Ian Saunders is Managing Director at Crystal Semantics. Here he discusses the company's offering in Europe, the recent deal with AppNexus and the importance of semantic targeting in a post-directive world.

For those not familiar semantic targeting, can you give a quick overview of the technology and why buyers would use it?

IS: Semantic targeting offers buyers an ability to identify the complete environment in which an ad placement could be made. For each and every page analysed, the technology can identify the language used on the page, the subjects identified on the page, the genre of the content (blog, forum shopping etc) and detect the presence of content that would compromise the brand safety of the advertiser . This entire process takes place in a matter of a few milliseconds and is currently being applied to billions of pages and impressions internationally. Buyers can use the system to place ads only on relevant content, avoid being compromised by brand safety issues and in a real time bidding environment, making a more informed bidding decision.

How does semantic targeting work? How does it differ from contextual?

IS: Semantic targeting looks beyond contexual’s simplistic analysis of the presence of individual keywords and instead, examines every word present on the page in order to determine the true context of the page. Only through such a thorough analysis can a correct categorization take place. To illustrate the scale of the problem that contextual targeting cannot resolve, it is a fact that for each word in the English Language there are an average of 2.7 alternative meanings, with poor keyword selection leading to misplaced and irrelevant ads. A key aspect of semantic targeting is to being able to determine the senses and meanings of words to ensure that ads are not assigned due to an incorrect assumption that a word does not have more than one possible sense In Crystal Semantics’ case, all the words and phrases that would be used in relation to a given subject have been scored for their distinctiveness and assigned to our hierarchical framework of categories. It is an enormous task and took a team of over 40 expert linguists over 10 years to construct. Therefore, when a page is analysed, the technology compares all words on the page with the words in our taxonomy and is able to determine the most relevant categories that depict the content.

Will semantic targeting become even more important in the display market once the directive on privacy comes into law – and curtails some of the excesses of third party cookie tracking?

IS: Undoubtedly the directive and its potential disruption to the display market has instigated research into alternative targeting methods. As semantic targeting is not cookie based and is non-invasive, it is totally compliant with the directive and is therefore, garnering a great deal of interest. Whatever the end result from legislation, targeting is data dependent and context is a reliable indication of interest. Semantic targeting is already playing a part as marketers come to terms with the legislation, either as a solus solution or as part of a hybrid semantic/behavioral targeting system, enhancing first party data or building profiles to enable predictive analysis.

How is the semantic data passed and interpreted in a real-time ad call? Is it just another variable you bid on?

Semantic data is essentially page level intelligence with categories and brand safety data being made available as further data points, operating in conjunction with other targeting variables. Clients can access the data as and when a real time ad call is made and then can therefore make an informed purchase decision. A bid can then be modified depending on the relevance to the campaign criteria and clients can avoid wasteage by only bidding on impressions matching their pre-determined category and brand safety criteria.

Aren’t semantic targeters and ad verification specialists offering similar solutions?

IS: Similar but different. The aim of both semantic targeting and ad verification solutions is to identify content that could compromise the brand values of the advertiser. However our approach is to identify such content and to offer pre-emptive information in order to deter the placement of ads alongside such content, pre- delivery, rather than highlighting the occurrence of compromising placements, post activity. We see our capability as more of a carrot rather than a stick approach.

How are you differentiating yourself from the likes of Peer39 and others working in this area?

IS: Our approaches are considerably different. Our patented proposition is linguistics based, leveraging one of the largest language engineering projects ever undertaken, combined with proprietary technology to analyse the core language and its usage on the page. The assignment of words to categories has been made by and is maintained by an in-house team of linguists, operating in the various languages we cover and carrying out daily updates to add new words and concepts. Our competitors are entirely technology based, leveraging algorithms based upon the statistical probability of words and their occurrences to determine context and requiring substantial volume of webpages to train the algorithm to identify the senses and meanings of words. Perhaps our 10 years of human led linguistic development puts our method beyond the ROI needs of most of our VC funded competitors !

Can you give some insight into your new partnership with AppNexus – and what it means for European exchange buyers?

IS: Appnexus are one of the key innovators in the new ecosystem and we have collaborated to integrate our technology into their platform, leveraging both their cloud infrastructure as well as their core real time bidding technology. This enables us to scale up to meet the growing demand for multilingual targeting of content in the European market and to be tied to all the major sources of inventory. Our technology will therefore see the vast percentage of European traded inventory and we will be able to provide our data to Appnexus’ European clients. Our deployment means that buyers can access our functionality if they use Appnexus’ own Displaywords bidder or have invested into developing their own bidding technology.

Can the tech work across the different languages in the European market?

IS: We have made substantial investment into developing language specific versions of our functionality with languages such as English, German, Dutch, Danish, French, Italian, Swedish rolled out and Spanish, Norwegian, Finnish & Portuguese being rolled out shortly. Our language detection technology automatically identifies the language of the page and categorises the content in the respective language.

Are we likely to see more integrations with other DSPs and ad servers over the coming months?

IS: Certainly, we are currently working on multiple integrations on both the buy and sell sides in the US & Europe as well as projects with other technology vendors in the likes of ad serving, analytics etc. and emerging areas such as iTV & mobile. Every day seems to bring new opportunities and open up new markets.