Evaluation of Session-Suitability of Papers in Conference Programs

Abstract

Every year thousands of conferences are organized in many parts of the world. These conferences are realized in specific sessions. While determining the papers to be included in these sessions, the topic similarities are taken into consideration. Papers of a specific area are included in the same sessions. However, sometimes there are papers that are not very related to each other in the same session. This causes a participant to leave after listening to one or two papers. This is an undesirable situation for participants who are interested in the session topic. The participants' satisfaction with the sessions attended shows the efficiency of the conferences. The aim is to enable the participant to benefit from the conference in the most effective way. In this study, the similarity of the papers in the past conferences sessions was investigated and the efficiency of the conference sessions for the participants was discussed. In order to find similarity relations between papers, the deep learning based Doc2Vec method, which has recently received great interest in text processing, was used. The method was applied to full texts of the papers and the similarity score of the papers in one session was calculated. Then, the overall score was found with the average of the similarity scores of all sessions. The high similarity score of the papers in a session indicates that an efficient session was created in common topic. The study was conducted using papers of the ASONAM 2018 conference program. The efficiency of this conference program is discussed with the results obtained.