| 摘要 |
A content intelligence module implementing a unique polarization algorithm may operate to identify products with polarized sentiments in user-generated content. The module may be configured to, at a first computer:
calculate an overall average rating on a product across segments in a dimension;for each segment in the dimension, calculate a segment average rating, a total number of ratings, a segment distance, and a weighted variation score;calculate a total weighted variation score for the dimension;analyze a distribution of overall sentiment variation in the dimension;set a polarization score for the dimension utilizing the total weighted variation score; andrepeat the above steps for each feature, each product, and each dimension to produce polarization scores. Visual representations corresponding to the polarization scores may be presented through a browser application running on a second computer communicatively connected to the first computer over a network connection. |
