soil organic carbon (SOC) is crucial for sustainable agriculture and climate change mitigation; however, conventional laboratory methods remain expensive and time-consuming, particularly for smallholder farmers in developing countries. This study evaluated five machine learning models (Random Forest (RF), Artificial Neural Network (ANN), Support Vector Machine (SVM), Boosted Regression Tree (BRT), and eXtreme Gradient Boosting (XGBoost)) for SOC prediction across the Indo-Gangetic Plains using smartphone imagery and topographic parameters. Two hundred soil samples were collected from agricultural fields in Karnal and Kaithal districts of Haryana, IARI farm of Delhi, Varanasi district of Uttar Pradesh, and Nadia district of West Bengal, representing diverse agroecological conditions. Three modelling approaches were tested: Model-1 used smartphone images as sole predictors, showing poor performance with testing R2 values of 0.005-0.06 and normalized root mean square error (nRMSE) of 0.31-0.34. SVM achieved the best performance (R2 = 0.06, nRMSE = 0.31). Model-2 employed topographic parameters derived from digital elevation models, including slope, aspect, openness measures, terrain ruggedness index, and total curvature. This approach outperformed Model-1, with RF achieving R2 = 0.147 and nRMSE = 0.29. Model-3 integrated smartphone imagery with topographic features, producing exceptional results. XGBoost achieved optimal performance (R2 = 0.717, nRMSE = 0.16, Nash-Sutcliffe Efficiency = 0.710, concordance correlation coefficient = 0.82, d-index = 0.906), representing a significant improvement over individual approach. The integrated framework enables real-time soil assessment, eliminating the need for costly laboratory testing and enabling resource-limited farmers to make informed management decisions.

Real-time assessment and prediction of soil nutrients are essential for precision soil nutrient management. Digital mapping of soil nutrients, influenced by topographical attributes, offers a promising approach for site-specific nutrient management. Although, multispectral and hyperspectral remote sensing data have been utilized separately for the digital mapping of soil nutrients, the integration of both data types for soil nutrients especially for soil micronutrients remains unexplored. Therefore, this study attempted to enhance the accuracy and reliability of predicting four important Diethylene Triamine Pentaacetic Acid (DTPA) extractable micronutrient cations such as zinc (Zn), copper (Cu), iron (Fe), and manganese (Mn) in the Katol block of Nagpur district, Maharashtra, India by integrating multispectral data of Sentinel-2, and hyperspectral data of Airborne Visible Infrared Imaging Spectrometer-Next Generation (AVIRIS-NG) along with the Shuttle Radar Topographic Mission (SRTM) digital elevation model (DEM) data, soil and climatic data with three machine learning (ML) models. These four micronutrients were selected for digital mapping owing to their crucial roles in plant growth and development, and their continuous depletion under intensive cultivation of high-yielding varieties, which may adversely affect crop productivity and soil health in the study area. 132 georeferenced surface soil samples were collected during airborne hyperspectral spectra acquisition by the AVIRIS-NG. Micronutrients were analysed in laboratory and lab-generated data combined with multispectral, hyperspectral, soil and climatic data with the help of three ML models viz. Random Forest (RF), Quantile Random Forest (QRF) and Cubist for predicting and mapping of those four key micronutrients. Based on the prediction performances especially in the validation dataset, the RF model was found more promising in comparison to other models in micronutrient prediction. Since the RF model demonstrated superior performance, it was utilized for mapping those four micronutrients. This RF model demonstrated moderate to satisfactory accuracy, with the value of the coefficient of determination (R2) 0.71, 0.65, 0.62 and 0.63 for Zn, Cu, Fe, and Mn prediction respectively. The study utilized eight different types of datasets and applied them to each model and it was revealed that the combined use of multispectral, hyperspectral, soil and climatic datasets improved the prediction accuracy of micronutrients compared to using each dataset individually. The study also revealed that soil variable i.e., soil pH was the most influential variable for soil micronutrient especially for Zn, Cu and Fe predictions. Next to this, hyperspectral variables (PC2 for Zn, Fe, PC1, and PC3 for Mn, and PC6, and PC7 for Cu) were found crucial in predicting and mapping soil-available micronutrients. Furthermore, it was observed that after soil and hyperspectral variables, DEM-derived variables (valley depth, elevation etc.) were also found important in predicting micronutrients particularly for Mn. Prediction uncertainties, assessed through confidence and prediction intervals as well as standard deviation and standard error, were lower for Zn and Cu, but slightly higher for Fe and Mn, indicating areas needing additional soil sampling for more reliable estimation. Thus, this study underscores the enhanced predictability of micronutrients, achieved by integrating multispectral and hyperspectral remote sensing data with terrain, soil and climatic data, providing valuable digital maps of micronutrients for precision micronutrient management in the study area.

