Some have said that a picture is worth a thousand words. In the artistic concept I would agree, but in the realm of employee, social or B2B marketing research, words are the currency we use. Pause for a moment and contemplate the number of open-ended questions you have administered on your surveys within the last year. Now multiply that by the number of survey participants and you have a healthy mountain of untapped data.
Why untapped? Well if you are like me then the scaled questions get analyzed first. If there is time I might go back and take a quick peak at the responses to the open-ends, but certainly nothing exhaustive. Well that was before text analytics came into play. The major statistical packages (SPSS, SAS, etc.) have had text mining engines available for some time, but they are often a significant added expense. Now Cvent and others have added text analysis capabilities as a feature on their platforms.
For those steeped in qualitative research you are keen to the fact that understanding text data is an iterative process that requires reviewing the data at a high level and then creating categories, and possibly collapsing those categories. In the end you have added a series of categorical questions to your data file. The ‘ah ha’ moment comes when you realize you can now crosstabulate these new categories with other variables in your survey. This allows you the ability to tie these opinion statements to profiling variables (e.g. customer status, demographics, market segments, etc.)
To wrap up this intro to text analysis I will end with the beginning. The word cloud below was from a recent survey of application developers. The question was "What trends do you see affecting the market for application development?" The larger the word, the more times it was mentioned in the text. This serves as a starting point for creating your categories.