Towards efficient string processing of annotated events

Abstract

This paper explores the use of strings as models to effectively represent event data such as might be found in a document annotated with ISO-TimeML. We describe the translation of such data to strings, as well as a number of operations, such as superposition, which may be used to manipulate these strings in order to infer new information. Some advantages and limitations of the operations are discussed, including issues of over-generation, which can be mitigated though the use of suitable constraints. In particular, we look at how Allen Relations, which might be extracted from a document annotated with ISO-TimeML, can be understood as useful constraints, and translated to strings.