In recent years, measuring three-dimensional (3D) surface information has gained a great interest in plant phenotyping because it can represent the nature of plant architecture better than conventional 2D images. This paper presents an approach for processing 3D point clouds converted from 2D RGB images in the context of high-throughput plant phenotyping. High-resolution RGB multi-view imagery of a Chickpea plant was collected using a high-end camera. Based on these image sequences, 3D point cloud reconstruction of the canopy was conducted and analyzed. Later sophisticated 3D operations were performed on these images including 3D downsampling and after that clustering was performed on the processed point cloud. The information generated can help in the evaluation of crop traits and provide accurate statistics for the assessment of their growth parameters.

Object detection is a crucial aspect of computer vision used to address drought tolerance in the biological context. In this context, we address the crucial task of selection of drought-tolerant mustard genotypes (B. juncea) based on automatic and non-destructive deep learning YOLOv5-based silique count software MuSiC v1.0. Our approach utilizes YOLOv5, based on images acquired using a DSLR camera (SONY Alpha 7iii, 24.2MP), securely affixed to a tripod. The user-friendly desktop software named MuSiC v1.0, underpinned by the YOLOv5 model discriminates drought tolerant and susceptible mustard genotypes based on Multi-trait Genotypte Ideotype Distance Index (MGIDI) and the results are comparable with the manual count. This model was trained using an annotated dataset for each of the 30 genotypes, crafted with the assistance of the Roboflow annotator, comprising approximately 22,800 annotations. Drought effect on the silique count trait was comparable both in manual as well as MuSiC v1.0 software-based count method. Stress tolerance indices are derived using iPASTIC software for 30 genotypes and ultimately these traits are used for selecting drought tolerance based on MGIDI index. Superior drought-tolerant mustard genotypes were chosen with a 25

Drought is a major abiotic stress that adversely affects plant growth, physiology, and crop yield. Conventional methods for assessing drought stress tend to be fragmented, targeting either leaves, canopies, or roots, and are often expensive, low-throughput, and lack the ability to provide real-time, whole-plant insights. Addressing these limitations, this study presents a novel, integrated pipeline titled Intelligent Decision Support for Drought Stress (IDSDS) that leverages remote sensing and artificial intelligence (AI) for accurate, real-time monitoring of drought stress across entire plants. The IDSDS pipeline employs low-cost RGB images collected at various growth stages and uses a deep learning-based model to reconstruct hyperspectral data, which is typically costly and complex to obtain. This reconstructed data enables the extraction of key physiological traits, including greenness, saturation, and pigment content. A novel phenotyping metric-Greenness Coefficient (GC), was also proposed, offering precise spatial analysis of drought impact within the plant. The hyperspectral reconstruction model was validated using standard performance metrics such as the correlation coefficient, mean squared error, standard deviation of squared error, and spectral angle mapper (SAM). IDSDS further calculates a comprehensive set of spectral indices (e.g., greenness, leaf pigment, water content) that are closely linked to drought-induced changes. Finally, by integrating these indices with machine learning-based classification models, IDSDS accurately stratifies plant drought stress into seven distinct categories. The results showed that the proposed hyperspectral reconstruction model effectively converts RGB plant images into accurate hyperspectral data, achieving a SAM value between 0.14 and 0.30. This indicates strong spectral similarity, meaning the reconstructed pixel spectra closely align with the reference spectra. The GC, along with other reconstructed spectral indices, supports visual interpretation and enhances the traceability of the system's outputs, thereby increasing transparency. Additionally, the findings demonstrate statistically significant results (p < 0.001) for these indices in detecting plant drought stress, with a high classification accuracy of 99 % and an average area under the curve (AUC) of 1.00, reflecting precise differentiation of stress across the entire plant. Overall, the study introduces a breakthrough in drought stress monitoring, combining high-throughput and cost-effective RGB imaging with AIto support both scientific research and practical crop management. The IDSDS pipeline lays the groundwork for informed, drought-adaptive decision-making of agricultural crops.

