AUTOMATIC ANNOTATION MANAGEMENT THROUGH DOCUMENT ENRICHMENT AND TRANSFER

Abstract

Annotations in a document can provide readers useful information for understanding its contents and readers become more active in the flow of text by highlighting text and giving comments. However, traditional online collaborative annotation services suffer from two drawbacks. First, many uploaded documents have few annotations so that the activeness on those documents is low. Second, the same or similar documents may be uploaded several times by different users. As a result, redundant and inconsistent annotations arise. In this thesis, solutions are presented to those problems. For the first problem, in order to enrich the documents with few annotations, we propose linking relevant Q&A site contents to text documents. Q&A sites have a lot of high quality question-answer pairs. These contents can enrich the reading experiences of readers. For the second problem, it is resulted from the isolation of documents. We present a domination approach to link documents. When uploading a new document, annotations from similar documents should be transferred to this new document. By this means, the phenomenon of redundant and inconsistent annotations is alleviated. Extensive experiments demonstrate the efficiency and effectiveness of our proposed approaches.