Sentiment analysis is a field of study that automatically analyzes sentiments, judgments, opinions, feelings, and/or emotions, primarily, in written texts on a particular topic, but also in other multimedia resources. Due to the advent of the Internet and especially, the Web 2.0, the importance of this field has increased in recent decades. Since then, many platforms have appeared such as social networks, online stores, blogs, etc., where people can exchange information and interact with each other at any time. All this information has become extremely important to companies wanting to know what consumers are saying about their products, or customers looking for suggestions on a wide range of activities.

People always care about the opinions of others when buying a new product, deciding what movie to watch and doing many other activities. Particularly, in the medical field, patients may rely on the opinions of others when choosing a hospital or doctor. As a result, medical social platforms have become a key information source  to know the patients' opinions, whether it is for a patient who wants to choose a hospital or for the hospital owners to assess and know what patients need. Nevertheless, processing all these data is a challenging undertaking, highlighting the need of developing advanced techniques and tools to process and exploit all this information in an automatic manner.

This thesis focuses on the development of new techniques and methods for improving some typical sentiment analysis tasks applied to the medical field. The main body of thesis is composed of four articles published in a re-known conference and three top-tier journals. The first article proposes  a new ranking algorithm for selecting hospitals based on intuitionistic fuzzy sets. The second article improves the previous one presenting a detailed methodology for customizing hospital recommendations based on aspect-based sentiment analysis and intuitionistic fuzzy sets. The third article proposes a new deep learning architecture for detecting multiple emotions from patients opinions as well as an automatic mechanism for labeling opinions according to their emotions. The last article studies a new mechanism for representing the polarity of the sentiments towards different aspects of hospitals considering terms such as indeterminacy using simplified neutrosophic sets. All these proposals have been broadly tested against other state-of-the-art approaches and implemented proposing possible real applications on the field of Medicine.

​​