#NL2SQL

📊 Latest #AI Research – Highlights Why do Chinese NL2SQL systems fail on structural/schema errors? New research says single-encoder is the culprit. Meet CIR-SQL: intent-first, dual-model, and built for Industry 4.0 databases. #NL2SQL #Industry40 Paper:

I want to build a Text-to-SQL system from scratch. 🧠 I’m comfortable with coding basics, but the actual NL ➡️ SQL logic is new territory. 🗺️ Q for the pros: If you had to learn this all over again today, what resource would you read first? Drop your gold mines! 👇 #NL2SQL #AI

I am thrilled to share DIY (Debug It Yourself), a novel technique that helps #SQL non-experts to 1. inspect for, 2. isolate, and 3. fix errors in #NL2SQL (#NaturalLanguage to #SQL) scenarios. (1/3)

I asked in plain language and Oracle AI Database answered in SQL 🧠➡️🗄️ AI + MCP on OCI, NL→SQL over the HR schema, with joins, a subquery and even a data-quality check (106 vs 107) explained by the AI itself. Full post 👇 #OracleAI #MCP #NL2SQL #OCI

Fabric Data Agents are still English-first. I published a hands-on guide on a quick-win multilingual setup: trial and error, prompt tuning, and guardrails to make it usable today. Hope you'll find it useful! #MicrosoftFabric #DataAgents #NL2SQL #SQL

@tim_kraska and his team have developed impressive NL2SQL technology and it is enabling Bedrock Knowledge Bases for structure data from the Lakehouse, including Redshift Storage, S3 and S3 Tables. #nl2sql #kb #bedrock #structured #data #lakehouse #redshift #s3 #s3tables

Very proud to see Bedrock Knowledge Bases for structured data stored being launched at #reinvent24. Under the hood it uses our novel NL2SQL engine, which we developed over the last year, to make it easy to access structured data as part of Generative AI assistants.

Saturdays mornings used to be dominated by Bert & Ernie, now they are by self-supervised BERT translation of natural language to SQL #NL2SQL Two sets of Papers/Python: #1 #2

573 - Smart answer accuracy on SQL jumped from 78% to 90%. The data agent stopped writing queries in one pass and started exploring schemas and checking filter values first. #microsoftfabric #dataagent #nl2sql #ai

‼️ We are pleased to announce the release of MySQL AI, an option to MySQL Enterprise Edition‼️ Learn more at: #MySQL #HeatWave #NL2SQL #AI #OLTP #cloud #GenAI

Power BI Copilot predicts your revenue. Trinetro calculates it. Same question. Very different answers. → #DeterministicAI #FinanceAI #NL2SQL

first impression of #qwen3 30B-A3B-Instruct-2507 Local analyzing a wide table in my laptop #mcp #nl2sql #sql

Just ask: "check the bearing-temperature trend for turbine 3 and predict 24h failure risk." TimechoCLI turns natural language into validated SQL, queries the DB, calls the LLM, writes results back. #NL2SQL #TimechoCLI #ApacheIoTDB #Database

ADAPT and @Huawei have won the prestigious Emerging Technology Innovation of the Year award at the 2024 Tech Excellence Awards for their groundbreaking Natural Language to SQL (#NL2SQL) translation system. Learn more: #techex24

Check out my latest article: Develop With OCI Real-Time Speech Transcription and Oracle Database NL2SQL/Select AI To Speak With Your Data #oracledatabase #python #nl2sql #selectai #genai #oci #speechai #realtimetranscription

最近在琢磨特定领域NL2SQL的RAG方案应该怎么设计才更合理。还对比了下完全不走RAG、纯靠模型本身能力跑出来的效果。#ai #agent

Research phase done. ✅ Time to get my hands dirty. Goal for tonight: Connect a simple LLM to a local SQLite database and get it to answer "How many users signed up last week?" without hallucinating. Starting small. If this works, I tackle JOINs next. 🕷️ #BuildInPublic #NL2SQL

🛠これまでデータベースからデータを収集する必要があったものも、#自然言語 から SQL へ変換する #NL2SQL を活用することで、誰もが分析できる環境を実現できます。Google Cloud では、実用的な NL2SQL ソリューションを実装するためのツールセットを提供しています。 #gcpja

gpt-oss-120b accuracy on #duckdb goes to top3 in #nl2sql #llm_eval

Deep NL2SQLを搭載したPrivate Agent Factory 26.10を発表

Oracle：AI 代理不能繞過底層權限控管 Oracle 提出 Deep Trust AI Architecture，主張將身分驗證、授權與商業規則下放至資料庫及企業技術堆疊，讓代理人即使透過 MCP 伺服器或 NL2SQL 存取資料，也無法避開底層管控。 治理也延伸到資料與執行環境。AI Data Platform 預計利用既有資料、文件與中繼資料建立可重用的商業脈絡，讓代理人在一致、受控的環境中工作；Fusion Cloud 則保留流程的權限、控制與核准要求，並產生「執行結果憑證」（Outcome Receipt），記錄授權依據、使用證據、決策過程與實際交易明細，供後續稽核。 Oracle 也預告未來治理框架將監控代理人的表現、產出品質、使用量、營運成本與模型漂移，並支援測試、認證及合規管理。 這項主張的關鍵不只是替 AI 增加一道安全閘門，而是把治理點移到「意圖真正變成資料存取與交易行動」的位置。企業若讓代理人處理核心流程，能否追溯每次行動的依據、限制與結果，可能比代理人能做多少事更重要。