A NLP text generation model in Nepali language. For the dataset, I scraped the 100 headlines from the official Ekantipur website and then built a word dictionary by tokeinzing them. Each headline is converted into sequence and last item in the sequence is the output whereas items before that are input. This is how model is trained. So, when user input some words (seeds), the model finds the nearest word in the dictionary. Limitation: It can only generate the learned words. Since, the dataset is extremely small, the metrics are:

• Accuracy: 1.0000
• Val Accuracy: 0.0000e+00
• Loss: 0.4187
• Val Loss: 8.6558

Future Work: Increase the text corpus by 10x, Redefine the Model Architecture.