Next generation annotation interfaces for adaptive information extraction

Abstract

The evolution of the Internet into the largest existent digital library is bringing about new challenges. One of the biggest problems is the location of information. The most promising approach seems to be performing searches semantically however this cannot work without semantically annotated documents. These documents are few and the manual annotation process to make them is both time consuming and error prone. To solve this problem Information Extraction (IE) technologies can be used to automatically annotate these documents, but be- fore doing so, IE tools require training examples. These examples are normally created manually by human annotators. Currently, there exist very few tools designed to support such people. This paper proposes a methodology aimed at supporting annotators by reducing the number of annotations required by an IE system therefore having effective learning. The whole methodology is implemented in the Melita system which will also be described in this paper. Finally enhancements to the existing methodology are being proposed in order to make IE accessible to a wider range of users, from inexperienced to expert users.