John Snow Labs, an award-winning healthcare AI company, has selected Oracle Cloud Infrastructure (OCI) to deploy and scale its new generative AI offering. This new service is based on John Snow Labs’ own healthcare GPT large language model (LLM) and chat platform.
As we stand on the edge of a new era in healthcare innovation, the John Snow Labs team has been diligently working towards this moment for years to help unleash the power of LLMs to enhance the capabilities of medical professionals and researchers. The latest results of this work is a new offering that provides real-time, evidence-based medical information in response to natural language queries. It offers relevant insights into medical questions which are deeply rooted in medical literature. John Snow Labs’ medical chatbot has been designed to facilitate medical conversations, ensuring that healthcare professionals can engage in an interactive dialogue to extract precise information needed for clinical decisions or for building patient cohorts.
This example focuses on dry eye disease, starting with a general inquiry. The chatbot explains what the disease consists of and highlights its symptoms with references from recent literature. When you ask follow-up questions, the LLM can put those into context and provides treatment options for the disease mentioned in the previous interactions.
You can also refer to parts of the previous responses to ask for more details. Here, the query asks for more info on point 1. The model interprets that “point 1” refers to artificial tears and gives a detailed overview of the available types of artificial tears while citing to sources used to compile the answers.
You can also guide the chatbot to provide more specific information in the current context. In this example, a query was made to learn more about high availability-based artificial tears. The answer doesn’t generically describe high availability-based drops but takes into account the context of the conversation and explains how those artificial tears can be used for treating dry eye disease. You can also specify in you question if you’re interested in studies, clinical trials, or other types of documents.
John Snow Labs’ commitment to its users is a tool that not only delivers information but also supports the rich, multifaceted conversations that are the heartbeat of medical practice.
The credibility and provenance of each data source hold paramount importance in healthcare. With this idea in mind, John Snow Labs’ medical chatbot has been engineered to offer more than simple information snippets in response to medical inquiries. It serves as a portal to the underlying research from which its answers are derived. This level of transparency is attained through the chatbot's citation feature.
The chatbot uses a constantly updated knowledge graph for its responses, sourced from an extensive array of medical texts and datasets. The chatbot's replies are grounded in data from this knowledge graph, referencing peer-reviewed articles, clinical databases, or user-specific custom knowledge bases. This approach supports traceability of information and significantly reduces the occurrence of false or misleading information. This functionality not only adds depth and context to the responses but also aligns with best practices of academic research and evidence-based practice.
Furthermore, you have the flexibility to direct your questions to specific knowledge bases, enabling access to information tailored to your unique medical specializations or areas of interest. This specificity ensures the explainability of the responses, enhancing user comprehension of the logic underlying its conclusions. Each piece of information is substantiated with citations, offering users the opportunity to further investigate the sources.
Nowadays, the volume of new medical research expands far faster than healthcare professionals can keep up with. More than a million new papers are published on PubMed every year. Staying up to date with the latest findings is paramount, but the quantity of information is often overwhelming. Here, the medical chatbot comes into play because it’s designed to offer easy access to the latest studies, clinical trials, and medical insights through a natural and intuitive chat interface.
The chatbot's advanced algorithms allow it to examine knowledge coming from extensive databases, journals, and publications and deliver the most relevant and up-to-date research findings. The primary knowledge bases like Pubmed and Medrxiv are refreshed daily, ensuring that you have the latest information at your fingertips.
With natural language processing (NLP) at its core, the chatbot interprets medical queries and offers precise responses. The previous video demonstrates how the tool can transform the way a user interacts with medical literature. By inputting a question in natural language, you can access the most recent studies on the subject and have the chatbot summarize or explain them. The chatbot backs up each response with literature references that you can explore for further information.
In the medical domain, where information is as rich as it is overwhelming, the ability to distill complex content into concise, digestible summaries can prove very convenient. John Snow Labs addresses this challenge head-on, employing advanced tuned models to extract key points from lengthy medical texts, research papers, and clinical guidelines, presenting them in a clear and succinct format.
In the following example, the chatbot takes a dense, multipage document and, within moments, provides a three sentence summary that captures the essence of the material. It can highlight the most pertinent information, so you can quickly grasp the salient points without wading through the entire text.
The summary feature is more than a simple truncation of text. This intelligent synthesis predicts the context, retains critical details, and omits the less essential ones, ensuring that the heart of the content is preserved and made more accessible.
Whether for academic purposes, patient care, or personal edification, this summarization tool is crafted to save time, reduce cognitive load, and enhance understanding. It empowers healthcare professionals to stay informed and focused, freeing up precious time for the human aspect of medicine—patient interaction.
In medicine, knowledge isn’t a one-size-fits-all resource. The specific needs of a practice or research domain require tailored information architectures. Recognizing this fact, the medical chatbot is equipped with features that allow it to analyze repositories of private documents, whether clinical guidelines, research papers, clinical trial data, safety reports, or other institutional records. It then transforms them into custom knowledge bases and applies state-of-the-art NLP to understand private documents and create intelligent knowledge repositories tailored to unique informational ecosystems. By building a custom knowledge base, you gain a tool that directly fits your needs, collects your organization’s medical dialect, and provides rapid, pinpoint access to the information you rely on.
The following video details the steps for creating a custom knowledge base.
