arXiv subjects

Xiuwen Zheng

Publications and source records attributed to Xiuwen Zheng.

15 recordsLinked to original sources

arXiv2026

Look Less, Hear Better: Jointly Rewarded GRPO for Streaming ASR

Streaming automatic speech recognition (ASR) must be judged jointly on what it transcribes and on how quickly it commits each word. Delayed streams modeling (DSM) has become the dominant paradigm for streaming large audio-language models, exposing a structural delay $τ$ that bounds the decoder's lookahead. We show that $τ$ is a poor proxy for user-perceived latency, and that the alignment-based supervision of DSM leaves latency on the table: the same forced-aligned transcript is used at every $τ$, forcing the model to withhold words it could already commit. We introduce AWED, a word-level emission-delay metric defined relative to the acoustic end of each word, and post-train a DSM recognizer with GRPO under a reward that scores transcription accuracy and measured delay jointly. Trained at a single operating point ($τ=6$ frames), our model dominates both its supervised fine-tuning initialization and the Voxtral Realtime backbone across all evaluated lookahead budgets: it cuts WER by 30.8\% relative at an 80\,ms structural delay, and by 5.7\% relative at 480\,ms while lowering median AWED from 1.17\,s to 1.04\,s. Latency-rewarded post-training thus advances the accuracy--latency Pareto frontier of streaming ASR without architectural change.

cs.SD↗

arXiv2026

GigaSpeechBench: A Real-World Multilingual Speech-to-Text Benchmark

While modern ASR systems achieve low error rates on high-resource benchmarks, such performance often overestimates real-world robustness. Existing evaluations address challenges in isolation, lacking a unified benchmark for domain terminology, age variation, dialects, accents, and low-resource languages, particularly across the Middle East and Southeast Asia, representing over one billion under-evaluated speakers. To address this gap, we introduce GigaSpeechBench, a comprehensive multilingual and multidimensional in-the-wild ASR & AST benchmark comprising 680 hours of human-annotated speech. It features five modules: (1) 12 low-resource Middle Eastern and Southeast Asian languages, plus challenging Japanese and Korean; (2) 6 Chinese dialects; (3) 6 English accents; (4) dense terminology across 12 vertical domains for Chinese and English; and (5) older adult and child speech. We further provide human-annotated Chinese and English translations for 11 languages to support AST evaluation. Extensive evaluations of leading foundation models and commercial APIs reveal significant performance degradation in these challenging settings, exposing critical evaluation blind spots.

eess.AS↗

arXiv2026

Something from Nothing: Data Augmentation for Robust Severity Level Estimation of Dysarthric Speech

Dysarthric speech quality assessment (DSQA) is critical for clinical diagnostics and inclusive speech technologies. However, subjective evaluation is costly and difficult to scale, and the scarcity of labeled data limits robust objective modeling. To address this, we propose a three-stage framework that leverages unlabeled dysarthric speech and large-scale typical speech datasets to scale training. A teacher model first generates pseudo-labels for unlabeled samples, followed by weakly supervised pretraining using a label-aware contrastive learning strategy that exposes the model to diverse speakers and acoustic conditions. The pretrained model is then fine-tuned for the downstream DSQA task. Experiments on five unseen datasets spanning multiple etiologies and languages demonstrate the robustness of our approach. Our Whisper-based baseline significantly outperforms SOTA DSQA predictors such as SpICE, and the full framework achieves an average SRCC of 0.761 across unseen test datasets.

eess.AS↗

arXiv2026

MICRO: A Lightweight Middleware for Optimizing Cross-store Cross-model Graph-Relation Joins [Technical Report]

Modern data applications increasingly involve heterogeneous data managed in different models and stored across disparate database engines, often deployed as separate installs. Limited research has addressed cross-model query processing in federated environments. This paper takes a step toward bridging this gap by: (1) formally defining a class of cross-model join queries between a graph store and a relational store by proposing a unified algebra; (2) introducing one real-world benchmark and four semi-synthetic benchmarks to evaluate such queries; and (3) proposing a lightweight middleware, MICRO, for efficient query execution. At the core of MICRO is CMLero, a learning-to-rank-based query optimizer that selects efficient execution plans without requiring exact cost estimation. By avoiding the need to materialize or convert all data into a single model, which is often infeasible due to third-party data control or cost, MICRO enables native querying across heterogeneous systems. Experimental results on the benchmark workloads demonstrate that MICRO outperforms the state-of-the-art federated relational system XDB by up to 2.1x in total runtime across the full test set. On the 93 test queries of real-world benchmark, 14 queries achieve over 100 speedup, including 4 queries with more than 100x speedup; however, 4 queries experienced slowdowns of over 5 seconds, highlighting opportunities for future improvement of MICRO. Further comparisons show that CMLero consistently outperforms rule-based and regression-based optimizers, highlighting the advantage of learning-to-rank in complex cross-model optimization.

