“Complexity is empowering”, argues Howard G. Cunningham. Most famous for inventing the first wiki and one of the pioneers of software design patterns and Extreme Programming, he is no stranger to it.
Seen through the three days of Ontotext’s Knowledge Graph Forum (KGF) this year, complexity was not only empowering but key to the growth of knowledge and innovation. The event held the space for presentations, discussions, and one-on-one meetings, where more than 20 partners, 1064 Registrants from 41 countries, spanning across 25 industries came together. It enriched their understanding of the full spectrum of knowledge graph business applications and the technology partner ecosystem needed to turn data into a competitive advantage.
Content and data management solutions based on knowledge graphs are becoming increasingly important across enterprises. So, KGF 2023 proved to be a breath of fresh air for anyone interested in topics like data mesh and data fabric, knowledge graphs, text analysis, large language model (LLM) integrations, retrieval augmented generation (RAG), chatbots, semantic data integration, and ontology building.
Bikes to the moon
The event opened with Atanas Kiryakov who invited us to consider that getting to the moon on a bike is impossible unless you have a really well-tuned bike. “Knowledge graphs are often criticized for being too complex”, said Atanas, “but such initiatives actually are bound to be complex. The question is not how to avoid complexity but how to embrace it and take advantage of it.”
With new business lines, leading to new tools, a lot of diverse and siloed data inevitably enters enterprise systems. For example, one of Ontotext’s clients is a global firm providing financial services in over 50 countries, which has over 5000 different IT systems. Needless to say, competitiveness in such an environment, requires synergies, economies of scale, and cross-selling.
Only data integration through semantic metadata can drive business efficiency as “it’s the glue that turns knowledge graphs into hubs of metadata and content”. This, of course, has its challenges and turning points, but Atanas provided solid advice for those wondering where to start at the end of his presentation.
The question is not how to avoid complexity but how to embrace it and take advantage of it. Share on XHow knowledge graphs power data mesh and data fabric
The second day’s keynote continued the idea of reframing the way we approach knowledge graph construction and enterprise data management. It was presented by Summit Pal, Strategic Technology Director at Ontotext and former Gartner VP Analyst.
Sumit started his talk by laying out the problems in today’s data landscapes. One of the major challenges, he pointed out, was costly and inefficient data integration projects. He further explained what data fabric and data mesh were, how these two related to knowledge graphs, and how they provided solutions that helped enterprises fight infobesity.
Another term Sumit introduced was datastrophe. All these terms show the huge chasm enterprises need to cross if they want to stop drowning in data while thirsting for context. As he put it, “We’ve allowed data to lose consistency and precision of meaning”.
To back this observation Sumit also reminded us of a quote by sir Tim Berners-Lee that is little known, yet very relevant to today’s enterprise scenarios:
Any enterprise CEO really ought to be able to ask a question that involves connecting data across the organisation. Most organisations are missing this ability to connect all the data together. (from Q&A with Tim Berners-Lee)
Finally, Sumit highlighted the importance of knowledge graphs to advance semantic data architecture models that allow unified data access and empower flexible data integration. If we want to understand and describe how things work and relate, unambiguously exchange data, or reuse it, we need a semantic data model.
Abstract art and knowledge graphs: embracing your mess!
Lance Paine, founder of Semantic Partners, took the ball from Sumit to talk about “why you’re not ready for knowledge graphs”.
Lance introduced himself as an ”engineer who avoided databases at all cost before discovering SPARQL”. He outlined the challenges of working effectively with AI and machine learning, where knowledge graphs are a differentiator. In an engaging narrative built on the premise that most organizations are not ready for a knowledge graph, Lance talked about the usual pitfalls when building such a solution. According to him, “failing to ensure data quality in capturing and structuring knowledge, turns any knowledge graph into a piece of abstract art”.
These ideas were further addressed at the panel “How can organizations streamline and speed up knowledge graph implementation of enterprise semantic applications?”. It was hosted by Ashleigh Faith, Founder at IsA DataThing, and featured James Buonocore, Business Consultant at EPAM, Lance Paine, and Gregory De Backer CEO at Cognizone. The panel offered an interesting discussion about strategies for overcoming these challenges, avoiding the pitfalls, and applying the best practices in semantic knowledge graph building.
The rise of knowledge graphs that power marketing and sales processes
Call me biased, but I see tangible progress for knowledge graphs powering marketing and sales. The marketing thread in the knowledge graph narrative is now going out of the academic approach (take for example Anna Fensel and team’s Knowledge Graphs for Online Marketing and Sales of Touristic Services), into practical application for enterprises.
Three presentations at the KGF 2023 proved it.
The first one was by François Scharffe, CEO at The Data Chefs, and one of the people behind The Knowledge Graph Conference. François shared the story of (and the methodology for) a boutique knowledge graph. He presented the case of building several production applications for a large global investment bank. It has helped hundreds of bankers and analysts to improve their intelligence capability and to better follow and target potential clients.
Aurelije Zovko CEO at Zenia Graph, presented another example of a knowledge graph humming under optimized sales processes in his talk “Leveraging the Power of Knowledge Graphs, LLMs, and ML to Turbocharge Sales Growth”. He shared their approach to knowledge graph building and architecture. It aims to bring tangible benefits to sales and marketing operations, locate new and relevant customer accounts, and save time and resources. In a live demo, Aurelije provided a sneak peek into how LLMs and machine learning were used to enhance their lead generation process.
