- 👋 Hi, I’m @malinphy
-
- Classification of named entities in a given sentence. System is designed with pre-trained BERT model.
-
Question Answer Matching by univesal sentence encoder
- Tensorflow/Keras implementation of Question answer matching system. In short, this project is designed for finding the best possible answers for related questions.
-
Natural Language Inference Sentence embeddings
- Detection of whether given sentence logically follows to next sentence. Sentences are vectorized using Universal Sentence Encoder(USE) and system has been trained to optimize similarity of sentece vectors.
-
- Sentiment classification on twitter dataset. Deep neural network system on multi-class classification.
-
- Question answering with BERT model on SQUAD v1.1..
-
Active Learning for Hate Speech Detection
- Problem of detecting hate speech, offensive language, and neutral tweets with minimal labeled data through iterative model improvement
Recommender Systems Deep learning based modern recommendation systems has been worked.
-
- Retrieval part to reduce the number of potential items
- Ranking part to provide a recommendation list of items for each user.
-
Convolutional Ranking based recommender system
- Personalized Sequential Recommender System
-
Neural Collaborative Filtering
- Neural network approach of matrix factorization
-
- Engineered a three-stage e-commerce search systemusing fine-tuned deep learning models, (retrieval, pre-ranking, re-ranking)
- Two-tower and cross-encoder architectures optimized for product relevance.
-
Next Trip Recommendation System
- Sequential Recommendation System
- Next hotel prediction based on user history
-
- Click-Through (CTR) and Conversion Rates (CVR) prediction
- personalized hotel search