cs.DB↗

arXiv2026

Towards Robust Dysarthric Speech Recognition: LLM-Agent Post-ASR Correction Beyond WER

While Automatic Speech Recognition (ASR) is typically benchmarked by word error rate (WER), real-world applications ultimately hinge on semantic fidelity. This mismatch is particularly problematic for dysarthric speech, where articulatory imprecision and disfluencies can cause severe semantic distortions. To bridge this gap, we introduce a Large Language Model (LLM)-based agent for post-ASR correction: a Judge-Editor over the top-k ASR hypotheses that keeps high-confidence spans, rewrites uncertain segments, and operates in both zero-shot and fine-tuned modes. In parallel, we release SAP-Hypo5, the largest benchmark for dysarthric speech correction, to enable reproducibility and future exploration. Under multi-perspective evaluation, our agent achieves a 14.51% WER reduction alongside substantial semantic gains, including a +7.59 pp improvement in MENLI and +7.66 pp in Slot Micro F1 on challenging samples. Our analysis further reveals that WER is highly sensitive to domain shift, whereas semantic metrics correlate more closely with downstream task performance.

eess.AS↗

arXiv2026

AVMeme Exam: A Multimodal Multilingual Multicultural Benchmark for LLMs' Contextual and Cultural Knowledge and Thinking

Internet audio-visual clips convey meaning through time-varying sound and motion, which extend beyond what text alone can represent. To examine whether AI models can understand such signals in human cultural contexts, we introduce AVMeme Exam, a human-curated benchmark of over one thousand iconic Internet sounds and videos spanning speech, songs, music, and sound effects. Each meme is paired with a unique Q&A assessing levels of understanding from surface content to context and emotion to usage and world knowledge, along with metadata such as original year, transcript, summary, and sensitivity. We systematically evaluate state-of-the-art multimodal large language models (MLLMs) alongside human participants using this benchmark. Our results reveal a consistent limitation: current models perform poorly on textless music and sound effects, and struggle to think in context and in culture compared to surface content. These findings highlight a key gap in human-aligned multimodal intelligence and call for models that can perceive contextually and culturally beyond the surface of what they hear and see. Project page: avmemeexam.github.io/public

cs.SD↗

arXiv2025

Aligning ASR Evaluation with Human and LLM Judgments: Intelligibility Metrics Using Phonetic, Semantic, and NLI Approaches

Traditional ASR metrics like WER and CER fail to capture intelligibility, especially for dysarthric and dysphonic speech, where semantic alignment matters more than exact word matches. ASR systems struggle with these speech types, often producing errors like phoneme repetitions and imprecise consonants, yet the meaning remains clear to human listeners. We identify two key challenges: (1) Existing metrics do not adequately reflect intelligibility, and (2) while LLMs can refine ASR output, their effectiveness in correcting ASR transcripts of dysarthric speech remains underexplored. To address this, we propose a novel metric integrating Natural Language Inference (NLI) scores, semantic similarity, and phonetic similarity. Our ASR evaluation metric achieves a 0.890 correlation with human judgments on Speech Accessibility Project data, surpassing traditional methods and emphasizing the need to prioritize intelligibility over error-based measures.

cs.LG↗

arXiv2025

The Interspeech 2025 Speech Accessibility Project Challenge

While the last decade has witnessed significant advancements in Automatic Speech Recognition (ASR) systems, performance of these systems for individuals with speech disabilities remains inadequate, partly due to limited public training data. To bridge this gap, the 2025 Interspeech Speech Accessibility Project (SAP) Challenge was launched, utilizing over 400 hours of SAP data collected and transcribed from more than 500 individuals with diverse speech disabilities. Hosted on EvalAI and leveraging the remote evaluation pipeline, the SAP Challenge evaluates submissions based on Word Error Rate and Semantic Score. Consequently, 12 out of 22 valid teams outperformed the whisper-large-v2 baseline in terms of WER, while 17 teams surpassed the baseline on SemScore. Notably, the top team achieved the lowest WER of 8.11\%, and the highest SemScore of 88.44\% at the same time, setting new benchmarks for future ASR systems in recognizing impaired speech.

cs.AI↗

arXiv2024

Fine-Tuning Automatic Speech Recognition for People with Parkinson's: An Effective Strategy for Enhancing Speech Technology Accessibility

This paper enhances dysarthric and dysphonic speech recognition by fine-tuning pretrained automatic speech recognition (ASR) models on the 2023-10-05 data package of the Speech Accessibility Project (SAP), which contains the speech of 253 people with Parkinson's disease. Experiments tested methods that have been effective for Cerebral Palsy, including the use of speaker clustering and severity-dependent models, weighted fine-tuning, and multi-task learning. Best results were obtained using a multi-task learning model, in which the ASR is trained to produce an estimate of the speaker's impairment severity as an auxiliary output. The resulting word error rates are considerably improved relative to a baseline model fine-tuned using only Librispeech data, with word error rate improvements of 37.62\% and 26.97\% compared to fine-tuning on 100h and 960h of LibriSpeech data, respectively.

