September 2019 Summaries

The ISO/IEC JTC1 Joint Technical Committee has approved a new declarative property graph query language, GQL, after a three-month New Project ballot closed on September 9, 2019. The approval reflects the growing interest in property graphs and prior work done before the project was requested. The GQL project will complement existing work in SC32 WG3 to add a new part to the SQL Standard, SQL/PGQ, defining how to integrate property graph queries within SQL queries and present relational data as property graphs. The standards development process has several steps and ballots that must be completed before GQL officially becomes a standard, with an estimated completion time of 2 or 2.5 years requiring significant effort from multiple people.

This week, the Neo4j community is gearing up for the NODES 2019 conference and previewing talks on Data Science and Graph Algorithms. The Brexit Graph has also been revisited with analysis on average positions of political parties and MPs. Max De Marzi is starting a series of blog posts on building a chat bot using Neo4j, while František Hartman builds a custom analyser for Full Text Search in Czech. Additionally, Graplytic modeling tool is now available as a Graph App, and APOC string matching has been showcased in an analysis of Game of Thrones dataset. The community is also celebrating Danielle Monteiro, this week's featured community member, who lectured at over 80 events in 2018.

GraphConnect 2020 Early Bird pricing is ending soon on October 1, don't forget to sign up and save. Tamika Tannis, a software engineer working on Amundsen at Lyft, shared insights into how Neo4j improved data discovery by significantly reducing time in her recent GraphTour Santa Monica presentation. To learn more about Lyft's journey with Neo4j, check out the blog post featuring Mark Grover, Product Manager at Lyft. For additional great graph videos and updates, subscribe to the Neo4j YouTube channel, updated weekly.

Mark Grover is a product manager at Lyft, working on data projects and is a founder of the open-source data discovery platform Amundsen. The company has grown significantly over the years, resulting in an increasing amount of data that needs to be processed and analyzed to provide better experiences for riders and drivers. However, this growth has also led to issues with data trust and discovery, as users may not know if a particular dataset exists or is still being populated. Grover and his team at Lyft set out to solve this problem by building Amundsen, an open-source platform that uses metadata to provide a searchable interface for discovering and accessing trusted datasets. The platform has been adopted widely within the company, with over 200 scientists using it every week, and has seen significant improvements in productivity. Grover also discusses the future of Amundsen, including its potential use in other applications such as compliance and ETL, and how it aims to provide a personalized experience for users, showcasing their experience and relevant information to improve trust and reduce manual curation.

Neo4j's Intelligent Recommendation Framework is a data model agnostic tool designed to help organizations design and manage their graph-based recommender systems. It includes an admin console for building out recommendation pipelines and exposes a GraphQL API for accessing recommendations, aiming to minimize development efforts while maximizing value from graphs. Various industries are leveraging the framework for interesting use cases such as material management, predicting flight risk, building corporate hierarchy, empowering makers to meet audience demands, and analyzing consumer behavior. The tool uses a score-based approach combining multiple techniques like collaborative filtering, content filtering, business rules, and knowledge-based filtering to build the best-fit recommendations. By utilizing graph technology, organizations can connect all their data and enable powerful suggestions to increase revenues, optimize margins, and delight customers.

The Neo4j Graph Algorithms library has recently undergone significant improvements, including optimizations at multiple layers, improved configuration and usability, and specific feature requests. The updates enable users to compute memory requirements ahead of time, specify different concurrencies for reading data into memory and writing results back to the graph, conduct faster reads and writes, load graphs more efficiently, use smarter information, and terminate algorithms during graph loading and result write. Additionally, the library has seen enhancements in its algorithmic capabilities, including PageRank, Label Propagation, Connected Components (Union Find), and Louvain Modularity, with features such as tolerance parameters, seeded start nodes, and parallel implementations to improve performance and accuracy. These improvements aim to provide better results, increased performance, and enhanced usability for data scientists working with graph algorithms in Neo4j.

This week the Neo4j Developer Relations team is previewing talks at the NODES 2019 conference, including a session on building knowledge graphs and a talk by Kenny Bastani on sentiment analysis with Neo4j and GCP NLP. Stefan Dreverman continues his series on building questionnaires using Neo4j, while Max De Marzi builds a property graph and discusses data import from the Marin County, Florida Property Appraiser's database. The team also highlights various community contributions, including a paper on querying multi-dimensional events and a blog post by Félix Revert on getting data into Neo4j from a spreadsheet.

