Paraphrase Generation Using Deep Reinforcement Learning - MSc Thesis

Multi-label prediction for GitHub issues with state of the art NLP models and state of the art performance. 😎

keras implementation for paper - Enhanced LSTM for Natural Language Inference, tested on Quora Question Pairs dataset.

Code for experiments done for EMNLP2020.

An Automatic Paraphrase Detection System

Sentential paraphrase datasets and BERT-based paraphrase detection models for the Armenian language.

ESIM implementation by PyTorch, get 89.29% accuracy on Quora Questions Pairs dataset.

This is the official repository of the paper titled "BnPC: A Gold Standard Corpus for Paraphrase Detection in Bangla, and its Evaluation", accepted in The 17th Workshop on Building and Using Comparable Corpora (BUCC 2024) co-located with LREC-COLING 2024. It contains the codes and the dataset.

I built this to automate discovery of common text between two documents. I used porter stemming, windowing, and make it save to file. I built the GUI in Java. This project was so successful that a competitor quickly added similar features and Dr. Hilton III published papers on results he discovered when using it.

The official repository for the paper "Paraphrase Detection: Human vs. Machine Content".

ACL 2022 (Findings): Striking a Balance: Alleviating Inconsistency in Pre-trained Models for Symmetric Classification Tasks

Group Project of CSE 576 : Natural Language Processing at ASU

BERT for question paraphrase detection

Given a text and a reason, predict if text satisfies the reason. You can use the train file for any training and report metrics on evaluation file.

Check matching seven word sequences in two different pieces of text.

PAS2: A Python-based hallucination detection system that evaluates AI response consistency through paraphrasing and semantic similarity analysis. Features include response evaluation, similarity metrics, visualization tools, and a web interface for interactive testing.

Document-level paraphrase detection using Siamese Neural Network with SBERT encoder (all-MiniLM-L6-v2). Features full transformer fine-tuning, contrastive learning, and semantic similarity analysis. Includes FastAPI backend with AI validation agents, React frontend, and complete training pipeline with checkpointing.

Fine-tuning BERT and BART for sentiment analysis, paraphrase detection, and controlled text generation. Features Siamese networks, multi-objective loss, adversarial training, and Bayesian hyperparameter optimization.

Master thesis code and results on topic "Transfer learning in paraphrase detection for English and Spanish languages"