[ 08 of 08 · 4 years ]

NAACL

Sentimatic: Sentiment-guided Automatic Generation of Preference Datasets for Customer Support Dialogue System

Automatic, sentiment-guided framework that produces large-scale preference datasets without human annotation — improves emotional appropriateness in customer-support LLMs.

ACL-W

STAR: Strategy-Aware Refinement Module in Multitask Learning for Emotional Support Conversations

Disentangles decoder representations and dynamically fuses task-specific info — state-of-the-art in both strategy prediction and response generation.

ICASSPW

Optimizing Music Source Separation in Complex Audio Environments Through Progressive Self-Knowledge Distillation

Fine-tuning strategy for hearing-aid–oriented source separation. Softening targets with previous-epoch predictions gives +1.2 dB SDR over the baseline.

KCC

ASMR Sound Generation through Attribute-based Prompt Augmentation

LLM-based prompt augmentation that explicitly models ASMR triggers — produces more detailed sound descriptions and improves T2A quality.

ISMIR

Track Role Prediction of Single-Instrumental Sequences

Predicts the track role of single-instrument sequences automatically. 87% symbolic / 84% audio accuracy — reduces manual annotation in MIR pipelines.

TOOL

// PROJECT

NSynth-MIDI-Renderer for massive MIDI dataset

Open-source renderer that picks NSynth candidates and locks a single preset across a MIDI sequence — used to synthesize massive symbolic-to-audio datasets.

TOOL

// PROJECT

MIDI-Metadata-Extractor

Segregates a multi-track MIDI file into individual instrument tracks, extracts per-track metadata, and supports measure-based truncation / save.

TOOL

// PROJECT

Soothing Sound White Noise Generator

Soothing-sound audio generator with five distinct noise colors and three weighting filters. Copyrighted with the Korea Copyright Commission.