The evaluation of seed viability is pivotal in agriculture, biodiversity conservation, and ecological research. Traditional methods used for testing the seed viability are often destructive and pose challenges regarding labour intensity and seed wastage. The study was carried out during 2022–23 at ICAR-Indian Agricultural Statistics Research Institute, New Delhi with the aim of collecting the seed genotypes and NIR spectroscopic instrument and computational approaches and appropriate hardware and software resources. A diverse dataset of NIR spectral data from various seed species was used and analysed using three sophisticated ML models, namely Linear Discriminant Analysis (LDA), Random Forest (RF), and Artificial Neural Networks (ANN). The performance of the developed models was evaluated based on accuracy, precision, recall, and F1 score metrics. Furthermore, the experimental results demonstrated that NIR spectroscopy and ML could effectively classify viable seed. The integration of artificial neural networks (ANNs) has demonstrated significant potential in capturing intricate patterns within spectral data, achieving an approximate accuracy of 95%. This highlights their effectiveness in precise classification tasks. Additionally, machine learning (ML)-based approaches have shown promise in conserving valuable seed resources by offering scalable solutions adaptable to large-scale agricultural and conservation applications. To enhance model transparency, Local Interpretable Model-Agnostic Explanations (LIME) has been employed, providing deeper insights into the ANN’s decision-making process by identifying key spectral features that influence classification outcomes. It was observed that ML-based approaches have the potential to enable continuous monitoring, contributing to the conservation of valuable seed resources. Additionally, these methods may offer a scalable solution that can be adapted for large-scale agricultural and conservation applications.

