Objective: This work aims to the development of Hadiyya language named entity recognition which is widely used in text summarization, machine translation, and information retrieval to categorizing and predicting tokens of a given corpus into predefined named entity classes Method : In this paper, a method combining Bidirectional Long Short-Term Memory neural network with Conditional Random Field (BiLSTM-CRF) is proposed to automatically recognize entities of Hadiyya language (Location, time, person, geography and other nonname entity) from annotated Hadiyya language corpus, the experiment in this work was conducted to discover the most suitable features for Hadiyya NER system. We have collected the data from Department of Hadiya Language & Literature (DHLL) at Wachemo University, Ethiopia. Hadiyya TV, and Hadiyya Media Network (HMN) Therefore, a newly annotated dataset having 5,148 instances is used for this study. We have used 70 % for training and 30% for testing Hadiyya NER system. Finding: after training and validating BiLSTM-CRF model using the collected dataset we have obtained a result of precision, recall and f1-measure values of 95.49%, 94.93%, and 95.21% respectively. Novelty: Finally, we have contributed by hybrid NER system in Hadiyya language to obtain state-of-the-art result which is independent of other natural language processing tasks.

Objectives: To design a Bacterial Wilt Enset (BWENet) model that can detect and grade the level of bacterial wilt disease in Enset using a deep convolutional neural network. Methods: Convolutional neural network is used for detection of bacterial wilt and 4-way Softmax is used for grading bacterial wilt into a specific level (normal, early-stage, infected stage, and completely wilted Enset). About 1600 images were used to evaluate the performance of the model out of which 70% is for training, 15% for validation and 15% for testing. We collected dataset from kefa, Sheka agricultural research center and some selected areas in SNNPR. The proposed model was trained using these images and data augmentation techniques were applied to generate more images. Findings: BWENet model achieved an identification test grading accuracy of 90.53% bacterial wilt disease. BWENet model is found to be faster to train and has a smaller model size as compared to State-of-the-art models like Alex Net, Google net, and VGG19. Novelty: we are proposing a new CNN architecture to grade bacterial wilt disease severity measurement. By our experiments, we have shown the kernel size selection, poling layer selection, and segmentation achieved better accuracy results.

Objectives: This study aims to advance Sheko language name entity Recognition first of its kind. Named Entity Recognition (NER) is one of the most important text processing in machine translation, text summarization, and information retrieval. Sheko language named entity recognition concerns in addressing the usage of the bidirectional Long Short-Term Memory (LSTM) model in recognizing tokens into predefined classes. Methods: A bidirectional long shortterm memory is used to model the NER for sheko language to identify words into seven predefined classes: Person, Organization, Geography, Natural Phenomenon, Geopolitical Entity, time, and other classes. As feature selection plays a vital role in long short-term memory framework, the experiment is conducted to discover the most suitable features for Sheko NER tagging task by using 63,813 words to train and test our model. Out of which is 70% for training and 30% for testing. Datasets were collected from Sheko Mizan Aman Radio Station (MARS), Sheko southern region mass media, Language, and Literature Department. Findings: Through several conducted experiments, Sheko NER has successfully achieved a performance of 97% test accuracy. From the experimental result, it is possible to determine that tag context is a significant feature in named entity recognition and classification for Sheko language. Finally, we have contributed a new architecture for Sheko NER which uses automatically features for Sheko named entity recognition which is not dependent on other NLP tasks, and added some preprocessing steps. We provide a comprehensive Comparison with other traditional NER algorithms.