handwritten word recognition with IAM dataset using CNN-Bi-LSTM and Bi-GRU implementation.

Official repo of the article: Yousef, W. A., Ibrahime, O. M., Madbouly, T. M., & Mahmoud, M. A. (2019), "Learning meters of arabic and english poems with recurrent neural networks: a step forward for language understanding and synthesis", arXiv preprint arXiv:1905.05700

An AI-powered legal assistant designed to assist with legal queries using a RAG pipeline containing pretrained inLegalBERT model embeddings and a FAISS index trained on the Indian Penal Code (IPC). Gemini answers your queries and provides explanations based on RAG as well. It features a chatbot, judgement prediction and legal document generation.

Powerful XRP price forecasting using public data. Stacking ensemble (Bi-GRU/LSTM/CNN-LSTM + LightGBM/XGBoost, RidgeR). Fuses market OHLCV (CCXT), news sentiment & top50 whale activity. No API keys or signups. Easy setup. CPU/GPU-ready. Multi-horizon single run forecasting. Backtests + Predictions visuals: plot_charts & in-depth tensorboard dash

This paper consists of all source codes related to the paper "An Efficient Framework for Vietnamese Sentiment Analysis", SOMET 2020.

Implementation of CNN and bi-GRU in parallel to solve text classification. In this particular example, I and my colleague, Ida Novindasari, use the AG news dataset from torchtext.datasets.

Given 10 predefined relations like cause-effect, product-producer, etc., the goal was to define the relation and the direction of the relation between 2 entities in a sentence.

Stroke diagnosis

⚖️ AskLegal.ai is an advanced AI-powered tool designed to help with legal queries, court cases, and lawsuits. Trained on the Indian Penal Code (IPC) and real case documents, this assistant offers accurate, context-aware responses

A comparison of different machine learning models for hate speech detection. Trained on a twitter hate speech dataset with more than 25K records.

This project predicts AC.PA stock prices using deep learning. Among the models tested, LSTM + SVR performed best, followed by BiGRU + LSTM. GAN + Transformer underperformed due to instability, highlighting the strength of hybrid sequential models for financial forecasting.