I'm a postdoctoral researcher at University of Copenhagen in CoAStaL group since 2022, advised by Prof. Anders Søgaard. Before that I received my integrated M.S. and Ph.D degree of Computer Science from Korea University in 2021. I am interested in NLP tasks with human inspired computation model and conversational agents. Specifically, evaluation of dialogue agents, generation, dialogue summarization.
Hello, I’m a Seolhwa, NLP researcher. I have rich experience in NLP & Deep learning techniques. Also I am good at
Abstractive dialogue summarization is challenging due to the most important pieces of information in a conversation are scattered across the utterances through multi-party interactions. To address this challenges, this research propose the syntax-aware sequence-to-sequence model for abstractive dialogue summarization.
Will preprint soonThere are a multitude of novel generative models for open domain conversational systems; however, there is no systematic evaluation of different systems. Systematic comparisons require consistency in experimental design, evaluation sets, conversational systems and their outputs, and statistical analysis. In this paper layout a protocol for the evaluation of conversational models using head-to-head pairwise comparison.
This work conducted during visiting student at JHU.
[Paper]Programming mistakes frequently waste software developers’ time and may lead to the introduction of bugs into their software, causing serious risks for their customers. Using the correlation between various software process metrics and defects, earlier work has traditionally attempted to spot such bug risks. However, this study departs from previous works in examining a more direct method of using psycho-physiological sensors data to detect the difficulty of program comprehension tasks and programmer level of expertise.
This work cited by Microsoft Research team [Paper]
[Paper 1] [Paper 2] [Slide]To combat misinformation regarding COVID- 19 during this unprecedented pandemic, we propose a conversational agent that answers questions related to COVID-19. We adapt the Poly-encoder (Humeau et al., 2020) model for informational retrieval from FAQs. We show that after fine-tuning, the Poly-encoder can achieve a higher F1 score. We make our code publicly available for other researchers to use.
[Paper] [Code] [Slide]The text-based search method has limitations because of the nature of the fashion industry, in which design is a very important factor. Therefore, we developed an intelligent fashion technique based on deep learning for efficient fashion product searches and recommendations consisting of a Sketch-Product fashion retrieval model and vector-based user preference fashion recommendation model.
[Paper]Bilingual word embedding models focus on the induction of a shared bilingual word embedding space where words from both languages are represented in a uniform language-independent manner such that similar words have similar representation. These research goal make seed lexicon by aligned parallel data and make mapping function between seed lexicon and general-domain corpora such as Wikipedia. Therefore, we expect to applicate these research to other NLP tasks such as machine translation.
[Demo]Wrap is a clean and elegant Multipurpose Landing Page Template. It will fit perfectly for Startup, Web App or any type of Web Services. It has 4 background styles with 6 homepage styles. 6 pre-defined color scheme. All variations are organized separately so you can use / customize the template very easily.
It is a clean and elegant Multipurpose Landing Page Template. It will fit perfectly for Startup, Web App or any type of Web Services. It has 4 background styles with 6 homepage styles. 6 pre-defined color scheme. All variations are organized separately so you can use / customize the template very easily.
All variations are organized separately so you can use / customize the template very easily.
All variations are organized separately so you can use / customize the template very easily.