It may at first seem counterintuitive that math plays an important role in daily conversation, but let’s consider how frequently our word selection and sentence composition are supported by mathematical calculations. Then we’ll discuss the implications for Natural Language Generation (NLG), and the role of Arria’s advanced mathematical and linguistic functions in the assembly of sentences, paragraphs, and reports for use in business.
In daily conversation, even a simple observation that a residence is “big,” “large,” “huge,” “small,” or “tiny,” for example, relies upon an automatic comparison to other residences that we have in mind. Well, perhaps not to all the others. Perhaps we’re making a relative statement that requires a locational qualifier and a specification of residence type. “For Texas, that’s a small house.” “For Brooklyn, that condo is huge!” Still speaking of the Brooklyn condo, we might go on to say, “But for $1.5 million, 1,200 square feet is tiny compared to the mansion you could buy in Texas.”
Our inherently mathematical nature is a factor in our conversation, whether we are a math PhD, an English major, or a toddler pointing and babbling toward a particular piece of cake that, among its peers, has the greatest total surface area of frosting. Perhaps we don’t have the training to put the formula on paper, but to some degree, everyone—even the toddler—is choosing language and assigning importance in a manner that could be expressed in numbers. In turn, with such simple data sets, humans are able to convert the numbers back to language relatively easily.
Not so with the complex data sets found in the business environment. A tidal wave of numbers ensures that no one, no matter their training, can possibly express those numbers efficiently in words. Insights are lost in the flood. Employees, managers, and executives attempting to comprehend the data find it hard to identify the right patterns and outliers for conversion to the best insights. That is why it is critical for any business-focused NLG system to contain built-in mathematical and linguistic functions—not simply to mirror human thinking, but to do so at massive scale. Such a system helps to bridge the gap between the machine’s computational power and the human’s linguistic choices, giving each entity a measure of the other’s skill.
In the housing example above, the underlying math pertaining to size and price will differ based on geography, and therefore so will the language. Similarly, the relative value of measures in a BI dashboard or any other business analysis reporting depends on dimensions such as time, location, the performance of managers and other personnel, and other factors. The question, for example, of whether a monthly sales figure variance is “acceptable,” “unacceptable,” or “outstanding,” compared to the historical standard deviation is a math question that the system answers in natural language, with human guidance providing the numerical thresholds to select the appropriate words.
Obviously, I’m writing this because Arria offers the functionality described above, but don’t take our word for it. If you’re considering an NLG implementation, be sure to read The 4 key questions (plus one bonus question!) to ask vendors when it comes to NLG, especially the bonus question at the bottom: “Are you willing to have a bake-off against Arria?”
Meantime, we invite you to take the next step toward bringing NLG to your organization by requesting a brief, personalized demo today.