The availability of key soil nutrients, like soil organic carbon (SOC), plant-available nitrogen (N), available phosphorus (P), and available potassium (K), serves as a significant gauge of soil fertility and its operational capability. Soil analysis through spectroscopy has demonstrated itself as an expedited and reliable approach for forecasting overall soil attributes, especially in determining soil nutrients such as NPK. Spectroscopy conducted in the laboratory has been extensively employed for estimating soil nutrients. However, the prediction of nutrients using airborne spectroscopy, in particular employing airborne visible infrared imaging spectrometer-next generation (AVIRIS-NG), is still lacking internationally. Therefore, this investigation was conducted to explore the feasibility of airborne VIS-NIR spectroscopy using AVIRIS-NG for forecasting of SOC and NPK and to compare it with lab-based spectroscopy using four multivariate techniques namely partial least square regression (PLSR), support vector machine regression (SVMR), random forest (RF) and multivariate adaptive regression splines (MARS). 132 soil samples were gathered from the surface soil (0-0.15 m) of the study area, and they were georeferenced for laboratory analysis, which included both chemical and spectroscopic methods. The study findings indicated that, although airborne spectroscopy using AVIRIS-NG is less accurate than lab-based methods due to external factors affecting spectral quality, the application of advanced multivariate techniques can enhance airborne prediction accuracy, making the results comparable to those obtained from lab-based spectral analysis. In this study, the SVMR and RF models showed superior performance in comparison, while, MARS and PLSR models displayed lower performance in nutrient prediction. It was observed that the SVMR model excelled compared to the others for SOC (R2V = 0.78, RMSEV = 0.16 and RPDV = 1.64 for lab and R2V=0.72, RMSEV= 0.17, RPDV=1.61 for airborne), P (R2V = 0.78, RMSEV= 5.71 and RPDV=1.77 for lab and R2V=0.64, RMSEV=5.39, and RPDV=1.65 for airborne) and K (R2V = 0.83, RMSE = 89.13 and RPD = 2.25 for lab; R2V=0.78, RMSE=101.99 and RPD=1.93 for airborne) prediction, whereas, RF (R2V= 0.80, RMSEV = 33.47 and RPDV = 1.86 for lab and R2V=0.61, RMSEV=33.23, RPDV= 1.59 for airborne predictions) outperformed the others for N prediction. The results revealed that the best predictability was achieved for K for both lab and air-borne predictions using the SVMR model and the prediction accuracy could be categorized as very good and good for lab-based and airborne spectroscopy, respectively. The best-performing model i.e., SVMR was utilized to generate prediction maps of SOC, P, and K, while RF was employed to generate a prediction map of N from AVIRIS-NG data integrated with multispectral data of Sentinel-2 and digital model elevation data. Uncertainty was assessed and maps revealed fewer prediction uncertainties and better prediction accuracy. The relevant wavelengths for the prediction of each studied nutrient were identified for both lab-collected and airborne spectra. Hence, this study proves how applying a non-linear multivariate technique can significantly increase the prediction accuracy of the aerial AVIRIS-NG sensor and bring airborne evaluation closer to the laboratory-based spectral technique. This study will serve as a guide for the possible use of airborne spectroscopy particularly with AVIRIS-NG for more precise on-site or in situ soil property or soil nutrient prediction.

Biosynthetic gene clusters (BGCs) encode enzymatic pathways that produce diverse natural products with applications in pharmaceuticals, agriculture, and biotechnology. Broad-spectrum tools such as antiSMASH and DeepBGC cover many BGC classes but may face challenges in detecting atypical or hybrid architectures. Here we present RFBGCPred, an open-source machine learning classifier focused on five clinically and agriculturally important classes-PKS, NRPS, RiPPs, terpenes, and PKS-NRPS hybrids. Rather than replacing existing pipelines, RFBGCPred complements them by improving class-level discrimination within this subset. Using curated data from the MiBIG database, we applied Word2Vec for feature extraction, supervised UMAP for dimensionality reduction, and SMOTE to address class imbalance. Multiple classifiers were benchmarked, with Random Forest identified as the top performer using the TOPSIS decision-making criterion. The final model achieved an accuracy of 98.0 % (MCC: 0.9752, AUC: 0.9928) on a balanced test set, and maintained strong generalization on an unbalanced validation set (accuracy: 94.8 %, MCC: 0.89, AUC: 0.96). Compared with antiSMASH and DeepBGC, RFBGCPred showed improved recall for hybrid PKS-NRPS clusters while sustaining competitive precision, thereby reducing misclassification of atypical arrangements. RFBGCPred supports FASTA, GenBank, and CSV inputs, with full source code, curated datasets, and documentation available at: https://github.com/SHARANBASAPPA/RFBGCPred.git.

Accurate and automated detection of wheat spikes is essential for high-throughput phenotyping and yield prediction, yet traditional manual counting is labor-intensive and error-prone. This study compared two deep learning models, U-Net and FasterViT, for wheat spike segmentation using pseudo-RGB images derived from hyperspectral data (400–1000 nm). A dataset of 400 wheat plants was collected at physiological maturity and annotated pseudo-RGB images were used for model training and testing. U-Net achieved a pixel accuracy of 0.893, a recall of 0.834, and a Dice score of 0.761. FasterViT outperformed U-Net with a pixel accuracy of 0.922, Intersection over Union (IoU) of 0.836, and a Dice score of 0.860, demonstrating better generalization and sharper segmentation of spikes. In terms of computational efficiency, U-Net required 2.5 seconds per image, whereas FasterViT required 6.85 seconds per image, reflecting a trade-off between speed and accuracy. Although the controlled dataset size was limited, the findings highlight the feasibility of low-resolution hyperspectral imagery for spike trait analysis. Future extensions could focus on field-based validation and integration into yield prediction pipelines to advance scalable precision agriculture.

