Soft Computing for Recommender Systems and Sentiment Analysis Lorenzo Malandri a,b , Carlos Porcel c , Frank Xing d , Jesus Serrano-Guerrero e , Erik Cambria f a University of Milano Bicocca, Italy b CRISP Research Centre - University of Milano Bicocca, Italy c University of Jaen, Spain d National University of Singapore, Singapore e University of Castilla-La Mancha, Spain f Nanyang Technological University, Singapore 1. Introduction 1 The World Wide Web is becoming a bottomless source of unstructured 2 data, with quintillions of bytes of data generated daily and publicly accessi- 3 ble [1]. Social media, customer reviews, and online news articles, as well as 4 the comments associated with them, are just some examples of what the In- 5 ternet is producing in terms of text data. Text data is usually not standalone 6 like digitalized books, but associated with a lot of information about user be- 7 haviors and preferences. This has led to a growing interest in the research 8 of social media analysis and many applications, including sentiment anal- 9 ysis and recommender systems. Closely related to the two mentioned are 10 other tasks, such as opinion retrieval, opinion summarization, subjectivity 11 classification, sarcasm/irony detection and more. 12 We only see a strengthening trend of the online presence of text data 13 bounded with our daily activities, as we migrate to the metaverse. In such 14 a world, sentiment analysis and recommender systems are really two sides of 15 the same coin [2]: one about passively knowing about the users, the other 16 about actively reaching to the users. The main of sentiment analysis is to 17 understand the correct information from versatile expressions. The main 18 challenge of recommender systems is to filter and transform busy online in- 19 formation streams into structured data that can be used for content push 20 decisions. 21 Preprint submitted to Applied Soft Computing November 10, 2021