Greystones Group Greystones Group Greystones Group Greystones Group
  • Home
  • About Us
    • Leadership
    • Board of Advisors
    • Culture
    • Corporate Citizenship
    • Greystones Labs – What We’re Working On
    • Mission Insights
  • Solutions
    • Soleite (Greystones Analytics Platform 2.0)
      • DozerDB
      • KGAP Graph Explorer
      • Secure API Gateways
    • Soleite CoPilot
    • SustainIQ
    • Zero Trust Toolkit
  • Services
    • Data Analytics and Decision Support
    • Cybersecurity and Infrastructure Support
    • Information Technology Solutions
    • Mission Planning and Operational Support
  • Contract Vehicles
    • GSA Multiple Award Schedule (MAS)
    • SBIR
    • Seaport-NxG
      • Quality Assurance
      • Team Members
      • Points of Contact
    • IRS DAIS 2.0 BPA
    • OASIS+
  • Careers
    • Join the Greystones Team
    • Benefits
  • News & Media
  • Contact Us
Greystones Group Greystones Group
  • Home
  • About Us
    • Leadership
    • Board of Advisors
    • Culture
    • Corporate Citizenship
    • Greystones Labs – What We’re Working On
    • Mission Insights
  • Solutions
    • Soleite (Greystones Analytics Platform 2.0)
      • DozerDB
      • KGAP Graph Explorer
      • Secure API Gateways
    • Soleite CoPilot
    • SustainIQ
    • Zero Trust Toolkit
  • Services
    • Data Analytics and Decision Support
    • Cybersecurity and Infrastructure Support
    • Information Technology Solutions
    • Mission Planning and Operational Support
  • Contract Vehicles
    • GSA Multiple Award Schedule (MAS)
    • SBIR
    • Seaport-NxG
      • Quality Assurance
      • Team Members
      • Points of Contact
    • IRS DAIS 2.0 BPA
    • OASIS+
  • Careers
    • Join the Greystones Team
    • Benefits
  • News & Media
  • Contact Us
Nov 28

Leveraging LLMs to Explore Knowledge Graphs

  • November 28, 2023
  • John Mark Suhy
  • Blog

In today’s ever-evolving data landscape, the creation of knowledge graphs has unveiled new opportunities for profound insights and connections. Yet, for individuals without a technical background, navigating these intricate graphs can be a daunting challenge, largely due to the overwhelming presence of technical terminology. This is precisely where the expertise of Artificial Intelligence, particularly Large Language Models (LLMs) like ChatGPT, becomes invaluable.

A knowledge graph serves as a meticulously structured framework for representing knowledge. It comprises nodes, representing entities, and relationships, symbolizing the connections between these entities. Visualize it as an intricate flowchart, where each box signifies an entity, and each arrow denotes a relationship. These knowledge graphs wield immense power in extracting valuable insights from vast datasets. In the AI domain, Large Language Models (LLMs) like GPT4, Llama 2, and others are pivotal.

Trained on extensive text data, Large Language Models (LLMs) boast a vast knowledge base across diverse subjects, enabling them to generate text that mimics human language. This fusion of knowledge graphs and LLMs streamlines information retrieval, empowering AI for research, question-answering, and context-based response generation. LLMs occupy the realm between algorithmic reasoning and human creativity, reminiscent of Monty Python’s whimsy, albeit with a stochastic nature that must be acknowledged. LLMs excel at processing structured data like JSON, employing various tools and APIs to facilitate complex questions posed in human language, effectively translating them into queries or even source code. Notably, the more examples and data provided related to a specific model, the better an LLM can harness this information, significantly improving its accuracy. This technology represents a transformative breakthrough in data analysis and advances our understanding of natural language.

