VELDT has developed and commercialized a causal extraction technology using LLM
First in Japan : Official use of the digital medical encyclopedia “MSD Manual” as a data source
Enhanced Causal AI “xCausal™” for more reliable causal model generation

VELDT Inc. (Headquartered in Shibuya-ku, Tokyo, CEO Jin Nonogami, hereinafter referred to as “VELDT”) has released “CKE-LLM (Causal Knowledge Extraction LLM),” a new function that enables the extraction and utilization of known causal relationships using LLM (Large-scale Language Models). CKE-LLM (Causal Knowledge Extraction LLM) will be implemented as one function of xCausal™, a Causal AI (Causal AI) platform for causal reasoning and DX enhancement. This allows for more efficient construction of more reliable causal models. In addition, the MSD Manual(Merck Manual), a digital medical encyclopedia provided by Merck & Co. The MSD Manual, a digital medical encyclopedia provided by Robert M. Davis, Inc. We announce that, for the first time in Japan, we have been officially provided with the MSD Manual, a digital medical encyclopedia provided by Merck & Co. (Headquarters : New Jersey, USA, CEO : Robert M. Davis,) as a data source for causal relationship extraction. We plan to develop capabilities that will be available from the healthcare sector.

VELDT is developing xCausal™, a Causal AI platform that accelerates problem solving, planning and development by using data to quickly estimate causal relationships. Many companies and organizations spend a great deal of time in investigating the causes of problems by making and testing hypotheses based on correlations, and then redoing the hypothesis based on spurious- correlations. xCausal™” enables the rapid execution of trial-and-error to virtually estimate causal relationships through its causal hint variable recommendation technology and easy-to-use causal discovery and causal inference functions. It enables data-driven decision making, shortened R&D cycles, and productive use of highly skilled human resources.

Aiming for faster causal model building with higher confidence, we have developed CKE-LLM (Causal Knowledge Extraction LLM ), which can be set as prior knowledge in “xCausal™. This feature is scheduled to be released in January 2025. xCausal™” is equipped with several "`Causal Discovery"` algorithms that calculate causal structures from data, and users can select an algorithm to generate a causal graph. On the other hand, while the utilization of causal discovery has the advantage of new discoveries, a weakness exists in that the calculated causal structure does not necessarily lead to all correct results, regardless of the type of algorithm. Therefore, it is known that setting up already known causal relationships as prior information before the causal discovery improves the level of confidence. VELDT provides a function for extracting causal relationships using LLM (Large language Models) based on officially obtained reliable information sources, along with a process for human verification of the extracted results and adoption decisions.

Since the reliability of information sources is particularly important when extracting causal knowledge, VELDT is officially provided with the MSD Manual (Merck Manual), a digital medical encyclopedia, as a data source in the healthcare domain. The MSD Manual is a collaborative effort of an independent editorial board consisting of hundreds of medical experts and peer reviewers from around the world, and the editorial staff of MSD Manual. For more than 125 years, it has continued to provide the latest and best thinking on diagnosis and treatment while maintaining complete editorial independence. In order to provide more accurate setting information, VELDT uses the data from the “MSD Manual” after multiple people review and confirm the extracted results and supporting text and information. Since the overall domain is broad, we plan to extract relationships from time to time from symptoms and diseases that are generally considered common, and provide them as a function of “xCausal™ for Healthcare”. As a professional service, we will also provide custom support for companies when using the service.

VELDT will expand this technology to a wide range of industrial applications, not only in the healthcare and will accelerate problem solving and innovation in fields that require causal reasoning. In addition, by building and utilizing causal models as “Causal Assets,” assets of knowledge including tacit knowledge within organizations, we will accelerate the provision of value in various DX applications.

■ About xCausal™ and ”xCausal™ for Healthcare”

URL: https://xcausal.com/en/

xCausal™ is a SaaS-based causal AI platform that emphasizes ease of use, allowing users to start using it simply by uploading data. It has an in-house developed “Smallytics™” algorithm that recommends causalities from a large number of variables, giving you notice of important variables. The tool seamlessly performs “Causal Discovery,” which quickly calculates the causal structure with narrowed variables, and “Causal Inference,” which calculates causal effects among variables based on a structural causal model. The interface and performance are intuitive to use for those with domain knowledge, even if they are not data scientists, and can lead to quick cause identification and action decisions. The intuitive interface and high performance of the system allows non-data scientists with domain knowledge to quickly understand the cause of the problem and determine actions to be taken. The constructed causal models can be improved in accuracy and converted into digital assets as “Causal Assets” for custom use in DX applications such as generated AI and linkage with other systems. xCausal™ for Healthcare” adds healthcare-specific functions, and the CKE-LLM function using the MSD manual as a data source, which was announced this time, will also be provided in this service.

※This tool does not provide medical advice. It is intended for informational purposes only. It is not a substitute for professional medical advice, diagnosis or treatment.

■ About xCausal™ CKE-LLM (Causal Knowledge Extraction LLM)

This function utilizes LLM to extract causal relationships that are already known and set rules as prior knowledge to estimate causal relationships with a high level of confidence. In the healthcare area, we will adopt the MSD Manual, a digital medical dictionary, and in other areas, we will expand the development by limiting it to customer-provided manuals, academic papers, and other data sources with a high degree of trust.

■ About Professional Services

This service provides workshops and data consulting for selecting variables necessary for causal model construction, causal model construction and causal inference on behalf of the user, development of a dedicated environment, development of custom functions, implementation of system integration, and other development and consulting services for use environments that meet the objectives of corporate users.

■ About VELDT Inc.

VELDT is a data science company that develops data analytics technology for the real world and society. While technological innovations such as the internet and AI have brought convenience, modern society faces new social and environmental issues as side effects. Since its inception, VELDT has made “Life Tech Rebalance” its mission, aiming to create a positive spiral in human society and the global environment by innovating not only technological advancements but also the ways in which technology is utilized.