Kejun Zhang

Fenfei Guo, Chen Zhang, Zhirui Zhang et al. · tencent-ai

This paper develops automatic song translation (AST) for tonal languages and addresses the unique challenge of aligning words' tones with melody of a song in addition to conveying the original meaning. We propose three criteria for effective AST -- preserving meaning, singability and intelligibility -- and design metrics for these criteria. We develop a new benchmark for English--Mandarin song translation and develop an unsupervised AST system, Guided AliGnment for Automatic Song Translation (GagaST), which combines pre-training with three decoding constraints. Both automatic and human evaluations show GagaST successfully balances semantics and singability.

Xinda Wu, Zhijie Huang, Kejun Zhang et al.

Pre-trained language models have achieved impressive results in various music understanding and generation tasks. However, existing pre-training methods for symbolic melody generation struggle to capture multi-scale, multi-dimensional structural information in note sequences, due to the domain knowledge discrepancy between text and music. Moreover, the lack of available large-scale symbolic melody datasets limits the pre-training improvement. In this paper, we propose MelodyGLM, a multi-task pre-training framework for generating melodies with long-term structure. We design the melodic n-gram and long span sampling strategies to create local and global blank infilling tasks for modeling the local and global structures in melodies. Specifically, we incorporate pitch n-grams, rhythm n-grams, and their combined n-grams into the melodic n-gram blank infilling tasks for modeling the multi-dimensional structures in melodies. To this end, we have constructed a large-scale symbolic melody dataset, MelodyNet, containing more than 0.4 million melody pieces. MelodyNet is utilized for large-scale pre-training and domain-specific n-gram lexicon construction. Both subjective and objective evaluations demonstrate that MelodyGLM surpasses the standard and previous pre-training methods. In particular, subjective evaluations show that, on the melody continuation task, MelodyGLM gains average improvements of 0.82, 0.87, 0.78, and 0.94 in consistency, rhythmicity, structure, and overall quality, respectively. Notably, MelodyGLM nearly matches the quality of human-composed melodies on the melody inpainting task.

Yue Liu, Shihao Zhu, Jun Xia et al.

Intent learning, which aims to learn users' intents for user understanding and item recommendation, has become a hot research spot in recent years. However, existing methods suffer from complex and cumbersome alternating optimization, limiting performance and scalability. To this end, we propose a novel intent learning method termed \underline{ELCRec}, by unifying behavior representation learning into an \underline{E}nd-to-end \underline{L}earnable \underline{C}lustering framework, for effective and efficient \underline{Rec}ommendation. Concretely, we encode user behavior sequences and initialize the cluster centers (latent intents) as learnable neurons. Then, we design a novel learnable clustering module to separate different cluster centers, thus decoupling users' complex intents. Meanwhile, it guides the network to learn intents from behaviors by forcing behavior embeddings close to cluster centers. This allows simultaneous optimization of recommendation and clustering via mini-batch data. Moreover, we propose intent-assisted contrastive learning by using cluster centers as self-supervision signals, further enhancing mutual promotion. Both experimental results and theoretical analyses demonstrate the superiority of ELCRec from six perspectives. Compared to the runner-up, ELCRec improves NDCG@5 by 8.9\% and reduces computational costs by 22.5\% on the Beauty dataset. Furthermore, due to the scalability and universal applicability, we deploy this method on the industrial recommendation system with 130 million page views and achieve promising results. The codes are available on GitHub (https://github.com/yueliu1999/ELCRec). A collection (papers, codes, datasets) of deep group recommendation/intent learning methods is available on GitHub (https://github.com/yueliu1999/Awesome-Deep-Group-Recommendation).

Haoxuan Liu, Zihao Wang, Haorong Hong et al.

This paper introduces MetaBGM, a groundbreaking framework for generating background music that adapts to dynamic scenes and real-time user interactions. We define multi-scene as variations in environmental contexts, such as transitions in game settings or movie scenes. To tackle the challenge of converting backend data into music description texts for audio generation models, MetaBGM employs a novel two-stage generation approach that transforms continuous scene and user state data into these texts, which are then fed into an audio generation model for real-time soundtrack creation. Experimental results demonstrate that MetaBGM effectively generates contextually relevant and dynamic background music for interactive applications.

Zihao Wang, Shuyu Li, Tao Zhang et al.