Neo4j is a graph database that has been gaining popularity in the industry, and its user, Dee Pandian, a Deputy Director of CyForce at Jacobs Engineering, has had a transformative experience working with it after 18 years as a SQL developer. She initially found Neo4j's concept different from her usual work with SQL Server but soon discovered its power through a project involving a large data set. The results showed unexpected relationships between systems, which led to a "wow moment" and a shift in her approach towards graph technology. Since then, she has started using Neo4j for various projects and appreciates its speed, accessibility, and ease of creating relationships. Her experience with Neo4j has been so positive that she now leans towards graph technology and is looking forward to utilizing the platform's new ETL tool instead of her own custom one.

Neo4j, a global company with a diverse community of graph enthusiasts from around the world, is hosting viewing parties for its upcoming online developer conference and summit, NODES 2019. The events will allow local Neo4j and graph enthusiasts to watch the keynote presentation with other like-minded friends in various cities worldwide. With over 52 speakers scheduled, these viewing parties are an opportunity for community members to connect and engage with each other while also participating in interactive chats and contests. The events have already seen a massive success with community members hosting local events to celebrate Global Graph Celebration Day earlier this year.

In the realm of artificial intelligence (AI) and machine learning (ML), graph technology is gaining significant attention as a powerful tool for enhancing autonomous systems. Graph data science applications are expanding into various fields, including financial crimes, drug discovery, customer segmentation, cybersecurity, churn prediction, predictive maintenance, search and master management data, and more. Research indicates that graph networks are bigger than individual ML approaches due to their ability to abstract and generalize structure. To get started with a graph machine learning model, one needs to begin with data sources, move them to Neo4j for persistence, and then write back to the graph. Graph embeddings transform graphs into feature vectors, describing topology, connectivity, and attributes of nodes and edges. Graph neural networks are deep learning models that input a graph, perform computations, and return a graph, enabling new ways of working with data. With these advancements, companies like Neo4j are helping bridge the gap between innovative ideas and technology gaps, making it an exciting time for AI and ML research.

We're excited to announce that we've selected 52 talks from speakers in 10+ countries, including a keynote from Neo4j CEO Emil Eifrem. This year's event features leading experts in graphs, including core engineers and developers at prominent customers like Under Armour and Autodesk. The lineup includes presentations on various topics such as graph algorithms, versioning approaches, and the future of Neo4j drivers. Additionally, we have a range of talks from Neo4j Labs engineers showcasing the latest advancements in the field. Register now to attend this online event and take advantage of the opportunity to learn from industry experts.

The ISO/IEC Joint Technical Committee 1 has approved the Graph Query Language (GQL) project proposal, which aims to develop a new language for querying graph databases. The GQL project draws inspiration from existing languages such as Cypher, Oracle's PGQL, and SQL, with the goal of creating a composable graph query language that can "compose over graphs" and enable users to insert and maintain data in addition to querying it. Ten countries have voted in favor of the proposal, while five abstained due to lack of expertise, and only Japan voted against, citing existing languages already covering the ground. The project is expected to build on work in openCypher Morpheus and the inspiration of G-CORE from the Linked Data Benchmark Council to create a conceptual equal of SQL.

This week at Neo4j, the Developer Relations team is previewing talks on the GRANDstack and GraphQL at the NODES 2019 conference, including a beginner-friendly introduction to the GRANDstack framework. Stefan Dreverman continues his series on building questionnaires using Neo4j, while Jan Zak shares tips on scaling up d3.js graph visualizations. The team also showcases a GRANDstack movies demo app and highlights Betsy Hilliard, this week's featured community member, who is passionate about graphs and helping others advance their knowledge. Additionally, Nelson Sousa presents building a data warehouse with Neo4j, and the podcast "Connecting to Apache Kafka with Neo4j" features David Allen and Michael Hunger discussing integration with Kafka. The team also shares updates on hosting Neo4j on the cloud and celebrates Global Graph Celebration Day.

The Neo4j team is excitedly promoting GraphConnect 2020, a conference to be held in New York City from April 20-22, 2020. The event promises to offer more talks, trainings, and breakout sessions compared to previous years. Early bird pricing is available until October 1, after which the ticket price will increase. The team encourages registration and also shares its YouTube channel for accessing a weekly stream of graph technology content.

Neo4j has taken off as a transactional graph database management platform, with the introduction of Cypher query language allowing expressive and efficient querying. The platform supports graph analytics, including state-of-the-art algorithms and custom optimizations for its graph database model. Neo4j is now venturing into machine learning (ML) and deep learning, aiming to automate both steps of the process: learning about entities and the rules governing their behaviors and interactions with each other. Graph embeddings transform nodes into vectors, preserving topology and connectivity, and can be used as features for classification or clustering. The dominant school of thought in artificial intelligence is deep learning, which automates learning representations through statistical models and learns parameters from past data. Various options are available to implement graph embeddings in Neo4j, including Java-based implementations like Deeplearning4j, a hybrid approach using Jython and JyNI, Py4J for object sharing between Python and Java contexts, GraalVM's universal virtual machine, and Python drivers that read data from a running Neo4j instance. Benchmarking experiments were conducted on two graph datasets, BlogCatalog and Flickr, showing varying performance across different approaches.

