Text analytics is all about turning customer comments into quantifiable data. Computers like numbers and patterns, which customer comments and complaints, submitted via surveys or website comments, unfortunately, do not provide. Text analytics programs, also known as text mining programs, create accessible quantitative data from a natural-language input.
Most surveys ask general satisfaction questions and assign numerical values to the answers. Many will follow those questions with open-ended questions that ask why respondents chose the score they gave. Text analytics allow companies to understand the ‘why’ behind the score by making text responses into quantitative data.
How Do Text Analytics Work?
Several kinds of text analytics schemas exist. Keyword analysis can identify general themes of customer comments, whereas sentiment analysis distinguishes between positive and negative comments. Some programs measure word frequency, density, collocation (words that typically appear near each other), concordance (common context of a given phrase), N-grams (common short phrases), entity recognition (identifying places, names, dates, etc.), or dictionary tagging (searching for specific phrases in comments).
The most accurate way to derive meaning from text is to have an intelligent, trained human read the text and classify it. Unfortunately, this method is also the slowest and most costly and introduces risks of bias, misinterpretation, and other human error. For large companies with many sources of unstructured (non-numerical) data, human coding is not necessarily a reasonable option. Thus, enter text analytics.
The majority of companies who use any kind of text analytics use remedial or basic text analytics. These programs often use keyword or sentiment analysis and can be useful in some situations. However, they do not provide context and usually register false positives.
More advanced programs use natural language processing or NLP. They are programmed to understand grammar, phrases, slang, and language trends. They use stemming, categorising words by their roots (for example, the words fisher, fished, and fishing would be classified together). These programs provide precision and categorise comments efficiently, according to protocols established by programmers. Advanced text analytics has its problems, as taxonomy and classification must be well programmed before the analysis begins to yield useful results. However, it can provide more valuable and applicable data, at a more reasonable cost, than any other type of text analytics program.
E-BOOK
What Makes a Text Analytics Solution Truly Successful?
Get to the heart of what makes a text analytics solution truly successful. Download our free eBook to learn more about top terms, accuracy, solution necessities, and more!
