Introduction of EXPLORE

Introduction

In today's digital age, search engines have become indispensable tools for information acquisition. Compared to traditional search engines, AI-powered search engines can automatically read search queries and directly output target results, significantly reducing users' reading burden and greatly improving efficiency.

In the medical field, where specialized knowledge is abundant, there is an urgent need to leverage AI search engines to enhance the overall efficiency of the medical industry.

EXPLORE——AI-driven Medical Research Assistant

To achieve the goal above, NOAH AI has developed EXPLORE, an AI-powered medical research assistant. Powered by a large language model, EXPLORE connects to biomedical databases and is reinforced by a biomedical knowledge base using Retrieval-Augmented Generation (RAG) to enhance its expertise in the biomedical domain.

EXPLORE not only provides users with an independent AI-powered professional medical search experience but also supports features such as DISCOVER and WORKFLOW, expanding the comprehensive capabilities of AI search and analysis.

EXPLORE is significantly different from traditional search engines, offering four core features including One-Click Answers, Transparent and Traceable Process, Authoritative Medical Data and Deep Dive Capability.

Key Feature Highlights of EXPLORE

• One-Click Answers: AI-powered search delivers instant answers by intelligently scanning the web and extracting key insights, dramatically boosting efficiency.
• Transparent and Traceable Process: Real-time task breakdown and execution, along with clear source citations, address the trust issues associated with AI hallucinations commonly seen in LLMs
• Authoritative Medical Data: Sourcing information from authoritative medical websites and databases ensures the professionalism of medical content.
• Deep Dive Capability: Ask follow-up questions to delve deeper into the background information related to search results.

Currently, EXPLORE supports the "Health Search" function to meet the search for common medical-related questions in daily life, such as the following questions:

• What are the different types of depression?
• What is bipolar disorder?
• How is bipolar disorder clinically diagnosed?
• What are the common treatments for bipolar disorder?
• What are some standardized clinical assessment tools for bipolar disorder?
• What are the current limitations of depression medications?

User Guide and Feature Highlights

In the following section, we will illustrate the complete user workflow and key features through a case study with visuals and descriptions.

First, let’s enter a specific question, such as “What’s the current therapy for depression?”, into the AI search box.

Next, EXPLORE will begin to work. We see observe the real-time breakdown, search, and analysis processes for their queries in the "Thought Process" interface.

Later, we will be directed to the results page, which includes multiple sections such as references, thought process, answer body, recommended follow-up questions (for further search), and a search box.

After the search is completed, there will be recommended questions at the bottom to facilitate us to ask follow-up questions, or we can click on the answers to the collapsed sub-questions to read, and sub-questions results usually have more detailed information.

If we want more in-depth information, the original source material is readily available. Clicking on the citation card will open a brief description of the source article in the right-hand panel, providing a direct link to the full text.

Above is a brief overview of the workflow and page structure. We will now delve into the specific features and highlights of EXPLORE."

Highlight 1- One-Click Answers

Traditional search requires users to browse through multiple web pages and compile information. This process is both time-consuming and labor-intensive. In contrast, EXPLORE provides a highly organized and summarized answer directly, offering a "one-click" solution.

This is made possible by AI technology that automatically filters, analyzes, and integrates information in the background, delivering a concise final answer to the user, significantly improving efficiency.

Highlight 2-Transparent and Traceable Process

After reading the main answer, users can click on "Thought Process" (as shown below) to view a detailed breakdown of how the query was decomposed into sub-questions. This allows EXPLORE to provide both the convenience of a one-click answer and the flexibility to satisfy users' need for in-depth information.

General large language models are prone to "hallucinations," leading to unreliable content and decreased user trust. This is especially concerning in the medical field. EXPLORE addresses this issue by providing clear references and source attribution, ensuring that information sources are transparent and traceable, thus enhancing content reliability.

EXPLORE displays all source information, including article titles and URLs, in a user-friendly card format at the top of the results page. With just one click, users can access the original source. As shown in the image, clicking the first card will lead you straight to the American Family Physician article.

Beyond displaying overall references, EXPLORE introduces a content traceability design. You can pinpoint specific content through citation-like markers, such as the superscript "2" in the figure, which corresponds to the second article in the reference list, further enhancing the credibility and traceability of the information.

Highlight 3- Authoritative Medical Data

Since AI search results are heavily influenced by the quality and relevance of their source data, it is crucial to rely on professional and authoritative sources for specialized inquiries.

To ensure the accuracy and reliability of medical searches, EXPLORE's Health mode integrates data from reputable medical institutions such as the Centers for Disease Control and Prevention, the National Institutes of Health, and the Mayo Clinic.

Examples of authoritative medical data sources used in EXPLORE search：

• The Centers for Disease Control and Prevention (CDC), The National Institutes of Health (NIH) , Mayo Clinic
• Authoritative medical websites: The Centers for Disease Control and Prevention (CDC) , The National Institutes of Health (NIH)
• Clinic Professional biomedical databases : ClinicalTrials.gov (operated by the U.S. National Library of Medicine) , The Chinese Clinical Trial Registry
• Medical clinical guidelines such as those published by the National Comprehensive Cancer Network (NCCN)
• dditional data sources will be added in future versions. Stay tuned!

Highlight 4-Deep Dive Capability

Whether it's about general knowledge or medical articles, users often have a series of related questions after reading, such as wanting to delve deeper into a certain part of the content or explore other aspects of the topic. Typically, users can only rely on search engines for further searching and reading, which is a rather cumbersome process.

EXPLORE now effectively addresses this issue by allowing users to "dig deeper" within a single interaction. EXPLORE has placed a search box at the end of each article, enabling you to continuously ask follow-up questions simply by typing them in. This seamlessly integrates reading and question-based search.

Since the way a question is phrased can significantly impact the quality of the results, EXPLORE provides users with some recommended questions. Users can simply click on a question to ask it directly (see the example above) or they can enter their own custom question into the search box.

User Suggestions for EXPLORE

To enhance your EXPLORE experience, we recommend the following:

• Use English for better results: As large language models are more mature in English, we recommend using English for your queries and responses.
• Be specific with your questions: Unlike traditional search engines, the level of detail in your query directly influences the level of detail in the response. Please be as specific as possible.
• Break down complex questions: For complex queries, please break them down into smaller, more specific questions. This will help ensure a more focused and informative response.

Go to [EXPLORE] for an in-depth exprience.