Abstract
This paper proposes a document skimming environment for surveying documents in our research. Although there are a lot of on-line documents on our surroundings, people, generally, prefer printing out on-line research papers from computer screen. For this reason, although skimming is used for reading documents in our daily life, it is difficult for us to skim documents from computer screens.
Therefore, we developed a document skimming environment. The environment has a skimming support system and a recommendation system. The skimming support system supports skimming documents from computer screens by the interface, which is applied the Fisheye effect and the Overview+detail effect. Focus points of the Fisheye effect are the sentences selected by the original sentence extraction algorithm based on the value of standard distribution, and the Overview interface is displayed automatically the generated table of contents. The recommendation system generates personalized summaries by the collaborate filtering, which use users' log of the skimming support system. Furthermore, evaluation results show as follows; The value of F-measure of our sentence extraction algorithm is higher than it of the sentence extraction algorithm based on TF or Japanese lexical chaining method, the skimming support system is more effective method to skim documents from computer screen than paper, and the skimming support environment is more effective method to product research proposal documents than paper.