The rapidly evolving multimodal Large Language Models (LLMs) urgently require new benchmarks to uniformly evaluate their performance on understanding and textually describing music. However, due to semantic gaps between Music Information Retrieval (MIR) algorithms and human understanding, discrepancies between professionals and the public, and low precision of annotations, existing music description datasets cannot serve as benchmarks. To this end, we present MuChin, the first open-source music description benchmark in Chinese colloquial language, designed to evaluate the performance of multimodal LLMs in understanding and describing music. We established the Caichong Music Annotation Platform (CaiMAP) that employs an innovative multi-person, multi-stage assurance method, and recruited both amateurs and professionals to ensure the precision of annotations and alignment with popular semantics. Utilizing this method, we built a dataset with multi-dimensional, high-precision music annotations, the Caichong Music Dataset (CaiMD), and carefully selected 1,000 high-quality entries to serve as the test set for MuChin. Based on MuChin, we analyzed the discrepancies between professionals and amateurs in terms of music description, and empirically demonstrated the effectiveness of annotated data for fine-tuning LLMs. Ultimately, we employed MuChin to evaluate existing music understanding models on their ability to provide colloquial descriptions of music. All data related to the benchmark, along with the scoring code and detailed appendices, have been open-sourced (https://github.com/CarlWangChina/MuChin/).

Zihao Wang, Haoxuan Liu, Jiaxing Yu et al.

Amid the rising intersection of generative AI and human artistic processes, this study probes the critical yet less-explored terrain of alignment in human-centric automatic song composition. We propose a novel task of Colloquial Description-to-Song Generation, which focuses on aligning the generated content with colloquial human expressions. This task is aimed at bridging the gap between colloquial language understanding and auditory expression within an AI model, with the ultimate goal of creating songs that accurately satisfy human auditory expectations and structurally align with musical norms. Current datasets are limited due to their narrow descriptive scope, semantic gaps and inaccuracies. To overcome data scarcity in this domain, we present the Caichong Music Dataset (CaiMD). CaiMD is manually annotated by both professional musicians and amateurs, offering diverse perspectives and a comprehensive understanding of colloquial descriptions. Unlike existing datasets pre-set with expert annotations or auto-generated ones with inherent biases, CaiMD caters more sufficiently to our purpose of aligning AI-generated music with widespread user-desired results. Moreover, we propose an innovative single-stage framework called MuDiT/MuSiT for enabling effective human-machine alignment in song creation. This framework not only achieves cross-modal comprehension between colloquial language and auditory music perceptions but also ensures generated songs align with user-desired results. MuDiT/MuSiT employs one DiT/SiT model for end-to-end generation of musical components like melody, harmony, rhythm, vocals, and instrumentation. The approach ensures harmonious sonic cohesiveness amongst all generated musical components, facilitating better resonance with human auditory expectations.

Shulei Ji, Zihao Wang, Jiaxing Yu et al.

Video-to-music (V2M) generation aims to create music that aligns with visual content. However, two main challenges persist in existing methods: (1) the lack of explicit rhythm modeling hinders audiovisual temporal alignments; (2) effectively integrating various visual features to condition music generation remains non-trivial. To address these issues, we propose Diff-V2M, a general V2M framework based on a hierarchical conditional diffusion model, comprising two core components: visual feature extraction and conditional music generation. For rhythm modeling, we begin by evaluating several rhythmic representations, including low-resolution mel-spectrograms, tempograms, and onset detection functions (ODF), and devise a rhythmic predictor to infer them directly from videos. To ensure contextual and affective coherence, we also extract semantic and emotional features. All features are incorporated into the generator via a hierarchical cross-attention mechanism, where emotional features shape the affective tone via the first layer, while semantic and rhythmic features are fused in the second cross-attention layer. To enhance feature integration, we introduce timestep-aware fusion strategies, including feature-wise linear modulation (FiLM) and weighted fusion, allowing the model to adaptively balance semantic and rhythmic cues throughout the diffusion process. Extensive experiments identify low-resolution ODF as a more effective signal for modeling musical rhythm and demonstrate that Diff-V2M outperforms existing models on both in-domain and out-of-domain datasets, achieving state-of-the-art performance in terms of objective metrics and subjective comparisons. Demo and code are available at https://Tayjsl97.github.io/Diff-V2M-Demo/.

Shulei Ji, Songruoyao Wu, Zihao Wang et al.

The burgeoning growth of video-to-music generation can be attributed to the ascendancy of multimodal generative models. However, there is a lack of literature that comprehensively combs through the work in this field. To fill this gap, this paper presents a comprehensive review of video-to-music generation using deep generative AI techniques, focusing on three key components: visual feature extraction, music generation frameworks, and conditioning mechanisms. We categorize existing approaches based on their designs for each component, clarifying the roles of different strategies. Preceding this, we provide a fine-grained classification of video and music modalities, illustrating how different categories influence the design of components within the generation pipelines. Furthermore, we summarize available multimodal datasets and evaluation metrics while highlighting ongoing challenges in the field.

Le Ma, Shirao Yang, Zihao Wang et al.

