I started using NVivo – a qualitative research and content analysis software – last year when I worked on a content analysis project. NVivo’s purposes are simple – it allows us to assemble different documents (including pictures) and code them. The codes can be arranged in a hierarchy or individually. While most coding is manual, some of it can be automated (see below). The best part of NVivo is that it allows us to review contents carrying a particular code with ease.

For example, I was looking at newspaper articles on contracting and wanted to know what were the range of ‘negative’ opinions about the topic across 200 articles. After I had coded some passages that contained the ‘negative’ view I could just ask NVivo to retrieve just those sentences of paragraphs. If I cannot make sense of a particular sentence in the result, I could by a click of a button ask it to fetch me the associated paragraph or the whole article. Though the task is simple, it makes the coding process so much more easy. Apart from retrieving passages coded with a particular code, I can also ask it to retrieve parts that have two or three specific codes that overlap (e.g. negative stories from Indian newspapers). Finally, I can ask it to create ‘assay tables’ for specific parameters for some quantitative analysis (basically tabulation, but one could attempt fancier analysis as well).

Brief note on autocoding with NVivo

Autocoding is a challenge with textual data that a lot of people are trying to get around. As of today, I do not know of a reliable system that would do the work for us without getting into a lot of coding. So NVivo’s autocoding is quite primitive. What it can do is to find certain strings across hundreds of documents and associate codes with it. If we have defined the data well, we can ask it to code certain sets of documents (e.g. only Indian newspapers, not American newspaper). I found it very productive to use it along with Wordnet of Princeton University (http://wordnet.princeton.edu). Wordnet enables us to create sets of words related to a word of interest e.g. ‘anger’. Apart from synonyms, it will create a list of associated words and forms of these words which can enable us to capture the notion in a variety of forms across hundreds of documents.