RAG analyzes complex documents such as tables and figures and generates highly accurate answers without tuning
Quickly incorporate configurable internal knowledge and support business automation with advanced AI-based answers

Fujitsu today announced that it will begin selling its proprietary AI product, "Super RAG™ Package Edition," on Friday, January 24th. This product enables developers to build LLMs that analyze complex documents such as tables and figures and generate highly accurate answers without any tuning.

For details on how to install it, please contact us via the "Super RAG" information request page below. In commemoration of this release, we will also be offering a free trial to allow you to try out the basic functions and UI of the "Super RAG package version" for a short period of time.

Click here to download the "Super RAG" materials.
https://contents.cinnamon.ai/download/wp_superrag_dl_lp

"Super RAG Package Edition" uses unique document analysis technology to perform advanced analysis of unstructured data that is said to account for 80% of data held by companies and is unsuitable for data utilization, enabling optimal structuring according to content type, thereby automating business processes through highly accurate answer generation by LLM. It provides a one-stop service that includes highly accurate searches using a graph database that utilizes the relationships between content units within documents, and connections to various systems and applications with a simple UI and API base. (A function for creating AI assistants using no coding is planned for implementation in March.)

It can be introduced in a wide range of business processes, for example, by easily importing a huge amount of specialized documents such as financial statements, technical information, and internal regulations, and then quickly generating appropriate answers* through LLM, thereby improving business efficiency. It can also be used in tasks that require specialized knowledge, such as responding to inquiries, creating business plans, writing reports, and predicting incidents, as well as in low-volume, high-mix tasks that require tacit knowledge. *The accuracy of answers varies depending on the type of document.

A RAG system that provides advanced support for business automation can be built quickly and at low cost

More and more companies are introducing RAG (Retrieval Augmented Generation) as a way to use LLMs like ChatGPT to automate business processes. RAG is a technology that improves answer accuracy by appropriately combining specific information when LLMs generate answers. Companies often hold proprietary information in the form of complex documents (unstructured data) such as tables and graphs, and an advanced RAG system is required to accurately incorporate this data into LLMs and achieve answer accuracy that can be used in business.

Since March 2024, we have been providing "Super RAG," which enables answer generation and document generation by automatically structuring large amounts of complex unstructured data (sales brochures with charts, administrative procedures, internal regulations, etc.) that are difficult to handle with general RAG, customized for each company. "Super RAG" combines three unique technologies, "document structure analysis," "unique search system," and "prompt optimization," with existing LLM services such as Azure OpenAI Service to generate accurate and advanced answers. By incorporating generation AI that can utilize a huge amount of company data into various business scenes, we will promote essential DX and hyper automation, such as extracting information from complex documents and automatically generating materials.

"Super RAG" solves the problems of general RAG

Systems that use general RAG have issues with chunking that matches documents and use cases, adding metadata that matches the contents, data processing processes, etc., and to improve the accuracy of answers, individual customization (AI tuning work) is required for each document and use case. Therefore, from the perspective of ROI, it is often difficult to introduce RAG and expand the use of LLM business, due to the huge customization costs.

The "Super RAG package version" combines unique AI technology (group of AI modules) to achieve highly versatile data processing (chunking and adding metadata to large amounts of text, including charts and graphs), enabling highly accurate answer generation without tuning. 

"Super RAG" technology examples

　Conventional RAG technology performs vector searches using semantic similarities between texts, but in many cases it ends up searching for unrelated content with similar character strings, which is the cause of reduced accuracy of answers (see left in figure below). "Super RAG" analyzes the document structure, then divides and classifies it into content units, and then structures the relationships between each piece of content in graph form and stores it in a database, thereby improving accuracy by performing searches using the relationships between content (see right in figure below).

Furthermore, when complex documents are directly input into LLM, they are divided into uniform chunk sizes, which means that information unrelated to the prompt is included, reducing the accuracy of answers. The "Super RAG Package Edition" improves answer accuracy and reduces calculation load by optimally dividing documents into chunks (chunking) for each content type.

"Super RAG Package Version" usage fee

The "Super RAG package version" can be deployed in either the Cinnamon AI cloud environment or an external cloud environment (both Microsoft Azure). In addition, we also offer a "verification package" that includes verification of document utilization and various support. For more information, please contact us via the Super RAG document request page on our website.

Click here to download the "Super RAG" materials.
https://contents.cinnamon.ai/download/wp_superrag_dl_lp

Future updates for Super RAG will be announced in press releases and on our website from time to time.

*"Super RAG™" mentioned in this press release is a trademark of Cinnamon Co., Ltd.
*The contents of this press release are subject to change without notice.