The use of electronic textual documents is gaining importance in many and different environments. For example, the majority of educational resources is often shared through e-learning platforms. In many other contexts, the digitalization is increasing: one need only think of news retrieved from online newspaper or social media. Therefore the need for summarization techniques to extract information in short time is consistently raising. Many attempts have been done with different techniques and models. This thesis work aims at analyzing the impact of word embedding information to improve the performances of an itemset-based summarizer.