World Leading Research applied by World Class Engineers

Quote from Dr. Fergal Monaghan. CTO at Adoreboard

Emotionally Intelligent Insights

We’re pioneering a new approach in data analytics, we call it Emotionally Intelligent Insights. 

The Adoreboard engine delivers a unique range of insights based on years of high-level research and patent pending algorithms. It goes far beyond traditional ‘Sentiment Analysis’ that tends to deliver very simple, binary answers – good/bad; like/dislike; yes/no – or other types of cognitive analysis.

Human emotion is much more complex and all those involved in customer experience know there’s much more to find and understand. For example, rage and sadness would both traditionally be simply classed as ‘negative’, but they are radically different emotions – with different action and responses required.

Our expertise in this area of artificial intelligence is based on a combined experience of over three decades.

We Transform Data Into Business Impact

How it works

Adoreboard’s algorithms analyse text to reflect how people communicate in everyday life. To achieve this we apply an approach known as common sense reasoning. This creates an evolving knowledge graph connecting facts about the world to the emotions they commonly evoke. So feelings such as love, hate, anger, surprise, annoyance and trust can be identified and compared in content.

The emotions that we measure are generally accepted across cultures, but since the language used to express emotions – to “emote” – in text continually changes, we’ve built a system which can be adapted to change e.g. by incorporating popular concepts like emoticons and emojis.

Our technology takes the latest research from psychology, namely that the “message is just a tiny reference to a vast and presumed shared meaning.” Our common sense knowledge graph provides that shared meaning allowing us to pick up hidden but implied emotion in text like “I’m eating my birthday cake at Windrush Cafe” since we infer that birthday cake is ‘commonly’ associated with ecstasy.

As a spin out of Queen’s University, Belfast we benefit from the rich history and contribution to the field of affective computing by the School of Psychology who have influenced a new reference language for describing emotions online known as EmotionML, which is recommended as a standard by the World Wide Web Consortium.

A New Framework for Emotional Analysis

The Adorescore

The Adorescore is a high level performance indicator of how content or a brand is doing on an Index of -100 to +100. A high score means the content is largely positive in nature containing positive emotions such as joy, trust & amazement. A negative score would indicate mostly negative emotions expressed within the content such as rage, loathing and fear.

Emotional Indexes

A simple framework for communicating the position of a brand in relation to an industry average or competitor. We currently provide 4 indexes – Joy, Surprise, Anger and Trust. Each index is a spectrum of emotions, comprised of 6 emotions while still allowing for a brand to appear neutral on each separate index.

24 Emotions

We detect up to 24 emotions from any textual content to discover much deeper emotional context beyond traditional sentiment analysis approaches simple like or dislike. These cover the all key emotions in the accepted model of human emotions known as Plutchik’s wheel of emotions.

Emotional Theme Discovery

Our theme discovery engine acts as a powerful piece of the emotional puzzle, allowing us to make strong correlations around key topics and uncover emotions related to topics around a brand. This surfaces key drivers associated with each of the emotions and provides an understanding of why a brand is performing well or not.

Why we are Different

The analysis performed by Adoreboard is different from other types of analysis. Others like sentiment analysis rely on a one—dimensional approach known as the “valence model”, to identify positive, neutral and negative tones.

Adoreboard takes a common sense approach to understanding hidden meaning in language which reflects how we communicate in everyday conversation. This is layered on a multi-dimensional and multi-emotional knowledge graph to identify the expression of 24 emotions like ecstasy, rage, anger, surprise, annoyance and trust which is summarised into a single metric called the Adorescore. Other approaches which rely on keywords: they presume that the “meaning is in the message”.  They can identify the keyword “loving” as positive in the sentence “I’m loving my birthday cake at Windrush cafe” but are unable to infer meaning from “I’m eating my birthday cake at Windrush cafe.

Human Centric

The Adoreboard Engine understands how people speak in everyday language.

True Emotion

The Adoreboard Engine delivers multi-emotional analysis beyond traditional approaches

Emotional Intent

The Adoreboard Engine interprets the meaning of language to provide an emotion to an action

Science Based

The Adoreboard Engine is built on latest empirically verified theories on emotion applied to brands

Example of Application

The Adoreboard engine does not care if emotional concepts are expressed with 1 word or 5 as we analyse the underlying meaning. For example, the concept of “Christmas” is associated with excitement, joy and ecstasy whereas the concept of “buy Christmas presents” is tinged with annoyance. This semantic analysis extends to understanding emojis and emoticons so it knows the difference between expressions of an Emoji of a happy face happy, Emoji of a happy face laughing, Emoji of a happy face sad and Emoji of a happy face playfulness. This provides actionable recommendations to based on the specific insights linked to individual emotions.

The Adoreboard engine looks for patterns in text, identifies things like negation, modifiers, homonyms, stems, lemmas, parts-of-speech, and what we call “emotes” or expressions of emotion which form part of our analysis.

Our model measures multiple emotions and the intensity of each emotion. This is achieved by linking semantic concepts to emotions. This model is based on the empirically verified ‘Wheel of Emotions’ developed by Robert Plutchik in 2001. This model is used to understand emotion in text. Our semantic technology uses an internal knowledge graph, consisting of emotional concepts and their expression in emotes (words and other symbols) in everyday language. This graph allows us to see which particular emotes contribute weight to which emotions, explaining why a certain phrase has a certain emotional resonance.

Our team of data scientists have published over 30 academic publications and bring together their research at world leading semantic technology institutes such as the Insight Centre at National University of Ireland, Galway, Stanford University, Seoul National University, STI Innsbruck and Queen’s University.

“Analysis by itself is just that: analysis. It can tell you what has happened. Adoreboard makes the connection between concepts in text and the emotions to tell you why something has happened. That’s applied analytics. Data needs to power smarter decisions - emotional analysis that gives you an edge.”

Dr. Luis Trinade Adoreboard Data Scientist