Last but not least was Ontotext’s Solution Architect Krasimira Bozhanova who took the audience behind the scenes of the Ontotext Knowledge Graph (OTKG) project. In her presentation “Unveiling the value of the Ontotext Knowledge Graph and LLMs in Marketing”, she discussed the process of constructing our own knowledge graph and highlighted the pivotal moment involving the LLM that occurred during the project.
Krasimira touched upon the ways knowledge graphs can harness unstructured data and enhance it with semantic metadata. She also shared the architecture behind the vision of building useful semantic search, valuable insights platform, and powerful knowledge-discovery/” target=”_blank” rel=”noopener”>knowledge discovery environment. She showed how our knowledge graph works in concert with other, natural language processing tools and LLM integrations.
LLMs are all around
With LLMs all over the place since the launch of ChatGPT and other generative AI tools, available to the public, the LLM thread in KGF 2023 was inevitable. Apart from the presentations we already mentioned, several other talks provided participants with insights on approaches and best practices when searching for a sweet spot between LLMs and knowledge graphs.
Without diving into too many details, here’s a list of presentations at the intersection of LLMs and knowledge graphs.
Product Portfolio and Roadmap updates
Presenting Ontotext’s portfolio and roadmap updates, Vassil Momtchev, CTO at Ontotext, spoke about GraphDB’s integration with ChatGPT. It allows hooking your knowledge graph to an LLM and thus enabling a variety of applications. Vassil also shared that semantic knowledge graphs address key challenges of LLMs such as lack of common domain knowledge, herd incremental updates, and costly production processes.
Generative AI and personalization at scale
Seth Earley, CEO at Earley Information Science, talked about “Generative AI and Personalization at Scale: How Knowledge Graphs Enable Context and Eliminate Hallucinations of Large Language Model Powered Applications”. He explained the foundational role of information architecture (IA) and knowledge graphs for AI applications. Seth has coined the phrase “There’s No AI without IA”, so it was no surprise that he outlined the need for IA and knowledge graphs to power AI systems. He also pointed out that knowledge architectures can improve precision and recall for Generative AI, which would typically hallucinate.
Unleashing conversational excellence
Alex Walher, CEO at Onlim, presented “Unleashing Conversational Excellence: Harnessing the Power of Knowledge Graphs and Large Language Models for the Perfect Match”. He dived into the big topic of LLMs and knowledge graphs through the even bigger perspective of conversational experiences. Inviting us into the realm of conversational literacy, Alex argued that when seeking to enable smarter, engaging, and meaningful conversations, the capabilities of knowledge graphs and LLMs are the perfect combination.
Graph building with text analytics and LLM
In this talk, Ivaylo Kabakov Head of Solutions at Ontotext, and Borislav Ankov, Metadata Studio Product Owner at Ontotext, showed how knowledge graph development can be made easier, faster, and cheaper with the help of text analysis tools and LLMs technology. Starting with how text analytics and knowledge graphs complement each other, Ivaylo and Borislav further elaborated on the scenarios where LLMs can be useful. Some examples they gave covered Natural Language Querying, Zero-shot approaches for information extraction, synthetic training data for machine learning development, and many more).
What opportunities does LLM technology create for data and business strategies?
This was a panel, led by Doug Kimball, Chief Marketing Officer at Ontotext, Lauren Hawker Zafer, Product Office Team Lead & Head of Training and Education at Squirro, Alexander Wahler, CEO at Onlim, and Vassil Momtchev, CTO at Ontotext. The discussion revolved around how LLM technology impacts semantic technology vendors, consultants, and customers and the threats and opportunities such technology brings to the market.
A consistent theme among the participants was the observation that we were to expect a hybrid future where LLMS complemented knowledge graphs. For while LLMs have changed the data-driven business landscape, they still need to be more accurate and are only useful if combined with the right graph technology. Participants also agreed that we were on the verge of a tectonic change in the way we accessed and interacted with information.
The future of digital transformation is graph-like
Given the overarching complexity of today’s enterprise data landscapes, it is not surprising that many C-level and senior executives are forced to think strategically about graph technologies and how they fit the bigger picture of enterprise data architectures and more importantly the business processes these architectures power.
It was this strategic approach that Brandon Richards, General Manager – APAC at Ontotext, laid out in his keynote. He talked about how many organizations waste valuable resources on data and technology systems that only add to the piles of siloed data and called this phenomenon “bad data tax”.
Brandon’s presentation highlighted the semantic data layer as the best way to address this technical debt because connecting, sharing and reusing data across an organization brings tremendous savings opportunities. It also introduced Ontotext’s Knowledge Graph Center of Excellence, which aims to help organizations understand the bigger picture of the graph data landscape and optimize their business outcomes with an informed approach to graph technologies.
Overall KGF 2023 was a horn of plenty, giving practical steps, actionable advice, and many enterprise-grade approaches to using knowledge graphs and to turning complexity from a challenge into an advantage.
Want to learn more?
Share this article
