Topic Significance

By

Let d be a document, Zj be a topic in the user interest model, be matched patterns, k = 1,…, nj, to document d, and be the corresponding frequencies of the matched patterns within Zj, the topic
significance Zj of j to d is defined as:

sim(Z_j, d) = \sum_{k=1}^{n_j} { spe(PA^d_{jk}) \times f_{jk} } = \sum_{k=1}^{n_j} |PA^d_{jk}|^{0.5} \times f_{jk}

[1] Y. Gao, Y. Xu, Y. Li, Pattern-Based Topic Models for Information Filtering, in: 2013 IEEE 13th Int. Conf. Data Min. Work., 2013: pp. 921–928. doi:10.1109/ICDMW.2013.30.