SALLEE is an entirely new approach to uncovering emotion from language. It quantifies 14 different emotions, including 7 positive emotions, 5 negative emotions, and 2 ambivalent emotions. 

Traditional sentiment analysis offers only a limited understanding of the actual emotions behind language. SALLEE (Syntax-Aware LexicaL Emotion Engine; pronounced “Sally”) is Receptiviti's emotion analysis engine, and it recognizes that emotion is far more nuanced than simply being positive, negative, or neutral. SALLEE detects 14 emotions, strength of emotion and sentiment, and can be used to detect emotions in real-time.

Full Spectrum of Emotions


SALLEE knows that people can experience many emotions at the same time, so it tracks all of them. It keeps track of 14 specific emotions, including 7 positive emotions, 5 negative emotions, and 2 ambivalent emotions. The 14 emotions are then compiled into 6 summary metrics of good feelings, bad feelings, and ambivalent feelings. It also measures the overall degree to which the text sample is emotional vs non-emotional. Finally, a compound sentiment score is available to indicate the net of good and bad emotions.

SALLEE Example.gif

SALLEE Measuring Emotion in Esports

Unparalleled Accuracy


Its impressive accuracy comes from the ability to understand grammatical structure and context clues. It accounts for intensifiers, such as 'very', softeners, such as 'sort of', and negations, such as 'never', but also understands the many different ways people can use the same swear words and idioms based on context. It can tell the difference between 'not really happy' and 'really not happy'. It can recognize 'stepped all the way up to the plate' just as easily as 'stepped up to the plate', and can tell those apart from 'never once stepped up to any plate'. And of course, it understands emojis and hashtags.

Strength of Emotions


The raw scores that SALLEE provides indicate unbounded quantities of emotion. A text sample of 100 sentences might contain 100x more emotion than a text sample with a single sentence. These may be useful if you have data where the source is consistent and bounded, such as Twitter. The scaled scores represent the amount of emotion proportional to the size of the text. These will be best for most users, as they will always fall between 0.0 and 1.0, or between -1.0 and +1.0 in the case of sentiment. For example, the sentence 'Some thrilling content...' produces a raw excitement score of 3.75 and a scaled excitement score of 0.64.

SALLEE measuring the Presidential Debate

For Business Users

Go beyond sentiment to understand the full spectrum of people's emotions, and the strength of their emotions, from their language. Differentiate among positive emotions like love, desire, excitement, and negative emotions like anger, fear, pain and more.

For Data Scientists

Integrate SALLEE emotion detection capabilities into your stack to generate unparalleled insights and predictive capabilities based on the full spectrum of people's emotions.

For Developers

Our easy-to-use API and simple documentation makes it easy to integrate SALLEE Emotion Detection into your technology.

Get started with SALLEE 

Developer Tools