Text analysis is the process of deriving information from text-based sources. Researchers and commercial entities use text analytics for multiple purposes: summarization, sentiment analysis, explicative,
investigative, and classification. For the purpose of this post, only summarization and sentiment analysis are covered and their applications for business and marketing.
The huge amount of data available on the Internet provides opportunity to extract information that, when filtered on a particular subject or source, can provide value and business insight into service shortfalls, service successes, product information, trends, pop culture memes, product and service sentiment, brands, and so on.
Summarization
Summarization attempts to find key content across a large body of information or within a single document. The analysis identifies high-level concepts in text documents and delivers key ideas and actionable insights that you need with powerful visualizations and data exports.
As the name implies, the summarization process reduces documents down to their most important points. A summarization is a shortened version of a text that contains the main points of the original document or quantity of text. There are two approaches to summarization: extraction and abstraction. The essential part of an extraction-based approach is the identification of text that contains the important information contained within a document. Alternatively, abstraction is the process of paraphrasing sections of text from the source document. Abstraction can produce a more highly condensed summary, but requires the use of the far more sophisticated natural language processing. This method produces an abstract synopsis of the source document that is more human readable, but is a much more complex process.
Sentiment Analysis
Sentiment analysis or opinion mining is a process that evaluates the overall attitude of a piece of text that can range from a sentence or two to very long documents. It is an attempt to assign a general label to the sentiment of the author as positive, negative, or neutral. This type of general sentiment assignment is known as polarity, but there are further sentiments that reach beyond polarity that attach emotional descriptions to text, such as angry, happy, or sad.
There are several methods of assigning sentiment to text: statistical analysis, keyword spotting, lexical affinity, and concept-level techniques. In recent years, researchers have applied sentiment analysis to online social media outlets with a great deal of success. Twitter, for example, is a hot property for sentiment analysis and companies watch those feeds for sentiment about their products, their brands, and their customer’s interactions with customer service and brick and mortar outlets.
Use Cases
In industries where customer service is of particular importance, companies will perform text analyses on forums and opinion sites to extract sentiment information or customer experience information that isn’t company solicited. Surveys don’t often illicit 100 percent reliable or honest responses from individuals who might assess an experience too negatively or to positively. Companies scrape forums and opinion sites to gather usable and actionable data.
For product manufacturers, text analysis can take the form of keyword or key phrase extraction in order to make product improvements, to issue a fix, to post a workaround, or to prepare customer service representatives with a possible topic list. For example, if a company manufactures flashlights, it might setup a key phrase filter such as, “battery life” or “battery drain” in an attempt to find out what experiences customers have with their product under personal use conditions.
Some companies use text analysis to examine customer service chat sessions and transcriptions of telephone conversations to pick out keywords or key phrases that might lead to a product recall or other remedy.
But the news for text analysis isn’t all bad. Companies also use text analysis to extract positive testimonials to post on their websites or in their product marketing brochures. Positive testimonials from actual customers are powerful selling tools. For example, if you want to purchase a new mobile phone, but you can’t decide which brand or model to choose; you might look at reviews or testimonials to help you make your decision. If one of the products has 800 five-star reviews out of 920 total reviews, your opinion will probably sway in favor of that phone compared to another model that has 80 five-star reviews out of 87 total. Text analysis can glean those reviews from multiple sites and from various sources to create a concise report of product sentiment.
Text analysis has a variety of possibilities for companies and individuals who want to check product and service sentiment, customer feedback, customer service information, or any number of other applications. Every industry and business, regardless of size or market reach, can benefit from customer feedback analysis. Text analysis is one of the most cost-effective, highly accurate, and quickest to implement market research methods available.
