September 2023 Summaries

This week's news from the world of graph databases includes updates on Vector Search, OpenAI, LangChain, and Neo4j, as well as hands-on experiences with geospatial data and digital forensics. The upcoming NODES 2023 conference is just a month away, featuring workshops and sessions on various topics including GenAI, Geospatial Data, and GraphSummit Series. Featured Node speaker Tomaz Bratanic will demonstrate the benefits of using knowledge graphs for chatbots and avoiding hallucinations at the conference. Additionally, there are articles exploring Neo4j's integration with OpenAI vector embedding, Cypher improvements in Neo4j 5, importing geospatial data into Neo4j, and digital forensics analysis using graphs.

The author of the text reflects on a live-stream series called "Going Meta" that explored aspects of Semantics, Ontologies, and Knowledge Graphs. The series, which consists of 20 episodes, was created to provide practical insights and assets for practitioners in this field. The authors aimed to make the content accessible and actionable, rather than theoretical or abstract. They noticed a growing interest in knowledge graph projects that integrated elements from the semantic stack, which is why they focused on these topics. The series has received positive reception, with 40k aggregated views on YouTube and a Github repository containing all the assets used during each episode. A recap of the content was published to organize the episodes into four tracks: Data Engineering, Knowledge Management, Developer/Data Integration, and Applied Semantics. These tracks cover various aspects of working with semantics, ontologies, and knowledge graphs in Neo4j, including data pipelines, ontology management, interoperability/integration, and automated inference.

In a Neo4j Developer Blog post, the author explores the use of Cypher, the query language for the Neo4j graph database, to create a computer player for a five-in-a-row version of Tic Tac Toe, detailing a competition among Neo4j colleagues. The challenge involved writing a single Cypher query to make a move for a player, with participants employing strategies like brute force pattern matching and heuristic searching, including innovative approaches such as using Neo4j's new Quantified Path Patterns feature. The post highlights the simplicity and effectiveness of Cypher in pattern matching compared to traditional programming languages like Java. It also discusses the implementation of a recursive Negamax algorithm in Cypher, which initially seemed impossible, but was achieved by using the database to manage a stack for recursion. The winning strategy in the competition involved recognizing and exploiting the powerful "fork" construct, which allows a player to create multiple simultaneous threats that the opponent cannot block. The post concludes by inviting readers to further explore Cypher and mentions additional resources for learning more about Quantified Path Patterns.

This week's episode of This Week in Neo4j covers various topics including a GitHub repository on working with Bill of Material for Software, Graph Data Modeling and Vector Search capabilities. Additionally, it highlights upcoming events like NODES 2023 and workshops, as well as features on featured speakers, software bills of material, data modeling, leveraging Large Language Models (LLMs) in graph data science pipelines and efficient semantic search over unstructured text in Neo4j. The episode also includes a tweet from GraphTFT showcasing its latest TFT set.

The author of the article discusses the improvements made in Neo4j's Cypher query language, specifically focusing on two new features: COUNT{subquery} and Quantified Path Patterns (QPP). The author presents a case study using these features to detect transaction fraud rings, comparing their performance with an older version of Cypher. The results show that the newer version is significantly faster, even when dealing with complex patterns, such as 100-hop cycles. This improvement is attributed to the better expression and conciseness of Cypher in Neo4j 5, allowing for more efficient pattern matching and filtering. The author concludes that these features make Neo4j 5 a worthwhile upgrade, especially for analytical queries that can be refactored into real-time transactional ones.

Neo4j has achieved full compatibility with PrivateLink, earning AWS Service Ready status. This ensures seamless and secure integration within the AWS ecosystem while enabling private connectivity. AWS PrivateLink enables customers to securely access AWS services without traversing the public internet, offering a host of advantages including data privacy, low latency, and compliance readiness. Securely accessing Neo4j Graph Database through PrivateLink is crucial for protecting sensitive data held by this graph database, which acts as a repository for vital information in data centralization projects. Leveraging PrivateLink facilitates direct and secure connections to Neo4j, minimizing exposure to security threats and cyber-attacks, while promoting seamless data sharing and simplified architecture.

The Neo4j NaLLM project aimed to explore, develop, and showcase practical uses of Large Language Models (LLMs) in conjunction with Neo4j. The project focused on three use cases: creating a natural language interface for a knowledge graph, forming a knowledge graph from unstructured data, and a KG and LLM backed application where the LLM isn’t at the center. The findings revealed that LLMs have strong abilities in generating text and code, but also struggle with understanding deeper meanings and autonomously populating knowledge graphs. However, prompt engineering was found to be an essential factor in effectively using LLMs. Despite challenges, the project offers valuable insights into the potential combination of LLMs and Neo4j, forming a strong basis for future research aimed at further developing and refining this technology.

This week in Neo4j` brings news and updates on various topics related to graph databases, including Vector Search announcements, NODES 2023 preparations, and hands-on workshops. The newsletter also highlights upcoming events, such as live streams, conferences, and meetups, featuring speakers like José Diogo Viana who will be discussing Graph Theory at NODES 2023. Additionally, it covers articles on using Evernote API in a Spring Boot application to retrieve data from Neo4j, building a knowledge graph with Chat-GPT and Python, and creating a smart voice assistant using Google's PaLM 2 and Neo4j. The newsletter also includes a tweet from Rich discussing the CR lore database and invites readers to share their thoughts. `|

Planning a journey on a railway network can be done using algorithms based on Scalable Transfer Patterns or A*; however, brute force approaches are unlikely to work due to their computational complexity. A new Cypher feature called quantified path patterns (QPPs) was introduced in Neo4j version 5.9 and offers several advantages over existing variable-length relationships, including a more concise syntax, the ability to match any number of nodes and relationships, and inline filters that can prune unwanted results as the graph is traversed. Using QPPs, it's possible to find the shortest distance between two stations by anchoring the beginning with the calling point at Denmark Hill and ending with the calling point at Gatwick Airport, filtering out paths that move too far in the wrong direction, and connecting the calling points to specific stop nodes. The query performance is significantly improved when adding filters to eliminate London terminal routes, making QPPs a useful tool for real-time journey planning.

Neo4j is highlighting its focus on graph databases with recent developments in Vector Search, Chatbots, LLMs, and knowledge graphs. The company is seeking participants for a research group, inviting everyone to join regardless of their experience with Neo4j. Various events are coming up, including GraphQL Meetups, conferences, and the GraphSummit Series. Featured speaker Gal Engelberg will discuss using LLMs and knowledge graphs to generate countermeasures against cyberattacks at NODES 2023. Additionally, Morgan Senechal provides a step-by-step guide on integrating Neo4j with Single Sign-On (SSO) authentication for user applications and backend services. Adam Cowley shares his experience in building an educational chatbot using GraphAcademy content, LLMs, and Vector Search. The newsletter also highlights recent tweets from Joao Correia emphasizing the importance of embracing graphs over charts in modern data stacks.