Object detection is a challenging and important field in the domain of computer vision. Automated detection and counting of the number of siliques non-destructively, which is the most important yield component trait in Indian mustard (B. juncea), which is important as well as widely cultivated oilseed in India, is done here. We have used Deep Learning based one-stage object detection model, YOLOv5 in the counting of mustard siliques from the images captured by a DSLR camera (SONY Alpha 7iii, 24.2MP) manually mounted on a tripod. A GUI-based desktop software, MuSiC v1.0 is developed using Tkinter, Python backed by YOLOv5 model developed from annotated training dataset using Roboflow annotator. Approximately 22,800 annotations were used to train the deep learning YOLOv5 model and the whole dataset was partitioned into training, testing and validation in the ratio of 8:1:1. The Confusion matrix, F1 score and mAP showed a moderate result of 66

The evaluation of seed viability is pivotal in agriculture, biodiversity conservation, and ecological research. Traditional methods used for testing the seed viability are often destructive and pose challenges regarding labour intensity and seed wastage. The presented study introduces a non-destructive approach by integrating Near-Infrared (NIR) spectroscopy with advanced Machine Learning (ML) algorithms to enhance seed viability assessments. A diverse dataset of NIR spectral data from various seed species was used and analysed using three sophisticated ML models, namely Linear Discriminant Analysis (LDA), Random Forest (RF), and Artificial Neural Networks (ANN). The performance of the developed models were evaluated based on accuracy, precision, recall, and F1 score metrics. Furthermore, the experimental results demonstrated that NIR spectroscopy and ML could effectively classify viable seed. The integration of ANN particular promises due to their proficiency in extracting subtle patterns from spectral data with achieving an accuracy of 95% approximate which is crucial for accurate classification. It is observed that ML based approaches may help to conserve valuable seed resources and may offer a scalable solution that can be adapted for large-scale agricultural and conservation applications.

This study investigates aspect-based sentiment analysis of educational course feedback using a hybrid approach that combines keyword-based aspect extraction with traditional Machine Learning, Deep Learning, and transformer-based model predictions. The proposed methodology leverages the interpretability of keyword matching alongside the adaptability of Machine Learning, Deep Learning and transformer models to enhance overall performance, particularly when dealing with imbalanced, multi-aspect datasets like those derived from Massive Open Online Courses. The development of the hybrid approach involved training on a balanced synthetic dataset to establish a foundational understanding of aspect distribution, followed by fine-tuning on an imbalanced real-world dataset, thereby improving robustness in handling skewed data patterns. Aspect detection begins with a keyword-based extraction step, identifying relevant aspects within a review, which are then augmented with Machine Learning, Deep Learning and transformer model predictions to provide a more robust ensemble solution. Sentiment classification was applied to each detected aspect, focusing specifically on relevant segments of text to deliver more precise sentiment outcomes. This hybrid approach notably improved the interpretability and adaptability of traditional models. Transformer models such as Bidirectional Encoder Representations from Transformers and Robustly optimized Bidirectional Encoder Representations from Transformers approach achieved high precision, recall and F1-scores independently, demonstrating their capability in extracting complex, context-dependent sentiment without needing keyword support. Overall, the results highlight the benefits of integrating keyword-based and model-driven approaches for handling multi-aspect reviews, while also emphasizing the superior ability of transformer models to analyze nuanced feedback in educational settings.. KEYWORDS :Machine learning, Deep learning, Transformers, Aspect extraction, Sentiment analysis.