Annotation persistence over dynamic documents

Abstract

Annotations, as a routine practice of actively engaging with reading materials, are heavily used in the paper world to augment the usefulness of documents. By annotation, we include a large variety of creative manipulations by which the otherwise passive reader becomes actively involved in a document. Annotations in digital form possess many benefits paper annotations do not enjoy, such as annotation searching, annotation multi- referencing, and annotation sharing. The digital form also introduces challenges to the process of annotation. This study looks at one of them, annotation persistence over dynamic documents. With the development of annotation software, users now have the opportunity to annotate documents which they don't own, or to which they don't have write permission. In annotation software, annotations are normally created and saved independently of the document. The owners of the documents being annotated may have no knowledge of the fact that third parties are annotating their documents' contents. When document contents are modified, annotation software faces a difficult situation where annotations need to be reattached. Reattaching annotations in a revised version of a document is a crucial component in annotation system design. Annotation persistence over document versions is a complicated and challenging problem, as documents can go through various changes between versions. In this thesis, we treat annotation persistence over dynamic documents as a specialized information retrieval problem. We then design a scheme to reposition annotations between versions by three mechanisms: the meta-structure information match, the keywords match, and content semantics match.

(cont.) Content semantics matching is the determining factor in our annotation persistence scheme design. Latent Semantic Analysis, an innovative information retrieval model, is used to extract and compare document semantics. Two editions of an introductory computer science textbook are used to evaluate the annotation persistence scheme proposed in this study. The evaluation provides substantial evidence that the annotation persistence scheme proposed in this thesis is able to make the right decisions on repositioning annotations based on their degree of modifications, i.e. to reattach annotations if modifications are light, and to orphan annotations if modifications are heavy.