strawberry
Announcement of start of demonstration experiment of generative AI tool using RAG by Collinia Co., Ltd. and Yokosuka City
……
Our company is a sustainable infrastructure company that uses real estate as infrastructure for daily life, connecting people and businesses, and supporting people’s enriched lives. As a sustainable infrastructure company, we aim to further expand our business advantages through strategic business development.
Our group company Kolinia Co., Ltd. (hereinafter referred to as “Colinia”) is promoting DX using generative AI, and in collaboration with the Digital Government Promotion Office in Yokosuka City, Kanagawa Prefecture, we have developed a large-scale language model. We would like to inform you that we have started a demonstration experiment using a technology (RAG) that combines text generation using (LLM) with external information search (RAG).
This project aims to reduce the burden on employees in contract conclusion work, which requires reference to numerous manuals and regulations.
1. Background and purpose of this project
Yokosuka City is the first municipality in the country to implement advanced initiatives such as DX conversion and the use of generative AI, and is the first municipality in the country to start using the generative AI service “ChatGPT” across the city. I’m here. While considering ways to further improve the efficiency of internal affairs, we found that when employees carry out their duties, they have to choose from a large number of manuals and regulations, and spend a lot of time searching for the target section. It has become clear that there is a need for In order to solve this issue, we have begun research into a mechanism that uses AI to generate answers based on internal business documents, etc., targeting “contract conclusion work” in particular, which requires referencing many manuals and rules. In collaboration with Yokosuka City, we have started this project to support the development of AI and the use of RAG (*) in line with on-site work, with the aim of reducing the burden on staff. (Project period: until the end of September 2024)
(*) RAG (Retrieval-Augmented Generation) is a technology that combines text generation using a large language model (LLM) with external information retrieval. By using this technology, unique information is cited in response to user questions, making it easier to generate answers that comply with the unique rules of local governments, companies, etc.
2. Overview of this project
In this project, we first sought to gain a solid understanding of the work and find an appropriate approach based on the need to obtain results that were more in line with the work within the agency. In order to provide a practical and effective method for the field, we conduct interviews with the Digital Government Promotion Office and the person in charge of the field, including the current inquiry content and next actions after obtaining the results, to clarify the work content and business requirements. The resolution was increased. We then decided to adopt an approach that combines technology to reference and search documents that regulate internal contract work, such as contract manuals and regulations, and obtain answers from a generative AI service.
Furthermore, in interactions between humans, necessary information is derived by reading between the lines and asking questions as appropriate, but it is not easy for machines that cannot read between the lines to gather sufficient information. In preparation for cases where the information necessary for the user to obtain the answer he or she is looking for is not made clear, and questions are asked at a high level of abstraction, or questions where it is unclear what kind of answer the user is looking for. , we will also incorporate a step to improve the resolution of questions through dialogue, allowing us to provide detailed responses not found in existing packaged services.
[Image: https://prtimes.jp/i/36834/200/resize/d36834-200-fc1b96cef34ad66953a5-0.jpg&s3=36834-200-86ba334b4e313ab4945ee681d4598f42-3900×2194.jpg] ■ About Collinia Co., Ltd.
Coligner implements optimal solutions to help companies transform their business. In the field of generative AI utilization, we define business requirements by deepening our practical understanding and build models that fit business challenges. In addition, we will provide detailed support not found in existing packaged services, such as incorporating steps to extract the information necessary to obtain the answers the user is looking for through dialogue.
Collinia homepage https://clnr.inc/
3. Future outlook
Although the impact of this initiative on our consolidated business results for the fiscal year ending February 2025 will be minor, we will leverage the synergistic effect with Coliniar to provide solutions that utilize DX in a variety of areas beyond the framework of the real estate business. We will create new profit opportunities and increase shareholder value.
More details about this release:
https://prtimes.jp/main/html/rd/p/000000200.000036834.html