ANH TRUNG DUC DANG
Address: Kim Hoa Street - Dong Da Districs - Ha Noi - Vietnam Phone: +84378850290
Email: [email protected] - Github: https://github.com/ducanhdt
SPECIALTIES
Machine learning, Deep learning, Infomation retrival, Natural Language Processing
EDUCATION
Hanoi University of Science and Technology (HUST) 2017 - NOW
School of Information and Communication Technology
• Major: Computer Science GPA: 3.44/4 In-Major GPA: 3.64/4
WORKING EXPERIENCE
School of Information and Communication Technology - HUST DEC 2019 - NOW
• Role: NLP Research Assistant
• Job Description: Research on Vietnamese Natural Language Processing, focusing on problems such as Vietnamese diacritic res-
olution, multi-label intent classification, text normalization, etc
InfoRe Technology DEC 2018 - DEC 2019
• Role: NLP intern of Vietnamese Automatic Speech Recognition (ASR) Project
• Job Description: Develop tools to pre-process the input and refine the output of the ASR system. Focus on problems such as text
normalization, handle too long input, etc.
TECHNICAL SKILLS
• Languages: Python, C/C++, JavaScript, SQL
• Technologies/Libraries: Flask, Numpy, Mathplotlib, Sciki-Learn, Pytorch, Tensoflow, Git, React-Redux
PROJECT
SmartDialog - Intelligent conversational interaction platform DEC 2019 - NOW
NLP Research Assistant Number of member: 13 Custommer: Lumi (voice command in smarthome), Vietcombank (chatbot)
• Solve the problem of Vietnamese diacritic restoration using the Transformer-based model and adapt the model for the bank
domain. The project provides the solution with the accuracy of 98% and reduces predicting time by at least 3 times compared
to the previous methods.
• Research on common error types in Vietnamese to build dataset and train deep learning models for error correction. The model
works well with hypothetical data but not with actual data.
• Use Rasa to build a multi-intent classification model for bank chat data.
Movie Controller MAR 2020 - JUN 2020
Principle Investigator - The Chrome extension detects hand gestures from webcam to control the YouTube tab.
• Collect hand gesture video data form 150 facebook friends, with a total of 5 hand gestures.
• Extract and label video frames to get a training dataset containing approximately 6000 images each gesture.
• Fine-tune various pre-trained models such as Resnet, Inceptionnet, VGG16, etc. Use Pre-train object detection model such as
SSD mobilenet, Yolo, etc to get higher accuracy achieved 90%.
• Deploy Api by Flask, receive image capture by webcam from Chrome extension and return getsure to control YouTube tab.
Vietnamese Text Summarization OCT 2019 - DEC 2019
Principle Investigator - A React website for the extractive text summarize web for Vietnamese documents
• Build K-means clustering model using various vectorization method such as TF-IDF, Fasttext, etc to generate summary form Viet-
namese Document
• Deploy the model into a restful Api by Flask, build front-end web using ReactJS
AWARDS
3rd prize, SOICT-IBM Hackathon 2020 – Leader
• Lead a team of 5 members to build and present a solution with the theme of climate change prevention within 48 hours
• The final result is a propotype of Smart watering system using drones and AI to detect where are the plant lacks water.
1st prize, Vietnam region in Innoworks 2019 – Developer
• Developer of a 5-member team, construct a possible business product with given topics using Advantech’s technology
Top 10, Specialized Informatics Olympic of HUST, 2018
• Competive Programing Contest for HUST students
PUBLICATIONS
Anh, D. T. D., & Trang, N. T. T. (2020). TDP – A Hybrid Diacritic Restoration with Transformer Decoder. PACLIC 2020.