JADT 2010 : 10 th International Conference on Statistical Analysis of Textual Data Topic connections and clustering in text mining: an analysis of the JADT network Domenica Fioredistella Iezzi Department of Philosophical Studies -Tor Vergata University of Rome Riassunto Nei convegni scientiici, gli studiosi discutono lo stato di avanzamento dei loro lavori e instaurano nuove colla- borazioni. Le conferenze, quindi, costituiscono un luogo privilegiato dove avviare dei legami con altri studiosi e promuovere nuove attività di ricerca. La collaborazione tra ricercatori rappresenta un prototipo di reticolo sociale. Nel presente lavoro, sono stati esaminati i paper presentati nelle dieci edizioni del Convegno JADT dal 1990 al 2010. L’obiettivo della ricerca è stato di individuare i metodi più utilizzati delle diverse edizioni. In particolare, sono state evidenziate le strutture di collaborazione a livello macro (la rete globale che si è creata nelle diverse edi- zioni) e a livello micro (gli attori centrali e i temi più rilevanti). A questo scopo sono state utilizzate le tecniche di Social Network Analysis per misurare la centralità di alcuni nodi e classiicare gli argomenti trattati dagli studiosi. Abstract In academic conferences, researchers usually discuss their work and clasp scientiic collaborations. Conferences provide an important link for the exchange of information among scholars. Collaborations among scientists represent a prototype of social network. We present the macro (the whole network) and micro (both actor and topic centred) structure of collaboration in the last 20 years in International Conference JADT. We use centrality measures to evaluate roles of scholars and their topics and classify research topics. Key-words: centrality measures, cluster analysis, Social Network Analysis, text mining 1. Introduction Collaboration among scholars is a key mechanism of knowledge lows in research activities. It represents a complex phenomenon, which should be studied from several perspectives: individual scientists, research institutions, national and international research polices. Several studies have shown that scientiic productivity depends, among other things, on scientists’ attitude towards collaboration in research (Lee and Bozeman, 2005; van Rijnsoever et al., 2008). Text mining is a new research area that requires expertise in Information Technology, Linguistics and Statistics (Bolasco, 2005), because it tries to solve problems by using techniques from data mining, machine learning, natural language processing (NLP), information retrieval (IR), and knowledge management (Feldman and Sanger, 2007). The aim of this paper is to propose a method to analyse the macro (the whole network) and micro (both actor and topic centred) structure of collaboration in the last 20 years in International Conference on statistical analysis of textual data conferences (JADT). We want to discover patterns of collaboration in a closed network of scholars and to ind links in writing articles. Cluster Analysis encompasses many different techniques for discovering structure within