Features of Know Narrator Search
1. Application of RAG-based Generative AI to Specialized Tasks
To effectively utilize generative AI in your company’s operations and enhance its business impact, it is essential to reflect industry-, company-, and department-specific knowledge in the AI’s responses.
However, since generative AI answers based on general knowledge learned from internet data (books, articles, web pages, etc.), it cannot provide answers tailored to specialized content or company-specific knowledge.
Know Narrator Search utilizes RAG (Retrieval-Augmented Generation) to supplement the generative AI’s knowledge by inputting the necessary specialized knowledge simultaneously with the question, thereby generating answers tailored to individual tasks. By extracting information necessary for the answer from document data related to the business using a search system and inputting it as a reference document into the generative AI each time an answer is generated, it becomes possible to generate more specialized answers.
2. Three Benefits of Promoting the Business Application of Generative AI
Know Narrator Search supports Multimodal RAG, a technology that directly inputs image data within reference documents into the generative AI. Therefore, it can generate answers by referencing image data such as graphs and photographs. In addition to the text of the reference document, image data is directly input into the generating AI model.
- Since reference documents are managed only within the user’s own Azure environment, there is absolutely no risk of confidential information contained in the text or questions within the reference documents being leaked externally.
- The generation AI uses generation AI models such as ChatGPT provided by Azure OpenAl Service, so there is no risk of your data being used as training data for the generation AI.
- User usage can be understood from user activity logs.
- By using Know Narrator Insight, it is also possible to analyze the effectiveness of implementation and effective use cases for each user and department.
Six Features of Know Narrator Search
Know Narrator Search uses multimodal RAG to generate answers by allowing the AI to understand image data within reference documents without processing it.
By generating answers from data such as graphs and photos commonly used in business documents, it can provide answers that closely resemble those obtained when a human actually reviews the document and answers the question.
- It is also possible to build a hybrid cloud by connecting the internal network and the Azure environment where Know Narrator Search operates via a dedicated line using Azure ExpressRoute. This allows you to restrict access to the internal network and improve security.
You can see which documents the answer is based on and where it is referenced. For multimodal RAG graphs and diagrams, image-level references are also possible.
Applying AITC’s natural language processing technology, which has been developed over many years through the utilization of text data and the business application of language AI models, Know Narrator Search incorporates features to improve the accuracy of reference document searches and control functions for generated AI responses. These features enable higher response accuracy than the raw response accuracy of the generated AI model alone.
