5 text analytics approaches:
A comprehensive review

In this 2-part article, author Alyona Medelyan begins Part 1 by explaining what text analysis is and how a text analysis application works in simple terms. To conclude Part 1, the author explains the potential of her applications and how companies can benefit from them.

In part 2, the author develops in detail (and with diagrams) the advantages and disadvantages of the 5 approaches used in text analysis.

5 Approaches to text analysis:

• Method 1:
  Word spotting (or keyword spotting), using a correspondence model;

• Method 2:
  Manual rules, identical to the keyword spotting method. The difference is that this rule uses regular expressions instead of a single word.

• Method 3:
  Text categorisation, a method using machine learning to determine categorisation rules based on the first 2 methods.

• Method 4:
  Subject modelling, identical to method 3, the only difference being that this method uses unsupervised learning, which means that the categorisation is carried out without the data being labelled beforehand. In short, the algorithm categorises a text on its own.

• Method 5:
  Thematic analysis, with this method (unsupervised machine learning), the algorithm is based on themes rather than words.