First, you define a name for the knowledge base and provide a link to a bucket where the target documents reside and access credentials to the data. A quick preview of the files located in the bucket is shown so that users can check that the provided bucket is the correct one. After this validation, the ingestion process starts. Currently, the chatbot can process text and pdf files.
When the knowledge base is in place, the chatbot allows you to define role-based access permissions with granularity, ensuring that only relevant teams, departments, or research groups within your organization can access each knowledge base, when they need it, under the governance protocols you set. The medical chatbot ensures that sensitive information remains secure, while still being readily accessible to authorized personnel, enhancing collaboration without compromising confidentiality.
In the practice of medicine, each patient has a unique narrative: A confluence of history, symptoms, and treatment responses. Usually, this data is presented in an unstructured natural language format. After creating a custom knowledge base with patient data, the medical chatbot can provide powerful and focused search capabilities. It’s designed to quickly and securely access specific patient information from within your organization's vast data repository to deliver precise responses about your patient’s medical history, treatment plans, or progress notes upon request.
As usual, you can easily access the paragraphs used to compile each answer to get more details about the patients and even go back to the source document for further investigation.
Access to patient-specific information is governed by robust authentication and authorization protocols, ensuring that only authorized personnel can query and receive patient data. The chatbot operates within a framework that prioritizes security and privacy, upholding the highest standards of patient data protection.
John Snow Labs has engineered its chatbot with an advanced security architecture, incorporating rigorous authentication and authorization protocols. At the core of its security framework is a finegrained access control system. This module empowers administrative users to manage permissions and access rights, ensuring that sensitive medical data is exclusively accessible to verified personnel. Administrators have the flexibility to configure access on a granular level by roles, groups, or individual user credentials. This granular approach to data security not only helps protect sensitive information but also customizes each user's interaction with the chatbot based on their specific clearance and professional requirements, enhancing both the security and the practical utility of the information exchanged.
In terms of integration with existing organizational structures, the medical chatbot supports user federation, seamlessly interfacing with lightweight directory access protocol (LDAP) and Active Directory. This integration means that you can directly link user accounts and identities already managed within these systems to the chatbot, streamlining the authentication process and reducing administrative overhead.
The chatbot adheres to standard protocols, such as OpenID Connect and OAuth 2.0, which are globally acknowledged for secure identity management and authorization. This compliance ensures that the chatbot’s authentication methods are both robust and aligned with industry best practices. Additionally, the chatbot introduces social login through identity brokering, permitting user access through familiar LinkedIn credentials, balancing ease of access with stringent security protocols—a critical feature in managing sensitive medical data.
The chatbot is also designed with a focus on scalability and adaptability, available both as a software as a service (SaaS) for individual users and as a single tenant solution for enterprise deployments. The enterprise version runs on a customer’s infrastructure and provides heightened security and control. It has been validated and optimized for deployment in OCI’s global cloud regions and is scalable to support an expanding volume of documents and concurrent users and offers the flexibility to host an unlimited number of custom knowledge bases, users, and groups.
In enterprise implementations, you can customize the chatbot with a unique brand voice and extra safeguards, aligning communication style and security measures with the organization's specific brand identity. Features like single sign-on and application programming interface (API) access are included, granting developers comprehensive interaction with the chatbot’s functions through a robust API, further enhancing the system's utility and security in an enterprise environment.
The chatbot also provides logging and reporting capabilities, providing administrators with detailed insights into user interactions, access patterns, and potential anomalies. This level of transparency and oversight is critical for maintaining high standards of data security and aiding in forensic investigations in the event of a security incident.
Moreover, you can tailor the enterprise version to include specific organizational security policies and protocols, ensuring seamless integration into existing security frameworks. This customizability allows organizations to enforce their unique security standards, from password policies to multifactor authentication, aligning the chatbot's security posture with the organization's overall cybersecurity strategy.
The SaaS version of the medical chatbot is live. The models are tuned and deployed on OCI, offering the following advantages for such a demanding application:
John Snow Labs’ medical chatbot represents a significant leap in medical knowledge management and retrieval. By providing a natural language query interface, a current and comprehensive medical knowledge graph, explained answers from trusted sources, and a new healthcare GPT LLM tuned for medical tasks, it offers differentiated value for healthcare professionals and researchers.
The chatbot's integration of private knowledge bases and patient cohort-building capabilities expands its role in advancing personalized medicine. The enterprise-grade security, privacy, scalability, and operations capabilities enable organizations to safely deploy this platform today.
John Snow Labs’ choice of Oracle Cloud Infrastucture for its infrastructure ensures that the medical chatbot can scale as an efficient, reliable, and secure tool for healthcare and life science organizations.
To learn more, see the following resources:
Dan leads AI Cloud engagements for healthcare and life sciences at Oracle, as well as Oracle Cloud’s partnership with NVIDIA for HCLS. This aligns a life-long love of cutting-edge technology, with a passion for applying that to help transform healthcare and save lives. To discuss using AI/ML & Accelerated Computing (GPU compute) for your healthcare or life science organization, please email firstname.lastname@example.org.
Dia Trambitas is a computer scientist with a rich background in Natural Language Processing. She has a Ph.D. in Semantic Web from the University of Grenoble, France, where she worked on ways of describing spatial and temporal data using OWL ontologies and reasoning based on semantic annotations. She then changed her interest to text processing and data extraction from unstructured documents, a subject she has been working on for the last 10 years. She has a rich experience working with different annotation tools and leading document classification and NER extraction projects.