Howard W. Chung

Research Assistant of NLP Lab from SCU

Hi! My name is Howard, I specialize in data analysis, machine learning and natural language processing! I am moving towards being a successful data scientist! I like to challenge anything that is very difficult. Solving these problems allows me to find the achievements in life, and it is very interesting to be able to solve the problems in life with programs.


Picture

MoST Research Assistant

Project :Organizational behavior and performance
Content :Analyze CSR reports and data analysis using statistical software such as Access, Stata, python

Nov, 17 ~ Jun, 19
Movie

Research Assistant at IF.Lab

  • Developed a micro-expression recognition model to identify readers’ investment styles, and combined recommendation systems to recommend suitable financial articles for readers. Achieved 91% F1-score.
  • Deployed the financial recognition APP on Zenbo robot, combined with personalized dashboard to present financial information exclusive to users. This APP earned Honorable Mention in HNCB Fintechers competitions.
  • Mar, 18 ~ Jun, 19
    Picture

    Automatic Speech Recognition Summer Intern at Delta Electronics, inc.

  • Annotated noise span in recordings and modified the speech recognition models to detection noise.
  • Developed a noise detection deep learning model and deployed on the internal system with Flask. The model is stacked the CNN model with soft-attention mechanism and LSTM as prediction model. Achieved 89% accuracy rate in noise span.
  • Jul, 19 ~ Aug, 19
    Picture

    Data Analytics Intern at Deloitte, Inc.

  • Completed the deployment of machine learning model on the Django platform for economic indicator trend prediction.
  • Researched the practical application of OCR in financial statements and achieved a digital recognition rate of 86%.
  • Maintained news and social web crawlers, and developed robotic process automation (RPA) programs to handle unstructured data.
  • Sep, 19 ~ Jun, 20
    Movie

    Research Assistant at NLP Lab

  • Invented a combination model of neural network and statistic method for co-reference resolution. Improved F1-score from 78.3 to 79.2.
  • Developed a hierarchical attention network (HAN) model to learn the three different aspect of hope, trigger event and arousal from society community post. Improved the F1-score from 0.58 to 0.86 on the prediction of depression.
  • Invented a dynamic MRT station embedding model for passenger flow prediction. Replaced the traditional station embedding model Node2Vec and improved MAE from 1.47 to 0.93, 36% increased.Mainly research on natural language processing and text-mining.
  • Read more
    Sep, 20 ~ Jan, 2022