Text as Data

CHAPTER 2Text Analytics Process Overview

TEXT ANALYTICS PROCESSING

In this chapter, we identify a number of best practices in the areas of machine learning, data and text mining, and analytics processing. A few processing templates have evolved for data mining and machine learning.ⁱ The cloud-enabled approach adopted by SAS is summarized in SAS Institute Inc.ⁱⁱ This is a fast-moving area where new practices evolve constantly.

PROCESS BUILDING BLOCKS

A high-level view of processing for text analytics resembles many solution approaches in information technology. This section looks at the primary building blocks often used in text analytics:

• Preparation. Getting the text ready for analysis (data capture, text decomposition, mapping to a data representation)
• Utilization. Interpretation and deployment.

Figure 2.1 describes the life cycle of text analytics from capture to deployment in six major processes. We can map document capture, test-to-data transfer, and characterization in the preparation phase. We can map latent structure development, composite document assembly, and prediction/classification in the utilization phase.

Figure 2.1 Main stages of the text-mining process.

Source: B. deVille.

Preparation

• Capture documents. First, assemble the documents. Usually, text documents require some kind of preprocessing to bring them into the analysis environment. For example, articles ...