User-generated online content has met a significant increase in recent years, as comments, online conversations, wikis etc. are an integral part of everyday human life. The storage capability improvement of such data, combined with evolution of machine learning techniques, has led to the creation of Natural Language Understanding-Processing (NLU NLP) systems. These systems are capable of extracting useful information, without any human intervention. The sentiment that is expressed in the content contributed by users is valuable information for analysis and a number of efficient analysis systems have been built. However, there are few systems that perform sentiment analysis for the Greek language. The lack of datasets hinders the research in that direction. Within the context of the present diploma thesis, a data annotation system is developed, which is used to create a dataset containing technology product reviews and the sentiment that is expressed towards their aspects. The dataset is utilized to train sentiment analysis models. Firstly, a comparison between already existing models for Greek is conducted. Then, a new architecture for Aspect Category Detection (ACD) from reviews is proposed, which is used along with an existing architecture for Sentiment Polarity (SP) to forge an end-to-end model. Furthermore, a web interface is implemented, with the purpose of analyzing text from any given review and presenting the respective results in a user-friendly graphical interface. Experimentation with the trained models shows promising results. The end-to-end model manages to accurately recognize aspects that are included in a review and analyze the sentiment expressed towards them.