Research Article
Evaluation of Herbicides on Upland Rice (Oryza sativa L.) Production in Fogera Plain, Northwestern Ethiopia
Yinebeb Abebaw Bekele*
Issue:
Volume 11, Issue 2, April 2025
Pages:
17-22
Received:
27 March 2025
Accepted:
15 April 2025
Published:
24 May 2025
Abstract: This study explores the effectiveness of various herbicides in enhancing upland rice production in the Fogera Plain of Ethiopia, a region recognized for its agricultural potential yet challenged by significant weed infestations. Despite a notable increase in both cultivated area and total rice output, the productivity in Ethiopia remains below the global average, primarily due to the adverse effects of weeds. To address this issue, a trial was conducted using a randomized complete block design (RCBD) during the 2023 rainy season, evaluating the impacts of Keeper herbicide, Pallas 45 OD, and manual weeding on key growth parameters, including grain yield, dry weight of weed biomass, panicle length, and spikelet count. The results indicated that the treatment involving two rounds of manual weeding achieved the highest grain yield of 3337.72 kg ha-1, significantly surpassing the unsprayed control yield of 763.69 kg ha-1. The Keeper herbicide yielded 2625.00 kg ha-1, while the Pallas 45 OD herbicide resulted in a lower yield of 1686.40 kg ha-1, demonstrating their effectiveness in managing weed competition, although with reduced yields compared to manual weeding. Furthermore, the economic analysis revealed that the Keeper herbicide treatment generated a greater net benefit compared to manual weeding, affirming its practicality as a viable alternative for weed management. This research highlights the necessity for integrated weed management strategies in Ethiopia’s rice production systems, emphasizing the potential of combining herbicides with traditional practices to mitigate the challenges posed by weeds and enhance agricultural productivity.
Abstract: This study explores the effectiveness of various herbicides in enhancing upland rice production in the Fogera Plain of Ethiopia, a region recognized for its agricultural potential yet challenged by significant weed infestations. Despite a notable increase in both cultivated area and total rice output, the productivity in Ethiopia remains below the ...
Show More
Research Article
TweetGuard: Combining Transformer and Bi-LSTM Architectures for Fake News Detection in Large-Scale Tweets
Kowshik Sankar Roy*,
Farhana Akter Bina
Issue:
Volume 11, Issue 2, April 2025
Pages:
23-45
Received:
22 April 2025
Accepted:
3 May 2025
Published:
10 June 2025
DOI:
10.11648/j.ijdsa.20251102.12
Downloads:
Views:
Abstract: The proliferation of misinformation on platforms like Twitter, where rapid dissemination can significantly impact public discourse, underscores the urgent need for effective automated fake news detection systems. These systems are crucial in preventing the spread of falsehoods and maintaining informational integrity. Traditionally, one of the challenges in developing such systems has been the lack of comprehensive benchmark datasets, which are essential for reliably training and testing detection models. Additionally, the rapid evolution of deceptive tactics makes traditional methods less effective, necessitating new approaches that can adapt to emerging misinformation patterns. In response to the challenges, a robust model named "TweetGuard" has developed, leveraging the 'TruthSeeker' dataset, a recently published benchmark offering a rich collection of annotated tweets. This dataset provides a solid foundation for training and refining our detection techniques. The proposed model employs a novel classification architecture that integrates transformer and Bi-LSTM technologies in a concatenation mode, enhanced by advanced preprocessing steps, including BERTweet, for effective tokenization and contextual understanding. An ablation study highlights the individual contributions of the Bi-LSTM and Transformer components, as well as their combined effect, demonstrating their critical roles in enhancing the model's performance. Compared to conventional classifiers, including various CNN, LSTM, Bi-LSTM, BERT and Transformer configurations, the proposed model demonstrates superior performance, as evidenced by comprehensive statistical testing. TweetGuard achieves an accuracy of 94.02%, an F1-score of 93.84%, and a ROC-AUC score of 0.9614 on the TruthSeeker dataset. Additional metrics, such as a Matthews Correlation Coefficient (MCC) of 0.8802 and a fake news detection rate of 93.70%, also demonstrate the model's stability and robustness. Its effectiveness and generalizability are further validated through rigorous testing across three additional fake news datasets, confirming its reliability and adaptability in diverse informational settings. This evaluation not only highlights our model's superior ability to identify and classify misinformation accurately but also establishes a new benchmark for automated fake news detection on social media platforms.
Abstract: The proliferation of misinformation on platforms like Twitter, where rapid dissemination can significantly impact public discourse, underscores the urgent need for effective automated fake news detection systems. These systems are crucial in preventing the spread of falsehoods and maintaining informational integrity. Traditionally, one of the chall...
Show More
