Speech
16 notes, newest first. Includes the vault tags speech recognition, speech.
Finetuning Strategies for Querying Sounds by Vocal Imitation
The authors describe their winning strategy for a sound effect querying competition in 2025, which involved two fine-tuning strategies: contrastive learning with a frozen pretrained CED encoder and joint…
纪念江泽民同志诞辰100周年大会隆重举行 习近平出席大会并发表重要讲话
The 100th birthday celebration of Jiang Zemin was held in Beijing, with Xi Jinping attending and delivering an important speech. The event paid high tribute to Jiang Zemin's enduring contributions to the party and the…
2026中国产业转移发展对接活动（西藏）在拉萨举行
The Industrial Transformation and Development Coordination Activity in Tibet took place in Lhasa, with several high-ranking officials attending. The event was held to commemorate the 100th birthday of Jiang Zemin, a…
学习贯彻党的二十届四中全会精神
The Ministry of Industry and Information Technology (MIIT) held a grand event commemorating the 100th birthday of Jiang Zemin, with Xi Jinping delivering an important speech. The theme was to study and carry forward…
一图读懂《安全应急装备产业发展“十五五”规划》
A summary of the 'Safety and Emergency Equipment Industry Development '15-5' Plan'. The plan was announced at a meeting to commemorate the 100th birthday of Jiang Zemin, where Xi Jinping delivered an important speech.
Whisper-Aware LLM: Self-Supervised Uncertainty Learning for Robust Whispered Speech Recognition
The Whisper-Aware LLM is a new framework designed to improve the performance of ASR (Automatic Speech Recognition) systems for whispered speech. The model learns to perceive and react to the uncertainty associated with…
Bias Analysis of L2 Speaking Assessment Systems Using Concept Activation Vectors
Researchers used Concept Activation Vectors (CAVs) to analyze the bias of two automatic L2 speaking assessment systems: a text-based BERT grader and a multimodal grader based on Whisper. The study shows that the…
Bias Analysis of L2 Speaking Assessment Systems Using Concept Activation Vectors
Researchers analyze the fairness and interpretability of two automatic speaking assessment systems using Concept Activation Vectors (CAVs). They find that the recoverability and sensitivity to certain concepts depend…
Transcription Policy as a Latent Variable: Activating Controllable Verbatim ASR with Word-Level Timing
The authors propose a method to improve the accuracy of automatic speech recognition (ASR) models by controlling the activation of 'verbatim' versus 'intended' transcription styles. They argue that current models…
Autoresearch with Coding Agents: Generalizers and Metric-Maximizers on Quran Recitation Data
Researchers tested two coding agents, Claude Code and OpenAI Codex, on a real-world task of identifying Quranic verses in noisy speech transcripts. Both agents developed similar algorithms initially but diverged later…
Audio-Native Speech Recognition with a Frozen Discrete-Diffusion Language Model
Researchers train a discrete diffusion language model to transcribe speech instead of using autoregressive decoders that emit one token at a time. The model is based on a mixture-of-experts called DiffusionGemma, and…
When Synthetic Speech Is All You Have: Better Call GRPO
Researchers present a method called Group Relative Policy Optimization (GRPO) that improves the accuracy of synthetic speech-to-text conversion compared to traditional methods. This could be particularly useful in…
Reasoning LLM Improves Speaker Recognition in Long-form TV Dramas
Researchers present two contributions aimed at improving speaker recognition in long-form TV dramas. They create a large-scale benchmark called DramaSR-532K with over 900 characters and propose using a large reasoning…
Reasoning LLM Improves Speaker Recognition in Long-form TV Dramas
The researchers introduce a large-scale benchmark called DramaSR-532K for speaker recognition in long-form TV dramas and propose a new method, DramaSR-LRM, built on a large reasoning model (LRM). This method aggregates…
TRACE: Temporal Relationship-Aware Conversational Entrainment Detection in Dyadic Speech
Researchers introduce a dataset (DyadEE) for studying emotional entrainment in dyadic speech interactions, including both entrained and disrupted conversations. They also propose TRACE, a new framework that models…
TRACE: Temporal Relationship-Aware Conversational Entrainment Detection in Dyadic Speech
Researchers have introduced a new dataset called DyadEE for studying emotional entrainment in two-person conversations. They also developed a model called TRACE that uses acoustic embeddings from emotion-fine-tuned…
