The Goal

One of the most quickly growing media companies in Bulgaria approached Ontotext looking for a smart sector-specific soluion that would address the following goals:

• to increase user engagement on their online platforms;
• to have better control over the serving of advertisements and make them more relevant for their users.

The Challenge

Simple as they may sound, addressing both goals required a multi-faceted solution, which incorporated an ad rotation tracking system, integrated with knowledge extraction, sentiment analysis and recommendation services.

The Solution

To respond to the client’s needs, Ontotext implemented a solution, which performs the following tasks:

• semantic annotation of content;
• sentiment analysis;
• user-focused recommendations;
• serving of relevant advertisements.

All components of the solution were customized and tuned to work with content, written in Bulgarian.

Semantic Annotation

The semantic annotation service used a knowledge base,  comprised of Linked Open Data sources.

The service wrapped a set of text analytics components, which perform information extraction over the diverse types of content of the client – news articles, forum posts, user comments, advertisements.

The annotation served two purposes:

• it performed automatic tagging of the articles, written by journalists;
• it provided the fundamental extraction, which was used by the other services in the stack.

Sentiment Analysis

The sentiment service analyzes the comments of the users and aims to provide insights about their attitude towards the mentioned concepts, topics and products – both in the referred content and their own comments.

The sentiment analysis provides important feedback for the ad recommendation service as well. For example, the media publisher didn’t want to show an advertisement for a product, which the client clearly dislikes.

Recommendation Services

A crucial part of the solution are user-focused recommendation services. They provide suggestions for relevant content across the client’s platforms, increasing the engagement with the content across the sites of the media company. In addition, they support serving the most relevant advertisements to their users.

The recommendation services are based on a user profile model, which takes into account:

• the semantic fingerprints of the content a user has read and actively engaged with;
• the content, contributed by other users, which a user actively engaged with (by upvoting/downvoting);
• the advertisements a user clicked on;
• the sentiment, which the users have expressed towards certain entities, referred to in the content they have engaged with.

Ad Serving

As the client wanted to improve their control of the serving of advertisements and their relevance to the users, Ontotext built a custom ad serving platform. It provides the expected ability to register marketing campaigns, manage banner serving zones in the websites, integrated with it, and take care of meeting the rotation targets.

Apart from these standard features, a key differentiator of the system is the ability to target users more relevantly, by utilizing the user-focused recommendation services. By integrating the ad serving with these recommendations, which know about the users’ interests, preferences and sentiments about specific concepts, Ontotext was able to achieve more relevant user targeting.

As a result, the algorithm took into account not only the rotation targets for the advertisement but also calculated the proximity of a user profile to the ad’s content. Based on this semantic analysis, the system could select an advertisement from a number of candidates for a banner zone and match the media readers with the most relevant ads.

Why Choose Ontotext?

Ontotext’s solution for smarter recommendations and ad serving enabled the media publisher to:

• know more about their content as well as their readers, which resulted in an improved recommendation system and boosted user engagement;
• adopt a sophisticated ad serving system that could target a more segmented audience, which resulted in meeting their advertising targets better and increasing their ad revenue.

Do you think this case resembles your particular needs?