Timeline Probabilities (undergraduate thesis)

The 13 temporal relations on time intervals developed by Allen, and the finite temporality approach to temporal knowledge representation and reasoning are used to build an efficient temporal reasoning framework. The value of this framework lies in its universality, any sequence of events and their relationships in time can be encoded and manipulated to uncover new insights about the nature of events and their interdependencies. Timelines of events are reimagined as stochastic processes, and probabilisitic models are developed to simulate timeline generation. How often timelines are generated by these models point to their probability, the question of how likely is a timeline of events is considered with it being found that one may be much more or less likely than another. A thorough discussion of the model parameters is given, as well as an application of the methods developed on real world events and timelines, taken from the TimeBank corpus of news articles, where temporal information in the texts is annotated.

The full pdf can be found here. I hope to make the python library ftstring FOSS soon.