With emotion keywords abundant, emotions are difficult to understand as it is. We take it a step further. Emotion clustering takes thousands of emotion expressions and uses an algorithm called DELSAR (Document Emotion Latent Semantic Analysis Reducer) to understand what humans understand each emotion to mean, based on massive lexical co-occurrence matrices. It turns out that emotion keywords are not a discrete as you might think.
DELSAR, an algorithm based on Latent Semantic Analysis, shows us that a specific emotion keyword consists of a particular amount of all other emotion keywords. Its primary purpose was to discern the semantic distinctiveness of a language — namely the set of basic emotion keywords — and to reduce this set to the N most semantically distinctive keywords; this means that the language surrounding these keywords are the least similar to each other than any other keyword (any other keyword would be more similar to one of the keywords within N). As you can imagine, this algorithm has many applications both with respect to emotion and to language in general.
When combined with Emotional Relativity, you can begin to appreciate the scale of complexity inherent in human emotion, which is not surprising as consciousness itself is relatively complex. Each emotion experienced by a human is much more subtle than a single keyword.
Our current research includes clustering methods based on the Extended Fiedler Method on decomposed matrices.