NLP & Sentiment Analysis

Why do opinions matter?

Many decisions are based on other people’’ opinions and recommendations (where to eat, where to go on holiday, which computer to buy, etc). For a company, it is important to know what customers think about its products and services. These opinions can influence other potential customers, especially when they are being published on the internet. Knowing people’s opinions is therefore an invaluable business asset which can help improve products and services and enhance marketing strategies.

How is Sentiment Analysis done?

Traditional approaches to Sentiment Analysis rely on a polarity lexicon, i.e. a lexicon in which words are associated with positive, negative or neutral orientation (“polarity”). Some polarity lexica will also indicate the strength of polarity, e.g. by using scores. The sentiment analyser then identifies these polarity words and potential targets (products that the opinion is aimed at) in a piece of text, e.g. by taking into account syntactic structure. It then tries to identify plausible opinions about the given targets.
More recent approaches include the use of neural networks (so called deep learning) to learn associations between texts and opinions expressed in them.