Sentiment Analysis

The use of natural language processing to systematically identify, extract, quantify, and study affective states and subjective information.

How it works

Traditional sentiment analysis used hand-crafted lexicons of positive and negative words combined with simple rules. Modern approaches fine-tune pre-trained language models (BERT, RoBERTa) on labelled examples of positive, negative, and neutral text. The model produces a probability distribution over sentiment classes. Aspect-based sentiment analysis goes further, identifying both the target (e.g., 'battery life') and the sentiment (negative) for each aspect mentioned in a review.

Why it matters

Sentiment analysis enables organisations to process customer feedback at a scale impossible manually. Businesses use it to monitor brand perception on social media in real time, analyse customer reviews for product insights, and gauge public reaction to product launches. In finance, sentiment analysis of news and social media is used as a trading signal. As more customer interaction moves to text channels, sentiment analysis has become a core tool in any customer experience programme.