• Leader of a 4-person team to build a QA system. • Implemented pre-trained CNN to extract the information from word, then used RNN to extract information from sentence based on the context of a word. • Plan to add more network layers to process information on different levels, such as query, context, output.
2018-9 - 2018-12
Movie Genre Predictor: Movie Classifier Based On Posters
• Leader of a 5-person team • Implemented neural networks to achieve multi-label classification and improved the model architecture by combing style feature and object detection information • In evaluation, the model’s accuracy is much higher than previous works