Natural Language Processing news and analysis on Web Pulse: 5446 curated articles tracking the latest Natural Language Processing developments, tools, and research, updated continuously from vetted sources.
A new multilingual summary meta-evaluation dataset (BASSE) containing human judgments on 2,040 abstractive summaries reveals that proprietary judge LLMs correlate most strongly with human ratings, followed by criteria-sp…
Researchers introduced SABER-Math, the first fully automated benchmark for evaluating mathematical information retrieval without expert annotation, built from 283K high-school-level math problems. The benchmark reveals t…
Researchers propose Prior-Guided Tuning (PGT) with Contrastive Prior Steering (CPS), a training method that uses task-level natural-language priors as auxiliary learning signals for low-resource LLM training. On AmbiMath…
Researchers introduced TALKFA, a unified benchmark for Farsi dialogue generation and understanding, comprising three datasets: WIKI-FADIAL (4.2K Wikipedia-grounded dialogues), DAILYDIALOG-FA (6.6K dialogues with dialogue…
Researchers introduced EmoStance, a method that uses emoji distributions as weak supervision to control the affective orientation of empathetic response generation, achieving a 62.2% decisive win rate in blind pairwise e…
Indicate, a new open-source tool for transliterating 12+ Indic languages to and from English, was released on Hacker News, offering both a PyTorch-based local model and LLM backends with auto-detection of source scripts.…
AI voice agents are now handling customer service calls end-to-end, from recognizing speech to executing actions like rescheduling deliveries, with speech recognition accuracy above 95% in clean conditions but dropping w…
Google's September 2026 update to the Google Home app enables Gemini for Home to handle combined voice commands for media and smart home controls, such as 'Turn off the lights and stop the music,' and fixes bugs with set…
AI voice agents that answer calls instantly and resolve routine inquiries without hold time can cut call abandonment from 20–40% to under 5% and reduce cost per call from $5–$12 to $0.05–$0.15, according to a comparison …
A developer is building CallRing AI, an AI-powered call tracking and analytics platform that turns phone conversations into structured business data. The platform uses dynamic numbers, telephony APIs, and AI transcriptio…
Francisco Carlos Serra released Agent Voice Kit, an open-source tool that lets users talk to AI agents like Claude Code, pi, Codex, and Gemini CLI locally, with dictation and optional text-to-speech, running entirely on …
GlobalFoundries CIO rebuilt its data foundation with a real-time, governed platform before deploying AI agents, and will discuss this approach on September 10. The article explains that RAG systems' accuracy depends on t…
Comcast Technology Solutions (CTS) unveiled a suite of end-to-end AI-powered workflow applications for broadcasters, content owners, and operators ahead of the 2026 IBC Show, building on its VideoAI platform. The new app…
Filevine launched an AI-native citator and a companion brief-checking tool inside LOIS, its Legal Operating Intelligence System AI platform, claiming performance equal to or better than LexisNexis's Shepard's and Thomson…
Tether AI Research released open-source AI translation models, QVAC TranslatePsy-AfriSLM and QVAC TranslatePsy-AfriNano, supporting 19 and 8 African languages respectively, plus QVAC TranslatePsy-EuroNano for 9 European …
Perspective API, the de facto standard for toxicity measurement, will shut down at the end of 2026, exposing researchers' dependence on a tool they did not control, according to a paper by David Hartmann and colleagues. …
Meta launched Muse Voice Transcribe, an AI-powered real-time speech-to-text model supporting over 70 languages including Hindi, Tamil, Telugu, Malayalam, and Kannada, and topping the Artificial Analysis streaming speech-…
LLM tokenization, while a small part of overall latency, sits on the hot path and can occur multiple times per request, according to a technical analysis of GPT-2's reference encoder. The analysis shows that converting t…
WhisprNote, a new macOS app launched on Hacker News, offers offline dictation and autocomplete that works in any app without an account or cloud, with local transcription and AI cleanup. The app supports 25 languages, cu…
Meta Superintelligence Labs released Muse Voice Transcribe, a real-time audio perception model that performs streaming ASR, speaker diarization for 20+ speakers, and endpointing in a single autoregressive pass, available…