TamilEOT is a new dataset and two audio‑only models for detecting semantic end‑of‑turn in Tamil telephone speech, comprising 18,485 labeled turn boundaries from 116 real conversations. The models, fine‑tuned from Smart Turn v3, achieve 83.71% and 86.13% accuracy on a held‑out set, with ROC‑AUC rising from 0.751 to 0.921, and run in under 150 ms on a laptop CPU. The paper also reports the cost of building the dataset, the accuracy of rule‑derived labels versus human agreement, and the impact of encoder capacity on performance. whyItMatters":"The study provides the first publicly available Tamil end‑of‑turn detection resource, enabling more accurate voice agents for South Indian languages and demonstrating efficient, low‑latency models that outperform zero‑shot baselines."
arXiv:2610.35272v1 Announce Type: cross Abstract: A transcriber listening through one microphone also hears everyone near it. We train causal voice isolators: masking transformers of 140M, 243M and 3...
The study evaluates conversational speech models by placing them at 427 pauses and 482 turn ends from 176 unscripted conversations, scoring their responses on timing, latency, length, wording, and recall of recent dialogue. Hosted APIs (GPT Realtime, Gemini 3.1 Flash Live, Grok Voice) respond at almost every pause, with latency ranging from 1–4 s for silence-based endpointers and about 8 s for semantic turn detection, while full‑duplex models (PersonaPlex, Moshi) remain silent more often and can interrupt within human reaction time. Recall rates vary: hosted systems remember 41–60 % of facts under a Claude Opus 5.5 judge and 19–26 % with a strict word matcher; full‑duplex models recall 17–22 % and 12–15 % respectively.
arXiv:2609.14817v1 Announce Type: new Abstract: In Yor\`ub\'a, pitch alone separates \d{o}k\d{o} (husband, Mid), \d{o}k\d{\`o} (vehicle, Low), and \d{o}k\d{\'o} (hoe, High) -- the diacritics ARE the...
arXiv:2610.20263v1 Announce Type: new Abstract: Every multilingual speech pipeline begins with a language decision, an error there propagates silently through recognition, translation and routing. In...
The paper investigates how to evaluate generative audio large language models (Audio‑LLMs) on known closed‑set tasks by separating the decision to call a generative model from the use of acoustic evidence. It introduces a controlled call‑decision framework where a policy can choose between a transcript label, encoder evidence from CLAP, AST, or WavLM, or a generative call to Qwen2‑Audio, Qwen2.5‑Omni, or MOSS‑Audio, and measures the impact of generative calls on accuracy. Results on the VocalSound dataset show that while transcript‑only accuracy is low (0.296), encoder‑based controls achieve high accuracy (≈0.85) without any generative calls, and adding generative calls yields only a marginal improvement (0.925 vs. 0.921).
BanglaTurn is a new corpus of 35,374 Bangla podcast speech samples, each 3 to 15 seconds long, labeled for end‑of‑turn detection through speaker diarization, an LLM pass, and human verification. A Whisper‑based model with task‑specific classification heads achieves 84.33 % accuracy on a balanced test set, outperforming the Smart‑Turn v3 baseline (69.28 %) and reducing the false‑negative rate from 51.57 % to 7.55 %, though with a higher false‑positive rate. The study also details the contributions of encoder‑layer fine‑tuning, multi‑scale pooling, INT8 quantization, and reports inference latency of 165–191 ms on CPU.
The paper details a multi‑month engineering effort to build Sophea, a production Greek‑English automatic speech recognition system. It evaluates the system against nine production gates—including word error rate, language identification, and hallucinations—across twenty‑three training iterations and two model architectures, finding that no single training‑data composition satisfies all gates. The authors describe a six‑stage data pipeline that dramatically reduces discarded Greek audio, isolate a hallucination defect to a specific training package, and demonstrate that a three‑model ROVER ensemble achieves full gate coverage and a 29 % relative WER improvement. A separate learned arbiter model is released on the Open ASR Leaderboard with 4.26 % average WER on English tests and 25.88 % on live Greek noisy traffic.
The paper introduces WaxalNet, a benchmark and adaptation study for spontaneous speech recognition across 19 African languages using the WAXAL corpus. It evaluates zero‑shot and fine‑tuned performance of several multilingual foundation models (Whisper, MMS, Omnilingual ASR, SeamlessM4T‑v2), showing that fine‑tuning can substantially reduce word error rates and that LLM decoders generally outperform CTC decoders when the encoder is shared. The authors also demonstrate that a single multilingual model can serve all languages with only a modest WER penalty and provide a native‑speaker audit of failure modes.
The paper investigates how to evaluate audio‑language models by separating the use of acoustic evidence from the need to invoke a generative audio model. Using a controlled call‑decision framework, the authors compare policies that rely on transcript labels, encoder outputs from CLAP, AST, or WavLM, and optional calls to generative models such as Qwen2‑Audio, Qwen2.5‑Omni, or MOSS‑Audio. Results on the VocalSound dataset show that while transcript‑only accuracy is low (0.296), encoder‑only controls achieve high accuracy (≈0.85) without any generative calls, and adding generative calls yields only a marginal improvement (0.925 vs. 0.921).
ufakzeka-kulak is an open Turkish speech‑to‑text model with 31.6 M parameters, built using a FastConformer encoder, a token‑and‑duration transducer (TDT), and a CTC head. It was trained from scratch on 4,977 hours of data—281 hours of human transcription and 4,696 hours of machine‑generated subtitles that were verified by a second reading—under licenses permitting commercial use. The model outputs cased, punctuated Turkish text with numbers written as words, achieving a normalised WER of 0.0921 on FLEURS Turkish and 0.0916 on Common Voice 27.0 Turkish, outperforming Whisper base and small models, and it is released under Apache‑2.0.
O'ido is an inference engine that performs open‑vocabulary English speech recognition entirely on an ESP32‑S3 microcontroller, running a 13‑million‑parameter Conformer‑CTC model with int8 quantization and a 1.3‑million‑parameter language model. On LibriSpeech it achieves 3.70%/8.23% WER on test‑clean/test‑other, only 0.12 points below full precision, and outperforms the vendor’s recognizer and several laptop‑based tiny models. The firmware processes audio in about twice the duration of the input on a single core, with a streaming variant that delivers partial transcripts while speaking, though it is not yet real‑time. whyItMatters":"This work demonstrates that sophisticated, open‑vocabulary speech recognition can run on inexpensive, low‑power microcontrollers without dedicated neural accelerators, expanding the reach of voice interfaces to devices that previously could only support keyword spotting or fixed command sets."
MTVA-Bench is a new benchmark designed to evaluate the language model component of cascaded voice agents under realistic conditions. It simulates callers with an LLM, mocks backend tool responses, and scores both tool‑call correctness and conversational quality using two LLM judges, covering 49 agents, 490 scenarios, and 7 languages. The benchmark reveals that while models perform similarly on tool selection, they differ widely in argument handling, rule compliance, and dialogue quality, highlighting the nuanced challenges of real‑world voice interactions.