en.Wedoany.com Reported -

Oracle recently unveiled a new technology called Trusted Answer Search, which leverages vector search as an alternative to Large Language Models (LLMs) and Retrieval-Augmented Generation (RAG), aiming to provide enterprises with reliable semantic search results. Tirthankar Lahiri, Senior Vice President of Oracle's Mission-Critical Data and AI Engine, stated that this feature is accessible via download or API. Enterprises need to define a "search space" containing approved documents or application endpoints, configure metadata, and then map users' natural language queries to relevant targets through vector similarity matching.

The Trusted Answer Search system deterministically maps queries to specific documents, extracts parameters, and returns structured results such as reports or URLs. Unlike RAG systems that rely on LLMs, it avoids generating inconsistent responses. A user feedback loop allows tagging inaccurate matches and specifying expected results. Lahiri pointed out that enterprises increasingly require such deterministic systems to ensure compliance and auditability. Independent consultant David Linthicum believes the market potential for this technology lies with enterprises that value predictability, especially in regulated industries like finance and healthcare.

However, adopting Trusted Answer Search requires enterprises to weigh the pros and cons. Robert Kramer, Managing Partner of KramerERP, noted that while it can reduce LLM inference costs, it increases expenditures on data curation, governance, and maintenance. Linthicum also emphasized that enterprises must invest resources in document organization, taxonomy design, and continuous tuning. Scott Bickley, Advisory Research Director at Info-Tech Research Group, warned that keeping data up-to-date is a challenge, and the risk of providing incorrect results rises, particularly as the scale of source data expands.

Oracle's Lahiri responded that the system reduces reliance on static document repositories by treating trusted documents as parameterized URLs, pulling content from real-time data sources such as enterprise applications or APIs. Linthicum remained cautious, believing that in fast-changing fields, maintaining descriptions and mappings still requires rigorous management. Trusted Answer Search positions Oracle to compete with rival products like Amazon Kendra and Azure AI Search. Ashish Chaturvedi, Executive Research Leader at HFS Research, pointed out that the key differentiator is that Oracle's technology does not layer on generative AI capabilities.

Enterprises can evaluate Trusted Answer Search by downloading a package containing vector search, embedding models, and APIs, and running it via API or a built-in GUI application. The package includes two APEX-based applications, one for system management and one for an end-user portal, which aids in integration into existing interfaces.

This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com