Finding the semantic-level precursors on a blog network

In this work, we study semantic-level precedence relationships between participants in a blog network. Our methodology has two steps: a process to identify units of discussion at the semantic level and a probabilistic framework to estimate temporal relationships between blogs, in terms of the order in which they arrive at those units of discussion. We propose dyadic precursor scores that can be used to construct semantic-level precedence networks. From these scores, we derive global precursor and laggard scores. Dyadic precursor scores are compared with URL linking to show that the semantic-level temporal relationships we estimate are an indicator of influence. Global scores are compared to traditional link degree and PageRank metrics, and we uncover relationships between semantic-level temporal behaviour and popularity. We show that our method reveals information about the network that could not be obtained from structural links alone.