Sentiment analysis is a natural language processing technique used to analyze text and identify the dominant emotional opinion expressed within it. The primary function is to determine whether the writer’s overall attitude is positive, negative, or neutral. This analysis is a key component of semantic analysis.
Sentiment analysis is often performed on data points such as user-generated content, customer feedback, and product or service reviews. The output of sentiment analysis typically includes a sentiment score (which indicates polarity, e.g., positive or negative) and a magnitude score (which indicates the strength of the opinion or emotion expressed). For instance, visualisations may show the sentiment score vertically and the magnitude via the size of the plot element.
Sophisticated NLP tools, such as the Google Cloud Natural Language API, Amazon Comprehend, and IBM Watson NLU, include specialized modules for performing document-level sentiment analysis. Some models, like IBM Watson NLU, can also offer granular emotion analysis, detecting scores for sadness, joy, fear, disgust, and anger. The analysis can be enhanced by combining it with entity extraction (Entity Sentiment Analysis) to determine the sentiment specifically directed toward recognized entities within the text.