Sentiment Analysis Helps Find the Insights in Product Reviews

Product Reviews Have Become a Critical Data Source of Insight for Companies

Online reviews have become a very important source of data for companies who are seeking to understand how consumers are reacting to their products. Web sites like Amazon and Yelp contain huge amounts of data in the form of product reviews on how consumers feel about just about any aspect of a product.

These online product reviews give a product vendor access to many aspects of the customer experience, including:

• How customers perceive a product overall – is it great, good, so-so, bad, or worthless?
• Which features do they like or dislike?
• Which features are essential? Which are nice-to-have? Which can be dropped?
• How do they feel about the pricing?
• Do customers consider the customer support to be adequate?
• etc.

Product reviews also might reveal some serious problems that require a quick response on the part of the vendor before they get amplified and which, if unresolved, could have an impact on the vendor’s reputation.

Given their importance, product reviews are often an integral piece of a company’s product planning activities. Reviews provide critical feedback that help to ensure that a product is competitive and has all the features required for success.

How Sentiment Analysis Helps a Vendor Monitor Its Competing Products

Product reviews also enables a vendor to monitor its competitors and to compare how customers view their products versus its own. Unfortunately, having staff monitor its own social media presence is a big task for a company, especially a large, well-known one. Monitoring a number of products from competing vendors is an even bigger one. This is an area where an automated solution is required. Fortunately, there is an AI technology, Sentiment Analysis, which can meet the challenge.

How Sentiment Analysis Works

Sentiment Analysis identifies people’s attitudes, opinions, and feelings in unstructured data such as found in social media and other sources. Early generations of Sentiment Analysis could only identify sentiment-bearing words such as “Excellent!” or “Terrible.” It could not associate a sentiment with something mentioned in the context such as a product name or a specific product feature.

The latest generation, Advanced Entity-Based Sentiment Analysis, dramatically improves this situation, allowing an automated and very fine-grained analysis of sentiment. It allows the identification of sentiment regarding the specific features of a product, even when more than one sentiment is mentioned in the same sentence. For instance, in the example below, Sentiment Analysis will recognize a negative sentiment towards the iPhone design and a positive sentiment towards the new navigation feature.

I find the iPhone design too bland, but I like the new navigation feature.

Advanced Entity-Based Sentiment Analysis also makes possible the aggregation of the sentiment from a multitude of reviews. This is because it normalizes all the sentiments and entities it identifies. Human language can express the same sentiment in many different ways: “I find the design too bland,” “the design is too bland,” “too bland a design,” etc. All three statements mean the same thing. Advanced Entity-Based Sentiment Analysis converts each of these sentiments to the same structured representation of the sentiment, such as:

Sentiment: negative

Predicate: bland

Sentiment object: design

Associated entity: iPhone

This normalization allows sentiments to be aggregated, giving companies a statistically grounded view of what customers think about their products. The time stamps associated with the social media posts can also give companies a view of how customer attitudes develop over time. This supports a dashboard-style display and supplies specific quantitative evidence of very fine-grained customer attitudes towards all aspects of a product.

Summary

Based on AI techniques, Sentiment Analysis is an effective tool for market research, market monitoring, and competitive analysis and gives a company real-time access to what the universe of customers is thinking about its products and those of its competitors.