Today’s Scenario

Consider Sarah, a non-technical end-user, seeking information about the universities attended by an individual named Pat. In a traditional scenario without the assistance of a Large Language Model (LLM), Sarah would typically engage with an analyst. She would pose her query to the analyst, who would then proceed to formulate and execute a query (based on the underlying technology where the data is stored) within the knowledge graph to retrieve the relevant information. However, the innovative approach we’re discussing empowers Sarah to interact with the knowledge graph using plain English. She can easily access the information she needs, thanks to the integration of knowledge graphs with Large Language Models (LLMs). This represents a significant step towards user-friendly and efficient information retrieval.

An LLM, by itself, has the potential to generate queries even without the contextual metadata, especially when dealing with well-structured data that follows intelligent naming conventions like “first_name” rather than cryptic designations such as “field_x123.” However, the effectiveness of LLMs in generating relevant queries to address user inquiries can be significantly improved by providing specific prompts and illustrative examples related to the underlying data. A comprehensive grasp of the data models in question can greatly enhance the utility of LLMs.

For example – by simply defining the following information about a “University” along with examples, you can make the LLM more precise by simply providing some natural language explanations and examples that can be used as a prompt for the LLM.

University:   A higher educational institution.   It is also called a College in the United States, but for example in France, College is how they refer to grade school.

 

Attended: Usually means a person attended a university but may not have necessarily graduated with a degree. 

 

Universities are usually referenced by their acronyms such as MIT instead of Massachusetts Institute of Technology.

 

Sarah can now ask intricate questions with ease. For instance, she might inquire about complex relationships, such as “Tell me about all individuals who attended the same university as Pat and their areas of study.” The integration of metadata and templates allows Sarah to access comprehensive information effortlessly, and to reach her end goal of acquiring and understanding the information. The LLM will generate a query specific to the underlying database, and the results will be more likely to be accurate if you provided the LLM with this intrinsic knowledge that exists in simple notes, and conversations.

Conclusion:

Enhancing the effectiveness of an LLM is easily achievable through the provision of comprehensive natural language explanations, illustrative examples, and insightful commentary regarding the dataset at hand. This valuable knowledge often already exists within your documents, Slack conversations, and notes; the key is harnessing this reservoir of information to optimize the LLM’s performance.

  • Facebook
  • Twitter
  • Google+
  • LinkedIn
  • E-Mail

Comments are closed.

Archives

  • May 2025
  • April 2025
  • March 2025
  • February 2025
  • January 2025
  • December 2024
  • October 2024
  • September 2024
  • July 2024
  • June 2024
  • May 2024
  • April 2024
  • March 2024
  • February 2024
  • December 2023
  • November 2023
  • October 2023
  • August 2023
  • July 2023
  • April 2023
  • January 2023
  • July 2021
  • August 2020
  • June 2020
  • March 2020
  • September 2019
  • August 2019
  • June 2019
  • March 2019
  • September 2018
  • November 2017
  • August 2017
  • July 2017
  • April 2017
  • March 2017
  • November 2016

Start your Career with Greystones!

Browse available positions and submit your resume, quick and easy

© Copyright 2024 Greystones Consulting Group, LLC. All Rights Reserved.
  • Privacy Policy
  • Terms and Conditions
  • Contact Us
  • Facebook
  • X
  • Linkedin
Site Design by ACS Creative
  • Home
  • About Us
    • Leadership
    • Board of Advisors
    • Culture
    • Corporate Citizenship
    • Greystones Labs – What We’re Working On
    • Mission Insights
  • Solutions
    • Soleite (Greystones Analytics Platform 2.0)
      • DozerDB
      • KGAP Graph Explorer
      • Secure API Gateways
    • Soleite CoPilot
    • SustainIQ
    • Zero Trust Toolkit
  • Services
    • Data Analytics and Decision Support
    • Cybersecurity and Infrastructure Support
    • Information Technology Solutions
    • Mission Planning and Operational Support
  • Contract Vehicles
    • GSA Multiple Award Schedule (MAS)
    • SBIR
    • Seaport-NxG
      • Quality Assurance
      • Team Members
      • Points of Contact
    • IRS DAIS 2.0 BPA
    • OASIS+
  • Careers
    • Join the Greystones Team
    • Benefits
  • News & Media
  • Contact Us