A Hybrid Framework for Song Lyric Annotation Based on Human-LLM Alignment

Rashini Liyanarachchi; Frank Tran; Md Mahmudul Hasan; Aditya Joshi; Erik Meijering

cs.CLJun 28, 2026

A Hybrid Framework for Song Lyric Annotation Based on Human-LLM Alignment

Authors: Rashini Liyanarachchi, Frank Tran, Md Mahmudul Hasan, Aditya Joshi, Erik Meijering

Organizations: School of Computer Science and Engineering, University of New South Wales, Sydney, Australia

Abstract

Emotion recognition of song lyrics is a challenging task since lyrics may not necessarily align with the overall emotion of a song. As a result, lyrics annotation remains largely underexplored. Drawing inspiration from research in large language model (LLM) assisted annotation, we examine the alignment between humans and LLMs for annotation of lyrics by creating a new sentence-level dataset of lyrics. Our observations highlight the subjectivity of the task and the inherent challenges. Following this, we present a hybrid annotation framework that optimizes human and LLM annotation by predicting potential misalignment in annotation.

Explore similar work

Aligning Language Models for Lyric-to-Melody Generation with Rule-Based Musical Constraints
Apr 20, 2026Hao Meng, Siyuan Zheng, Shuran Zhou +2Music Generation Large Language Model Alignment
SHALA-LLM: Smartly Handling Ambiguous Labels in Aligning LLMs
Jun 3, 2026Jingyao Wu, Ashley Wang, Keane Ong +2Large Language Model Alignment Natural Language Inference
Aligning MusicLLM with Emotion using Instruction Tuning and Feedback-Driven Alignment
Jun 23, 2026Takuya Hasumi, Welly NaptaliLarge Language Model Alignment Instruction Tuning

Apr 20, 2026 · cs.SDJ/K move · Enter open · S save

Aligning Language Models for Lyric-to-Melody Generation with Rule-Based Musical Constraints

Hao Meng, Siyuan Zheng, Shuran Zhou, Qiangqiang Wang+1

Zuoyebang Education Technology, Beijing, China

Large Language Models (LLMs) show promise in lyric-to-melody generation, but models trained with Supervised Fine-Tuning (SFT) often produce musically implausible melodies with issues like poor rhythm and unsuitable vocal ranges, a phenomenon we term "constraint violation". To address this, we propose a novel alignment framework that instills musical knowledge without human annotation. We define rule-based musical constraints to automatically generate a preference dataset from an SFT model's outputs. The model is then aligned through a sequential process, first using Direct Preference Optimization (DPO) on paired preference data, followed by Kahneman-Tversky Optimization (KTO) on unpaired negative samples. Experimental results demonstrate that our aligned model substantially reduces rule violations and outperforms strong baselines in both objective and subjective evaluations, generating melodies with substantially improved musicality and coherence. An interactive demo with audio comparisons is available at https://arain233.github.io/AligningMelody-demo.

Music Generation Large Language Model Alignment Large Audio Language Models Large Language Model Generation Supervised Fine-Tuning

Open paper Ask about it PDF

Abstract

Explore similar work

Aligning Language Models for Lyric-to-Melody Generation with Rule-Based Musical Constraints

SHALA-LLM: Smartly Handling Ambiguous Labels in Aligning LLMs

Aligning MusicLLM with Emotion using Instruction Tuning and Feedback-Driven Alignment