Strings for Temporal Annotation and Semantic Representation of Events
Citation:Woods, David Adam, Strings for Temporal Annotation and Semantic Representation of Events, Trinity College Dublin.School of Computer Science & Statistics, 2022
DavidAdamWoods_Thesis.pdf (PDF) 1.426Mb
This work describes the use of strings as models for the representation of temporal data---that is, events and times, and their linear ordering and temporal inter-relations---to form the basis of a framework for reasoning about that data and using it to aid in the creation or validation of semantic temporal annotation. Some of the relevant motivating literature is examined, in particular Allen's interval algebra and relation set and the TimeML annotation schema. The finite-state temporality approach to semantics wherein the string framework originated is also detailed, and a breakdown is given of the work done to develop and flesh out the framework, including discussion on the various operations for manipulating and reasoning with the data. In particular, various flavours of a superposition operation allow for collation of the temporal information into compact, timeline or comic strip-like objects, which provide a useful visual reference or signpost for a document's temporal structure. A projection operation allows for the identification of temporal relations between arbitrary events and times which appear in the strings, and also for validating that data is not lost or corrupted from the original sources. Possible treatments of incomplete information are also described, leveraging the relation set associated with Freksa's semi-intervals. Applications in annotation and scheduling are discussed, and a proof-of-concept online tool is presented which uses strings as a basis for creating, editing, and removing inconsistencies from documents marked up with TimeML.
Author: Woods, David Adam
Publisher:Trinity College Dublin. School of Computer Science & Statistics. Discipline of Computer Science
Type of material:Thesis
Availability:Full text available