Tokyo, Japan — Kotozna Inc. (Head Office: Minato-ku, Tokyo; CEO: Genri Goto), a B2B SaaS company specializing in generative AI–powered multilingual communication platforms, today announced the launch of TocDex RAG (Table of Contents + Index RAG), a new enhancement to the company’s proprietary Retrieval Augmented Generation (RAG) architecture designed to significantly improve response accuracy, multilingual capability, and scalability.
As businesses increasingly rely on large language models (LLMs) to interact with internal knowledge bases and customer information, conventional RAG approaches often struggle with contextual accuracy and multilingual queries. TocDex RAG addresses these challenges through a dual-layer retrieval framework that combines category-based retrieval with index-based vector search.
By narrowing relevant information through structured categories while simultaneously performing precise index-level searches, TocDex RAG enables AI systems to retrieve contextually relevant information more efficiently before generating responses.
■ Difference Between TocDex RAG and Conventional RAG
Conventional RAG systems typically send fixed-length text chunks retrieved through vector search directly to the language model. Because these chunks are often split mechanically without considering semantic meaning, the model may receive incomplete or fragmented context.
In contrast, TocDex RAG takes a different approach. Instead of passing only the retrieved chunk, the system expands the retrieval to include a larger, semantically coherent passage surrounding the retrieved section. This allows the language model to better understand the broader context of the information and generate more accurate responses.
“The quality of retrieved text directly affects the performance of a RAG system. Only meaningful chunks can provide high accuracy in responses, which conventional RAG approaches often overlook,” said Genri Goto, CEO of Kotozna.
■ Key Advantages of TocDex RAG
Compared with conventional RAG implementations, TocDex RAG offers several benefits:
(1) Higher response accuracy through improved contextual understanding
(2) Enhanced multilingual query handling, supporting global communication environments
(3) Cost-efficient processing of large datasets without impacting response time
(4) Minimal operational overhead, requiring no additional maintenance
The architecture is particularly effective for industries requiring precise terminology and multilingual communication, such as tourism, hospitality, and technical documentation, where accurate retrieval of location names, product information, and specialized terminology is critical.
TocDex RAG is now available through Kotozna TPG 2.0, which offers a free plan for individuals and organizations to explore the platform. You can sign up here: https://www.kotozna.com/tpg
For a detailed explanation, Kotozna’s CEO discusses the challenges of conventional RAG and how TocDex RAG addresses them in this video: https://www.youtube.com/watch?v=btxqQlOddEA
About Kotozna Inc.
Official website: https://www.kotozna.com/about
Representative: CEO Genri Goto
Established: October 2016
Capital: 70,000,000 yen (as of December 31, 2025)
Head office location: 3F Hulic JP Akasaka, 2-5-8 Akasaka, Minato-ku, Tokyo 107-0052, Japan
Business: Providing services related to generative AI-powered multilingual communication tools
Contact: [email protected]
Company website: https://www.kotozna.com
