Natural Language Processing news and analysis on Web Pulse: 3944 curated articles tracking the latest Natural Language Processing developments, tools, and research, updated continuously from vetted sources.
A developer built Dub Any Video, a tool that dubs video courses into another language using local text-to-speech, after finding the best Godot course was in Spanish and dubbing services were too expensive. The tool uses …
A computer engineer explains that Siri and Alexa sound human through text-to-speech technology, which simulates the human voice by breaking words into phonemes and using machine learning to create natural speech. The tec…
Talkform.org, a new front-end AI tool, converts voice inputs into structured data directly in the browser, processing audio locally to enhance privacy and reduce latency. The tool is showcased on Hacker News as a 'Show H…
IBM Research introduced DocLang, a constrained XML sub-language designed to make document structure more interpretable for large language models, complementing its open-source Docling parser. Docling, IBM's most successf…
Morten Punnerud-Engelstad released mpe-lkg, a local knowledge graph tool that visualizes a language model's step-by-step reasoning using Ollama and embeddings, with blue edges for embedding similarity and green edges for…
Futuro Corporation's Futuro AI builds custom AI voice agents that answer business calls 24/7, trained on a company's own pricing, policies, FAQs, and brand voice via its MasterMind knowledge engine, which ingests up to t…
GeekyAnts has launched an Execution Intelligence AI Accelerator that adds an AI assistant to WhatsApp project groups, interpreting written updates in real time and recommending actions for human approval before syncing t…
Researchers introduced MPAR-Bench, a bilingual English-Chinese benchmark with 1,000 items that tests large language models' reasoning breadth through multi-point associative reasoning, finding that perturbations reduce a…
A study of 28,592 French news headlines from 2022 to 2025 found that La France insoumise (LFI) is more often framed through conflict, while Rassemblement National (RN) is more often framed through strategic-electoral ter…
Researchers introduced VialectBench, the first systematic benchmark for evaluating LLM robustness to Vietnamese dialects, containing 400 Standard Vietnamese instances and 2,400 human-written dialectal rewrites across six…
Researchers propose a lightweight, offline-trained logit correction for Grammar Constrained Decoding (GCD) that uses the parser and lexer states already computed during incremental parsing to restore the language model's…
A new arXiv study (2608.10046v1) analyzing 300 curated CVs of ML engineers, data scientists, and software engineers found that candidates disclose soft skills through narrative rather than keyword lists by a ratio of rou…
Researchers introduced PERCEPT, the first publicly available large-scale Persian-English code-mixed corpus annotated with Universal Dependencies POS tags, comprising 6,800 posts from X, Instagram, and Digikala. The corpu…
An audit of embedding-cosine similarity gates in agent frameworks found that these safety checks often fire backwards, with a production drift guard catching 0 of 56 meaning-breaking mutations and approving a reversal fr…
NVIDIA and academic collaborators introduced TAR (Traffic Anomaly Reasoning) and TAR-Bench, datasets with 44,040 chain-of-thought training annotations across 10 tasks for 3,670 CCTV videos (~26 hours) from eight public d…
LinkedIn's self-evolving agentic customer support system, integrating retrieval-augmented generation with evolutionary auto-prompting, increased QA self-serve by 9.0 percentage points, cancellation self-serve by 4.8 poin…
Most RAG pipelines fail because they use symmetric embedding models for asymmetric query-document pairs, according to Zilliz and developer Mario Ottmann. The fix is to use asymmetric embeddings or APIs like Gemini's task…
Telnyx has published a code example demonstrating a voice agent built on its Edge Compute platform that can hold a live phone call. The agent answers inbound calls, streams transcription, uses Telnyx AI Inference to gene…
New research from the United States concludes that speaking to AI models such as ChatGPT or Gemini produces worse results than typing prompts, even when typed prompts contain mistakes. The study found that transcription …
A new guide comparing eight AI meeting assistants as of August 2026 finds that most tools stop at producing a document, leaving follow-up work to humans, and ranks Zoye as the only one with native CRM write-back and draf…