The proliferation of large models has intensified the need for efficient data valuation methods to quantify the contribution of individual data providers. Traditional approaches, such as game-theory-based Shapley value and influence-function-based techniques, face prohibitive computational costs or require access to full data and model training details, making them hardly achieve partial data valuation. To address this, we propose Unlearning Shapley, a novel framework that leverages machine unlearning to estimate data values efficiently. By unlearning target data from a pretrained model and measuring performance shifts on a reachable test set, our method computes Shapley values via Monte Carlo sampling, avoiding retraining and eliminating dependence on full data. Crucially, Unlearning Shapley supports both full and partial data valuation, making it scalable for large models (e.g., LLMs) and practical for data markets. Experiments on benchmark datasets and large-scale text corpora demonstrate that our approach matches the accuracy of state-of-the-art methods while reducing computational overhead by orders of magnitude. Further analysis confirms a strong correlation between estimated values and the true impact of data subsets, validating its reliability in real-world scenarios. This work bridges the gap between data valuation theory and practical deployment, offering a scalable, privacy-compliant solution for modern AI ecosystems.

Jiaxing Yu, Xinda Wu, Yunfei Xu et al.

Lyric-to-melody generation aims to automatically create melodies based on given lyrics, requiring the capture of complex and subtle correlations between them. However, previous works usually suffer from two main challenges: 1) lyric-melody alignment modeling, which is often simplified to one-syllable/word-to-one-note alignment, while others have the problem of low alignment accuracy; 2) lyric-melody harmony modeling, which usually relies heavily on intermediates or strict rules, limiting model's capabilities and generative diversity. In this paper, we propose SongGLM, a lyric-to-melody generation system that leverages 2D alignment encoding and multi-task pre-training based on the General Language Model (GLM) to guarantee the alignment and harmony between lyrics and melodies. Specifically, 1) we introduce a unified symbolic song representation for lyrics and melodies with word-level and phrase-level (2D) alignment encoding to capture the lyric-melody alignment; 2) we design a multi-task pre-training framework with hierarchical blank infilling objectives (n-gram, phrase, and long span), and incorporate lyric-melody relationships into the extraction of harmonized n-grams to ensure the lyric-melody harmony. We also construct a large-scale lyric-melody paired dataset comprising over 200,000 English song pieces for pre-training and fine-tuning. The objective and subjective results indicate that SongGLM can generate melodies from lyrics with significant improvements in both alignment and harmony, outperforming all the previous baseline methods.

Zihao Wang, Le Ma, Chen Zhang et al.

Music as an emotional intervention medium has important applications in scenarios such as music therapy, games, and movies. However, music needs real-time arrangement according to changing emotions, bringing challenges to balance emotion real-time fit and soft emotion transition due to the fine-grained and mutable nature of the target emotion. Existing studies mainly focus on achieving emotion real-time fit, while the issue of smooth transition remains understudied, affecting the overall emotional coherence of the music. In this paper, we propose REMAST to address this trade-off. Specifically, we recognize the last timestep's music emotion and fuse it with the current timestep's input emotion. The fused emotion then guides REMAST to generate the music based on the input melody. To adjust music similarity and emotion real-time fit flexibly, we downsample the original melody and feed it into the generation model. Furthermore, we design four music theory features by domain knowledge to enhance emotion information and employ semi-supervised learning to mitigate the subjective bias introduced by manual dataset annotation. According to the evaluation results, REMAST surpasses the state-of-the-art methods in objective and subjective metrics. These results demonstrate that REMAST achieves real-time fit and smooth transition simultaneously, enhancing the coherence of the generated music.

Hang Zhao, Chen Zhang, Belei Zhu et al.

In this paper, we propose S3T, a self-supervised pre-training method with Swin Transformer for music classification, aiming to learn meaningful music representations from massive easily accessible unlabeled music data. S3T introduces a momentum-based paradigm, MoCo, with Swin Transformer as its feature extractor to music time-frequency domain. For better music representations learning, S3T contributes a music data augmentation pipeline and two specially designed pre-processors. To our knowledge, S3T is the first method combining the Swin Transformer with a self-supervised learning method for music classification. We evaluate S3T on music genre classification and music tagging tasks with linear classifiers trained on learned representations. Experimental results show that S3T outperforms the previous self-supervised method (CLMR) by 12.5 percents top-1 accuracy and 4.8 percents PR-AUC on two tasks respectively, and also surpasses the task-specific state-of-the-art supervised methods. Besides, S3T shows advances in label efficiency using only 10% labeled data exceeding CLMR on both tasks with 100% labeled data.

Zeqian Ju, Peiling Lu, Xu Tan et al.

