Chapter 5. Contextual AI

The last few years have demonstrated impressive improvements in AI predictive capabilities but with narrow application and sometimes disturbing results. For AI to reach its full potential, we believe it must incorporate wider contextual information from knowledge graphs.

We think of context as the network surrounding a data point of interest that is relevant to a specific AI system. Using knowledge graphs with AI systems is the most effective way to achieve contextual AI, which incorporates neighboring information, is adaptive to different situations, and is explainable to its users.

This chapter explains why AI needs the connected context of knowledge graphs and its benefits for more trustworthy data, higher accuracy, and better reasoning. We explain how this can be achieved with some straightforward knowledge graph patterns and showcase some successful systems where graphs and AI have been combined.

Why AI Needs Context

AI is intended to create systems for making probabilistic decisions, similar to the way humans make decisions. Humans make thousands of decisions every day, often without conscious thought, by matching observed patterns against contextualized experiences.

A person who says, “You saw her duck,” may be asking you directly whether you saw a woman get out of the way of a flying object. But perhaps they are stating that you or someone named Yu (于 in Chinese) saw a friend’s pet bird or are telling them to help butcher poultry for dinner. It’s ...

Get Knowledge Graphs now with the O’Reilly learning platform.

O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.