Zusammenfassung:

---

Stochastic models have a long history in IR. For modeling sequences of interactions, different variants of Markov models have been proposed by a number of researchers. Here a user moves stochastically between different model states, which are not directly observable; instead, some signal is emitted along with each transition. For applying these models to interactive retrieval, we aim at modeling search progress at a level that is comparable to cognitive models. This allows for user-oriented analysis of interactive IR, for user guidance and for stochastic simulations of interactive IR. As a second application domain, we regard social media interaction, focusing on rumor detection and veracity in Twitter streams. Experimental results from both domains show that these approaches are well-suited for dealing with real-world data.

Volltext als PDF (Vortrag)