Pracha Promthaw

Hi! My name is Pracha Promthaw. I am a senior year undergraduate student in Computer Science major at UMass Amherst. I’ve experience in competitive programming for 6 years but currently I’m interested in information retrieval and natural language processing. Specifically, I would like to learn more about how we could improve the conversational system robustness and efficiency in term of memory and computation. I’m currently doing independent study about conversational system in specific domain with Atharva Nijasure and Chris Samarinas as a PhD mentor, advised by Professor Hamed Zamani.

Fun fact:

My hobbies are readings, working out, watching movies,and playing piano. I also love learning new languages, currently I can speak Thai (Native), English (Fluent), Japanese (Beginner), and French (Beginner).

Extracurricular Club:

Thai Student Association (Senior Advisor)
UMass Machine Learning Club (Event Coordinator)
Film Production Club (General Body Member)

news

Jun 30, 2023	Our team get into the semi-final phase during the Alexa Prize Taskbot Challenge Competition!
Apr 19, 2023	My first medium blog got published! here :
Mar 27, 2023	I have been qualified into the final round of QTFT Optimization Hackathon!
Feb 17, 2023	Present a poster at the Undergraduate Research Volunteer session in topic "Continuous control of latent diffusion model through token embedding interpolation"
Dec 7, 2022	Got accepted into the Alexa Prize Project !

latest posts

Apr 20, 2023	Save your time with these numpy techniques
Apr 19, 2023	Numpy Blog

selected publications

Amazon

MarunaBot V2: Towards End-to-End Multi-Modal Task-Oriented Dialogue Systems

Chris Samarinas, Pracha Promthaw, Rohan Lekhwani, and 5 more authors

In Alexa Prize TaskBot Challenge Proceedings, 2023

Abs PDF

We introduce MarunaBot V2, an advanced Task-Oriented Dialogue System (TODS) primarily aimed at aiding users in cooking and Do-It-Yourself tasks. We utilized large language models (LLMs) for data generation and inference, and implemented hybrid methods for intent classification, retrieval, and question answering, striking a balance between efficiency and performance. A key feature of our system is its multi-modal capabilities. We have incorporated a multi-modal enrichment technique that uses a fine-tuned CLIP model to supplement recipe instructions with pertinent images, a custom Diffusion model for image enhancement and generation, and a method for multi-modal option matching. A unique aspect of our system is its user-centric development approach, facilitated by a custom tool for tracking user interactions and swiftly integrating feedback. Finally, we showcase the promising results of our end-to-end retrieval-augmented LLM taskbot, MarunaChef, and set a promising precedent for future task-oriented dialogue systems.