Sentiment Analysis is the process of determining whether a piece of writing is positive, negative or neutral. It’s also known as opinion mining, deriving the opinion or attitude of a speaker. A common use case for this technology is to discover how people feel about a particular topic.
The rise of social networks and blogs and forums, and the accompanying mountain of data available from these sources, has seen increased interest in the field of sentiment analysis. Many companies now exist that use a form of sentiment analysis to "listen in" on social conversations or to spot online trends. Several of the social media platforms or search engines themselves have opened up analytics capabilities to the public. Google or Twitter, for example, now allow individuals and brands alike to analyze events and reviews.
As well as being of value to consumer brands, financial analysts have started using sentiment analysis to gain an advantage when picking stocks or rating the financial performance of companies based on the general public’s views of their products or services.
Whilst great advances have been made in this field, there are still significant challenges in delivering not only more accurate, but also more informative, sentiment analysis. Most sentiment analysis algorithms use simplified and single words such as “great”, “angry” – single terms to determine whether the wider opinion is positive or negative.
Cultural factors, linguistic nuances and differing contexts make it extremely difficult to turn a string of written text into a simple pro or con sentiment. As they highlighted,
There are two restrictions with the approach of using these “anchor” words to determine overall sentiment.
The first is that the practice of using simple terms to express sentiment ignores the wider context of what is being said. The chocolate cake example above being a good illustration of this problem. Secondly, there are instances where there are no anchor words to guide us. For instance, factual statements can be difficult to analyze. Examples of factual statements that we have comes across include:
For each of these statements, traditional sentiment analysis algorithms would not be able to analyze or determine whether they are positive, negative or neutral.
As an Artificial Intelligence company that supports retailers in analyzing consumer opinion, we were keen at Aspectiva to solve this challenge. We recognized that up until now these sentences have been impossible to tag as positive or negative. As a result, we are missing out on both understanding more opinions as well as possible aspects and important facts about a product.
From August 2016 Aspectiva has updated its sentiment analysis algorithms and are now able to look at the wider sentence as a whole and include context to understand the true nature of an opinion. It does so by means of a machine learning algorithm applied in conjunction with Aspectiva’s core natural language processing algorithms. This new algorithm scans texts written by consumers and learns what people are saying when they are happy or unhappy with products they write about, beyond the obvious ‘sentiment words’ themselves.
As a result of this advance the insights we have collected have been much more significant and informative. Aspectiva’s algorithms are now able to understand people’s opinions about specific product features, determining sentiment also with factual sentences.
Aspectiva analyzes massive volumes of consumer opinions, turning them into comprehensive and valuable insights. Based on Artificial Intelligence and Natural Language Processing technologies, we leverage User Generated Content to help online shoppers search for the products they want and provide the recommendations to enable them to make informed purchasing decisions. Supporting eCommerce sites across any type of product or service, Aspectiva significantly increases shopper engagement and conversion rates.