Lyric-to-melody generation is an important task in automatic songwriting. Previous lyric-to-melody generation systems usually adopt end-to-end models that directly generate melodies from lyrics, which suffer from several issues: 1) lack of paired lyric-melody training data; 2) lack of control on generated melodies. In this paper, we develop TeleMelody, a two-stage lyric-to-melody generation system with music template (e.g., tonality, chord progression, rhythm pattern, and cadence) to bridge the gap between lyrics and melodies (i.e., the system consists of a lyric-to-template module and a template-to-melody module). TeleMelody has two advantages. First, it is data efficient. The template-to-melody module is trained in a self-supervised way (i.e., the source template is extracted from the target melody) that does not need any lyric-melody paired data. The lyric-to-template module is made up of some rules and a lyric-to-rhythm model, which is trained with paired lyric-rhythm data that is easier to obtain than paired lyric-melody data. Second, it is controllable. The design of template ensures that the generated melodies can be controlled by adjusting the musical elements in template. Both subjective and objective experimental evaluations demonstrate that TeleMelody generates melodies with higher quality, better controllability, and less requirement on paired lyric-melody data than previous generation systems.

Chen Zhang, Jiaxing Yu, LuChin Chang et al.

Automatic lyrics transcription (ALT), which can be regarded as automatic speech recognition (ASR) on singing voice, is an interesting and practical topic in academia and industry. ALT has not been well developed mainly due to the dearth of paired singing voice and lyrics datasets for model training. Considering that there is a large amount of ASR training data, a straightforward method is to leverage ASR data to enhance ALT training. However, the improvement is marginal when training the ALT system directly with ASR data, because of the gap between the singing voice and standard speech data which is rooted in music-specific acoustic characteristics in singing voice. In this paper, we propose PDAugment, a data augmentation method that adjusts pitch and duration of speech at syllable level under the guidance of music scores to help ALT training. Specifically, we adjust the pitch and duration of each syllable in natural speech to those of the corresponding note extracted from music scores, so as to narrow the gap between natural speech and singing voice. Experiments on DSing30 and Dali corpus show that the ALT system equipped with our PDAugment outperforms previous state-of-the-art systems by 5.9% and 18.1% WERs respectively, demonstrating the effectiveness of PDAugment for ALT.

Chen Zhang, Yi Ren, Xu Tan et al.

While neural-based text to speech (TTS) models can synthesize natural and intelligible voice, they usually require high-quality speech data, which is costly to collect. In many scenarios, only noisy speech of a target speaker is available, which presents challenges for TTS model training for this speaker. Previous works usually address the challenge using two methods: 1) training the TTS model using the speech denoised with an enhancement model; 2) taking a single noise embedding as input when training with noisy speech. However, they usually cannot handle speech with real-world complicated noise such as those with high variations along time. In this paper, we develop DenoiSpeech, a TTS system that can synthesize clean speech for a speaker with noisy speech data. In DenoiSpeech, we handle real-world noisy speech by modeling the fine-grained frame-level noise with a noise condition module, which is jointly trained with the TTS model. Experimental results on real-world data show that DenoiSpeech outperforms the previous two methods by 0.31 and 0.66 MOS respectively.

Chen Zhang, Xu Tan, Yi Ren et al.

Existing speech to speech translation systems heavily rely on the text of target language: they usually translate source language either to target text and then synthesize target speech from text, or directly to target speech with target text for auxiliary training. However, those methods cannot be applied to unwritten target languages, which have no written text or phoneme available. In this paper, we develop a translation system for unwritten languages, named as UWSpeech, which converts target unwritten speech into discrete tokens with a converter, and then translates source-language speech into target discrete tokens with a translator, and finally synthesizes target speech from target discrete tokens with an inverter. We propose a method called XL-VAE, which enhances vector quantized variational autoencoder (VQ-VAE) with cross-lingual (XL) speech recognition, to train the converter and inverter of UWSpeech jointly. Experiments on Fisher Spanish-English conversation translation dataset show that UWSpeech outperforms direct translation and VQ-VAE baseline by about 16 and 10 BLEU points respectively, which demonstrate the advantages and potentials of UWSpeech.

Xiao-gang Zhang, Shou-qian Sun, Ke-jun Zhang

Computation of semantic similarity between concepts is an important foundation for many research works. This paper focuses on IC computing methods and IC measures, which estimate the semantic similarities between concepts by exploiting the topological parameters of the taxonomy. Based on analyzing representative IC computing methods and typical semantic similarity measures, we propose a new hybrid IC computing method. Through adopting the parameter dhyp and lch, we utilize the new IC computing method and propose a novel comprehensive measure of semantic similarity between concepts. An experiment based on WordNet "is a" taxonomy has been designed to test representative measures and our measure on benchmark dataset R&G, and the results show that our measure can obviously improve the similarity accuracy. We evaluate the proposed approach by comparing the correlation coefficients between five measures and the artificial data. The results show that our proposal outperforms the previous measures.