Natural Language Processing news and analysis on Web Pulse: 5475 curated articles tracking the latest Natural Language Processing developments, tools, and research, updated continuously from vetted sources.
Minions.AI, a voice-based service dispatch platform for trade contractors, reduced voice AI response times on real phone calls from ~2,900ms to an average of 1,450ms (sub-1.8s in 99% of live calls). Developer Parvej Shah…
A study of 10,211 real scam and spam calls collected over 54 days by an AI voice-agent honeypot found that scammers spend about 15% more conversational turns per decade of a target's apparent age (rate ratio 1.15, 95% CI…
A developer is building an AI-powered real estate business assistant for the Indian market, designed to help brokers and agents manage leads and property matching through natural language conversations. The system uses G…
Dries Buytaert, founder and CTO of Acquia, added a related-posts feature to his blog using embeddings from BAAI's bge-base-en-v1.5 model, which returns 768-dimensional vectors per post. He applies mean-centering to corre…
Most RAG implementations are over-engineered, and a simple full-text search (BM25) often suffices, according to a technical guide that outlines a recipe-based approach. The guide recommends starting with full-text search…
A developer built a lightweight intent detector using embeddings to spot actionable messages in a chat, avoiding the cost of calling a large language model on every message. The detector achieves 92% recall on generated …
A developer built SummarizeVideoToText, a tool that converts YouTube videos into text workspaces with transcripts, AI summaries, and Q&A panels. The project revealed three unexpected challenges: YouTube's transcript endp…
Amazon Textract, a managed document analysis service, extracts text, forms, and tables from scanned documents, going beyond plain OCR to preserve structure. The service enables automated data extraction from invoices and…
Voice AI systems are shifting from cascaded pipelines to native multimodal speech-to-speech models, but the older architecture remains widely used due to its flexibility, security, and cost advantages. Cascaded systems, …
Researchers introduced LUX (Lesion-aware Unified eXplainable captioning), a graph-conditioned vision-language architecture for explainable endoscopic image captioning in ulcerative colitis, which constructs a lesion-cent…
Researchers introduced ESQ-Bench, an Oracle-first NL2SQL benchmark with six populated schemas (465 tables, 164,682 rows) and 550 gold-validated question-query pairs, finding that GPT-4o schema-linked execution accuracy d…
Traditional media asset management (MAM) systems fail because they rely on manual tagging, but AI-powered MAMs automatically generate metadata through visual recognition, audio analysis, and semantic search, cutting asse…
Vercel's AI Gateway now offers Google's Gemini 3.5 Transcribe model, which converts audio to text in two variants: a single-request mode and a live WebSocket mode that updates transcripts in real time. The model supports…
Hugging Face's Sentence Transformers library v6.0 introduces a fourth model type, MultiVectorEncoder, for ColBERT-style late interaction retrieval, along with a complete training approach. The blog post demonstrates fine…
SnapplAI, an open-source job-hunting pipeline created by TDK-99, scrapes LinkedIn listings, uses Google's Gemini 3.5 Flash Lite AI to summarize and score each job against a user's CV, and emails only the top matches. The…
Voiced, a new AI-powered content tool, turns a two-minute voice note into seven drafts — three X threads, three LinkedIn posts, and a newsletter draft — in the user's own voice. The tool, available at getvoiced.co, is de…
ChirpGo, a private desktop dictation app for macOS, Windows, and Linux, offers local-first voice-to-text transcription that works offline and keeps audio private, with an optional AI Polish feature for cleaning up text w…
Agentic retrieval techniques that measurably improve accuracy in production fall into six patterns, according to a 12,000-trajectory study covered in 'Before Reasoning Can Fail' (arXiv:2608.02011). A forced read-before-a…
IBM researchers released Granite Speech 5.0 Turbo CTC, two 470-million-parameter English speech-recognition models that remove the language-model decoder to achieve over 20 times the throughput of earlier Granite Speech …
Oro announced a $3 million strategic funding round co-led by MH Ventures and Mapleblock Capital, bringing its total raised to $4 million, to advance its natural-language DeFi platform that converts user goals into multi-…