Graph technology: The missing link to power AI

null

Companies are investing heavily in AI to get ahead of the curve, seeing it as the secret ingredient to enhanced personalisation and a better customer experience. AI is indeed a powerful tool when it comes to engaging customers, but delivering a personalised content experience isn’t as easy as it sounds. 

AI has enormous potential, but it is important to understand it is still very much in its adolescence and still has a way to go before it can grasp social and contextual awareness, for example. AI systems are therefore only as good as the data we feed them to train smart algorithms. AI itself isn’t biased, but any biased deficiencies it is fed will manipulate the results. This is an enormous problem in itself.    

Making connections

The issue is that no matter how much bigger or faster the computers you run AI on, the software today doesn’t possess super intelligence. It has no understanding of context. The big problem is that it is almost impossible to get the right data results when one crucial element is missing – that of relationships.

To turn data into actionable insights you need to understand its relationships. In other words, it needs to be connected. If your team is going to succeed in an area such as personalisation you have to be able to leverage these connections and join the dots between the relationships. This is where graph technology comes into its own, providing a cohesive picture of your big data.

Graph databases were designed to store data relationships as connections. This explicit storage and native record of every aspect of a relationship means fewer disconnects between your evolving schema and your actual database. Graphs captures evidence that links to the strength of these relationships. For example interconnections between people groups and organisations can more accurately predict if you drink alcohol than by turning to other more standard personal and socio-economic factors.

With relationships centre stage, graph databases are also incredibly efficient when it comes to query speeds, even for deep and complex queries, which can speed up the overall AI process. This is important to the? required agility in the digital economy.

In addition, the flexibility of a graph data model allows you to add new nodes and relationships as required. This avoids expensive data migration and all your original data remains intact.

The power of collaborative filtering

At the architectural level, graphs are different from traditional relational databases in that they document connections between data elements. This gives you the power to map relationships that would be complex, or even impossible, in these traditional databases.

This unique capability makes graph technology and AI a perfect marriage. Graphs can provide the context, helping the algorithm and also the trainer to see if there are any gaps. More context means smarter AI.

Personalisation is one area that graph technology is proving invaluable when it comes to the missing link with AI. Following pathfinders such as Amazon and eBay, companies are now looking to provide customers with intelligent, highly context-sensitive prompts to boost customer engagement and the bottom line. These ‘smart’ recommendations, however, can only be made by exploiting technology and building more intelligence into recommendation engines, for example.

Retailers have gargantuan amounts of data they can use, but finding the best paths and content to serve up to customers isn’t easy. When it comes to combining these many data sources into a personalisation engine, traditional relational databases can’t handle the complex recommendation computations in real time. Rather than heavy lifting this data into a centralised system, a graph database enables you to keep the data exactly where it is and simply add a graph analysis overlay.

Personalisation is a burgeoning area for graph technology and we can advance it still further. Coupled with graph software, AI can provide far more accurate predictions and provide a route to smart decision making.

Take the AI based eBay App for Google Assistant, for example, designed to provide a seamless personalised shopping experience and surge way ahead of traditional search techniques. It enables you to check out the prices of products you request and find the best deal, simply by asking.

The eBay App for Google Assistant is a chatbot powered by knowledge graphs that support conversational commerce. The app was developed to bridge the gap between regular search and natural language search. In the app, Natural Language Understanding (NLU) is utilised to break down queries into their component parts. Learning from one type of query can be captured and transferred to other contexts to further enrich the knowledge graph going forward.

Importantly here, it uses a graph database to process all the real-time data connections required. The graph technology helps to refine the search against inventory with context as a way of representing connections inside a retailer’s data sources, based on shopper intent. This lets the system build up its internal profile of the customer. Working with this image it can generate hyper-personal and relevant product suggestions.

Building and storing detailed customer images

A major bonus is that all the contextual data gets stored, so that the eBay App for Google Assistant can recall data for future interactions, providing an excellent example of real-time decision making. When a consumer searches for ‘red shoes’, for example, the app knows exactly what details to ask about next, such as the size, brand or budget. As it collects this data by traversing through the graph database, the app continuously checks it against the inventory to identify product recommendations that match.

Personalisation has come a long way. Hyper-personalisation is the next big step. It will allow online retailers to not only tap into a history of interactions, such as purchasing habits and behavioural traits, but also a shopper’s intent. With the help of AI and graph technology, retailers will be able to identify deep patterns in their data and analyse them to predict real-life outcomes. Finally they will be able to make recommendations in context that consumers actually want – which is exactly what time-poor shoppers are demanding from online shopping. 

Emil Eifrem, CEO, Neo4J
Image Credit: Sergey Nivens / Shutterstock