User Annotations as a Context for Related Document Search on the Web and Digital Libraries
Jakub Ševcech, Róbert Móro, Michal Holub, Mária Bieliková · 2014
In this digital age, a lot of documents that people read are accessed through the Web and read on-line. There are various applications and services which enable creating bookmarks, tags, highlights and other types of annotations while reading these electronic documents. Annotations represent additional information on a particular information source and indicate that documents or their sections are somehow interesting for the document reader. However, existing approaches lack immediate reward for content annotation. We propose a method for query construction enabling search for other documents related to the currently studied one using not only the document’s content, but also user created annotations as indicators of user’s interests. In our proposed approach, annotations are used to activate nodes in a graph created from the document’s content employing spreading activation algorithm. We evaluate the proposed method in Annota- a service for bookmarking and collaborative annotation of Web pages and PDF documents displayed in a web browser. Along with its main purpose, Annota is designed to support scenarios useful for a novice researcher working together with his or her mentor. Based on Annota usage data we also analyzed properties of various types of annotations. Discovered annotation properties served as a basis for simulation we performed to determine optimal parameters of the query construction. We compared the proposed method to the commonly used tf-idf based method which our method outperformed when using annotations in the query construction process by improving the overall precision of the document retrieval. Therefore, annotations proved to be a viable source of information for user’s interest detection. Povzetek: Razvita je metoda, ki pri delu s spletnimi besedili uporablja oznake v besedilu. 1