One aspect of Emcien technology that we don’t spend a lot of time highlighting is the flexibility that EmcienPatterns gives its users. While most data analyses use highly structured and very clean data sets, the majority of  data is messy and unstructured. A good example of that flexibility is how some Patterns users are tackling unstructured data. While data structure provides a lot of additional information, there’s still a lot of data to be unlocked from unstructured text.

There are several ways to approach a text analytics problem with Patterns, but if you begin with  the right “Learn” data the analytics portion is relatively simple. As a means of demonstration, we’ll take the text from The American Presidency Project:  the 2012 presidential election speeches. Political speeches are helpful because they demonstrate how bits of unstructured text can become useful data points through analysis. After analysis, those data points illustrate how prescriptive analysis gives more context and guide automated responses through predictions.

Unstructured Text Prediction Example #1:


Unstructured Text Prediction Example #2:


There are a few different ways to analyze unstructured text, but one of the things that we try to highlight in the Patterns analysis is the importance of whyMaking an accurate prediction or classification is critical, but a prescriptive model will identify why each prediction was made. This predictive metadata can help guide analysts or automated systems for corrective action.

The words and groups of words are instructive when predicting a speaker, but how would we apply this method to analyze customer complaints, software trouble tickets, or user feedback? Not only would you want to predict the nature of the text in order to route the ticket to the right person or even direct users to an identified solution, but you would also want to know what else the metadata had to tell you.

Predictive analytics on unstructured text could tell you how difficult a customer complaint will be to address, or even a most likely solution. A prescriptive analysis will give you that same answer, but also the reasons that the prediction was made and what effect each element had on the prediction.

What else could you do with unstructured text?

  • Automate text classification
  • Route trouble tickets
  • Discover the true causes of customer complaints
  • Unstructured machine log analysis