This technology is a topic modeling system that extracts word sets for each document topic using the Latent Dirichlet Allocation (LDA) algorithm, matches them with an emotion word database quantified by the arousal-valence model, and highlights representative emotion words in bold.
Because topic models present a topic only as a set of words, it has been difficult to grasp the meaning without prior knowledge of the document. The same topic could be interpreted differently depending on the reader, and in documents where emotion is key, such as literary works, the ability of word sets to convey meaning was significantly limited.
This technology uses a sentiment filter to categorize emotion words into positive and negative groups, selects the word with the highest emotion score from the dominant group, and displays it in bold to reveal the emotional direction of the topic at a glance. Emotion words are also visualized as circular nodes proportional to their score, increasing consistency in interpretation. It can be applied to literary text analysis and sentiment distribution analysis for social media posts and song lyrics, enhancing the reliability of text mining results.
N/A