eess.AS↗

arXiv2023

An Optimized Tri-store System for Multi-model Data Analytics

Data science applications increasingly rely on heterogeneous data sources and analytics. This has led to growing interest in polystore systems, especially analytical polystores. In this work, we focus on a class of emerging multi-data model analytics workloads that fluidly straddle relational, graph, and text analytics. Instead of a generic polystore, we build a ``tri-store'' system that is more aware of the underlying data models to better optimize execution to improve scalability and runtime efficiency. We name our system AWESOME (Analytics WorkbEnch for SOcial MEdia). It features a powerful domain-specific language named ADIL. ADIL builds on top of underlying query engines (e.g., SQL and Cypher) and features native data types for succinctly specifying cross-engine queries and NLP operations, as well as automatic in-memory and query optimizations. Using real-world tri-model analytical workloads and datasets, we empirically demonstrate the functionalities of AWESOME for scalable data science applications and evaluate its efficiency.

cs.DB↗

arXiv2022

AWESOME: Empowering Scalable Data Science on Social Media Data with an Optimized Tri-Store Data System

Modern data science applications increasingly use heterogeneous data sources and analytics. This has led to growing interest in polystore systems, especially analytical polystores. In this work, we focus on emerging multi-data model analytics workloads over social media data that fluidly straddle relational, graph, and text analytics. Instead of a generic polystore, we build a "tri-store" system that is more aware of the underlying data models to better optimize execution to improve scalability and runtime efficiency. We name our system AWESOME (Analytics WorkbEnch for SOcial MEdia). It features a powerful domain-specific language named ADIL. ADIL builds on top of underlying query engines (e.g., SQL and Cypher) and features native data types for succinctly specifying cross-engine queries and NLP operations, as well as automatic in-memory and query optimizations. Using real-world tri-model analytical workloads and datasets, we empirically demonstrate the functionalities of AWESOME for scalable data science over social media data and evaluate its efficiency.

cs.DB↗

arXiv2021

News Meets Microblog: Hashtag Annotation via Retriever-Generator

Hashtag annotation for microblog posts has been recently formulated as a sequence generation problem to handle emerging hashtags that are unseen in the training set. The state-of-the-art method leverages conversations initiated by posts to enrich contextual information for the short posts. However, it is unrealistic to assume the existence of conversations before the hashtag annotation itself. Therefore, we propose to leverage news articles published before the microblog post to generate hashtags following a Retriever-Generator framework. Extensive experiments on English Twitter datasets demonstrate superior performance and significant advantages of leveraging news articles to generate hashtags.

cs.CL↗

arXiv2020

An Algebraic Approach for High-level Text Analytics

Text analytical tasks like word embedding, phrase mining, and topic modeling, are placing increasing demands as well as challenges to existing database management systems. In this paper, we provide a novel algebraic approach based on associative arrays. Our data model and algebra can bring together relational operators and text operators, which enables interesting optimization opportunities for hybrid data sources that have both relational and textual data. We demonstrate its expressive power in text analytics using several real-world tasks.

cs.DB↗

arXiv2019

Scalable Community Detection over Geo-Social Network

We consider a community finding problem called Co-located Community Detection (CCD) over geo-social networks, which retrieves communities that satisfy both high structural tightness and spatial closeness constraints. To provide a solution that benefits from existing studies on community detection, we decouple the spatial constraint from graph structural constraint and propose a uniform CCD framework which gives users the freedom to choose customized measurements for social cohesiveness (e.g., $k$-core or $k$-truss). For the spatial closeness constraint, we apply the bounded radius spatial constraint and develop an exact algorithm together with effective pruning rules. To further improve the efficiency and make our framework scale to a very large scale of data, we propose a near-linear time approximation algorithm with a constant approximation ratio ($\sqrt{2}$). We conduct extensive experiments on both synthetic and real-world datasets to demonstrate the efficiency and effectiveness of our algorithms.

cs.DB↗

arXiv2019

Social Network of Extreme Tweeters: A Case Study

The number of posts made by a single user account on a social media platform Twitter in any given time interval is usually quite low. However, there is a subset of users whose volume of posts is much higher than the median. In this paper, we investigate the content diversity and the social neighborhood of these extreme users and others. We define a metric called "interest narrowness", and identify that a subset of extreme users, termed anomalous users, write posts with very low topic diversity, including posts with no text content. Using a few interaction patterns we show that anomalous groups have the strongest within-group interactions, compared to their interaction with others. Further, they exhibit different information sharing behaviors with other anomalous users compared to non-anomalous extreme tweeters.

cs.SI↗