How NLP suits content to sentiment and understanding

Jun 09, 2021 at 09:00 am by admin

A new brand marketing tool using natural language processing is delivering performance increases of up to 40 per cent.

Timmy Bankole, associate director for digital operations at Hong Kong’s South China Morning Post, says the SCMP Signal product is built directly into the CMS and applied both to editorial and advertising.

“It uses three components – sentiment analysis, sensitive keywords tagging, and article readability analysis – to better understand what drives pageviews,” he says in a new INMA Ideas Blog post.

SCMP Signal has evolved brand safety into brand suitability for marketers, providing them with customisable options for contextual, keyword, and sentiment targeting.

The natural language processing project uses data to understand the emotion of the article delivered; whether the article contains specified keywords; and how easily it is understood by the audience. 

The three components work together to provide article profiles, with SCMP Signal combining these innovations with its lexicon-analysis tool, Valence Aware Dictionary for Sentiment Reasoning (VADER).

Its ability to understand what drives pageviews means more content can be generated to expand readership while increasing page session depth and user engagement.

“At SCMP, we believe the application of brand suitability for all direct campaigns to be the ‘new norm’ to address the industry concerns including the evolution of brand safety, whitelisting – which still allows access to approved applications and websites – blacklisting (which blocks unwanted entities), and keyword targeting.

“While these methods have been effective, they can also limit reach,” he says.

Bankole says keywords without context could lead to brands missing out on key audiences for content producers such as SCMP, “for example, an article with the headline ‘New Canon camera shoots multiple coloured images’ could be negatively-flagged because of the keyword ‘shoots’.

“Brand suitability is more about inclusiveness and matching brands’ values to context.”

He expects ad industry changes – including Google’s plans to block third-party cookies on Chrome, and Apple’s coming restrictions on the Identifier for Advertiser that enables advertisers to track data for customised ads – will bring growth in contextual buying to fill in gaps in behavioural targeting.

Publishers are well-positioned to help brands find content for their audiences that builds relevance and engagement, and SCMP Signal is proving its value to advertisers with early results indicating a performance increase of more than 40 per cent in some cases. 
“For example, a health and fitness campaign that ran from May to August 2020 showed that applied sentiment targeting helped to drive a more than 35 per cent increase in performance compared with other strategies.

“The graph above showed us that campaign strategies with sentiment applied drove a significantly higher performance than other lines. The sentiment breakdown gave an insight into the role emotion can – and does – play in digital advertising. 

“We found that articles with a fairly negative score drove higher performance than those with a fairly positive score, which was astonishing to see during these challenging times. We were able to understand and confirm that users’ emotions affected their reactions to content.”

Bankole says campaign findings show brand suitability offers brands a unique opportunity to tap into their audience’s emotions without sacrificing on performance. “SCMP is currently adding engagement metrics and a viewability measurement to help prove campaign return on investment.”


or Register to post a comment