A graph database is a complex system represented as a network of nodes and connections called a labeled property graph. The nodes represent entities such as people, products, or orders, while the relationships between them are directional and know their beginning and ending node. A native graph database platform stores and accesses data in its native property graph form to maximize efficiency and performance. Non-native approaches, however, can be flawed with performance, integrity, ease-of-use, and scalability risks due to added processing layers that translate graph queries into relational-table-based storage and processing models used by underlying database technologies. To be a native graph DBMS, a technology must conform to 15 rules, including native storage and modeling, first-class relationships, real-time availability, index-free adjacency, comprehensive data management, discrete management, Cypher support, nonsubversion, ACID transactions, consistent reads, consistent writes, integrity independence, data independence, seamless presentation, and query performance optimization. Despite the rise of graph databases, relational technology is not dead, but rather, it's more suitable for tabular data with static schemas, while graph databases excel in handling highly connected or changing datasets with sub-second response times.

The Neo4j Developer Relations team has returned from a brief absence, with Mark Needham and the team sharing updates on various topics including new releases, tutorials, and upcoming conferences. The team previews the modeling talks at NODES 2019 conference, highlights a featured community member Matt Casters, who has been simplifying data import in Neo4j, and showcases new plugins for Kettle. There are also network analyses of books and video introductions to Spring Data Neo4j RX.

The Neo4j Innovation Lab is a global team that facilitates Applied Innovation workshops with customers and prospects to accelerate their time to validate Neo4j and explore the value proposition of connected data within enterprises. The lab helps companies identify and prototype around use cases to shorten the validation span, bringing together strategy, design, field engineers, and data science expertise to facilitate the process. Design thinking plays a crucial role in these workshops, allowing customers to bring their domain expertise and craft the use case together. Participants have reported that they accomplish an impressive amount of work in just a week, even on projects that they've been thinking about for years. The lab's methodology is well-suited for graph technology, which can help address cultural problems within large enterprises and connect disparate data sources to build better products and services. To find their next use case, companies should look at the adjacent possible by adding new data sources or perspectives to what they're already doing, inviting a natural fit for graphs and a creative space to discover new opportunities.

Neo4j, a native graph database, is well-suited for recommender systems due to its explainability, rapid development capabilities, and ability to support personalization and contextualization. The graph data model is intuitive and easy to understand, making it ideal for non-technical users. Neo4j's Cypher query language allows for pattern matching and traversal of relationships in constant time, enabling real-time recommendations. The database's schemaless nature makes it flexible and adaptable to changing requirements, reducing the need for retraining models. Additionally, graph algorithms such as centrality, community detection, and path finding can be applied to generate recommendations, making Neo4j an exceptional tool for producing real-time recommendations. Neo4j's Intelligent Recommendation Framework, Keymaker, is a data model agnostic tool designed to help organizations design and manage their graph-based recommender systems, minimizing development efforts while maximizing value.

The Global GraphHack has officially started, with over 450 registered hackers participating in the world's largest graph hackathon. The theme is extending the graph ecosystem, and participants are encouraged to build something using or extending Neo4j that benefits others in the community. A combination of community votes and a Neo4j panel of judges will determine the winning teams, with judging criteria including creativity, usefulness, completeness, and quality and clarity of written documentation. Participants must submit their project directly on DevPost, publishing code to GitHub and creating a 1-2 minute video explaining their project. The top three winning teams will receive prizes such as flight and hotel vouchers, free tickets to GraphConnect 2020, and private access to the invite-only Ecosystem Summit.

The growth of enterprise networks and connected devices has made Identity and Access Management (IAM) a top concern for IT organizations worldwide. Traditional directory services are often unable to cope with the complex, ever-changing relationships found in everyday business, such as organizational changes, promotions, mergers, acquisitions, divestitures, supply chains, facilities, and network resources. This can lead to slow query times, impacting enterprise applications and user productivity. In contrast, graph databases can handle complex relationships and provide fast query performance, making them an ideal solution for IAM. Telenor Group's experience with a Neo4j graph database IAM solution demonstrates the benefits of this technology in delivering performance, scalability, and maintainability required by self-service portals, reducing query response times from minutes to milliseconds.