Making bold statements about artificial intelligence (AI) is very much in vogue right now, and it’s often difficult to distinguish the signal from the noise when it comes to the implications of this broad set of technologies. While digital technology has been increasing levels of advertising personalisation for years, the current wave of innovation does have a lot of promise.
At The World Economic Forum held in Davos the CMO of Unilever described AI as a “huge wave” coming down on advertising and marketing: if that’s the case, it’s important to understand what this technology actually is and how it’s being used.
Moving beyond categories
Digital-first readers are fickle, with ever-changing desires and preferences. What this requires from content providers is a more flexible approach, and this can be delivered with more accurate categorisation of content so that content and reader can be more closely matched faster. Companies such as the BBC and the Financial Times have realised this, and are becoming more interested in technologies that allow them to personalise their content.
Publishers are leading the way in this area, and advertisers should follow. The rise of ad blocking shows that readers are finding ads increasingly intrusive, and part of the reason for this is that ads are not targeted correctly. Ads are just content by another name, and so the same rules and technologies that apply to publishers also apply to advertisers.
Advertisers are already using extensive categorisation to help them target audiences better: the Interactive Advertising Bureau (IAB) uses advanced taxonomies to categorise types of advertising segments, such as “barbeques” within “food and drink” or “immigration” within “law and government services” This is a great platform which organisations are expanding and developing, however it’s not quite enough to deliver high-performing ads in the future.
Categories are by their nature imperfect, they are estimations which can lead to inappropriate results: airline ads next to news about a plane crash is the time-honoured example of this. One can amend categorisations to create certain rules avoiding this, however this is costly and unscalable as it requires human intervention each and every time a new case appears. What’s needed is a different approach.
Introducing nuance to advertising with semantics
Semantic technologies enable machines to “understand” written content in a much more human-like way, and for this reason they are able to support much more detailed and robust methods of describing the digital landscape to programmatic ad exchanges.
Semantic technology suites pair natural language processing with graph databases to allow systems to parse through a sentence, identifying entities, and understanding the grammatical relationships between those entities. Crucially, these technologies also allow computers to infer relationships between entities that it hasn’t been explicitly told about yet. For example, the word ‘apple’ can have several different meanings: it can mean the fruit, the singer Fiona Apple, and the electronics company. Using a semantic graph database powered by natural language processing it is possible for a content management system (CMS) to recognise the fact that these entities are distinct and different: in other words, to disambiguate. Taking this further, semantics can also allow a CMS to infer that in a story with the phrase “earnings report” is relevant to Apple the company, whereas an article frequently using the word “nutrition” is relevant to apple the fruit.
This approach becomes very useful in the world of high performance programmatic advertising. Publishers use this technology to provide advertisers with information about what their articles are about on a very granular level: the topics that are discussed, specific people or places, their timeliness and relevance to other articles. By the same token, advertisers would be able to use the technology to tag their ads with search engine-friendly tags to create a more accurate fit between ad and impression. Advertisers would be able to target their ads based not just on an article’s title or tagged keywords, but on the actual specific ideas and topics within that article.
The technologies we’ve been describing are already in wide use for other purposes: Google’s search engine is in part built on these open standards, and if advertisers can align themselves to this kind of infrastructure they’d gain a great deal of additional precision and performance out of their ads.
These technologies are not new or untested: companies like the BBC, the Financial Times and AstraZeneca are already reaping the benefits of sector-specific applications. The flexibility of the underlying technological principles mean that advertisers are beginning to take notice and appreciating the powerful analytics that semantics can provide. Semantics text analysis and graph database technology can be the missing piece that will add some intelligence and nuance to existing advertising technologies such as categorisation vocabularies and taxonomies.