I doubt you're actually lazy, if you're working in the customer experience or research world. Your to-do list is probably about 3 miles long, and you likely have a variety of stakeholders and executives pushing you for insightful, actionable, business-transforming insights – today. All the more reason to find a faster, hands-off approach to text analytics.
Late last year, we started revamping our text analytics approach at Walker to account for our CX programs' need for speed. Namely, we're incorporating an exploratory text analytics method that relies on mostly automated Natural Language Processing (NLP) to identify main topics or concepts found throughout a data set.
The overall process is pretty simple: load data, process data, find themes. (Of course, you still want to figure out how these themes are important to your overall program, but at least getting to those themes is simplified.)
The analysis outputs high-level words and phrases found throughout the data set, mostly nouns and adjectives. These groupings are generated based on NLP principles, which are looking for subjects across unstructured feedback. The outputs are fairly straightforward, allowing for a simple understanding of what's in the data. A comment can also have unlimited numbers of these outputs, or "concepts," which also provides fuller coverage of the data set.
In this example below, I've taken a small portion of the concepts generated for one of our survey questions. There are many more concepts than this, of course, but these are the top volume for demonstration purposes. You can see that the output is very high-level, but still provides a good indication of what we would find in the data:
Although the tool is mostly automated, the user does have some control over grouping these concepts together. For example, looking at the table above, we might decide to group all "licensing" concepts or all "online" concepts. This process starts to form a two-level hierarchy, giving more shape to the data, as seen in the image below. Licensing, for example, contains 54 items, or concepts, that the tool automatically generated. This group includes some of those mentions we saw in the previous image: licensed version, new licenses, licensing information, etc.
Now, instead of having to read through hundreds of concepts to understand the data, we can simply read through these Concept Groups to get a feel for what's being mentioned the most.
Volume isn't the only feature in this automated approach. There's also sentiment analysis that works based on "trigger" words and phrases. Sentiment is assigned at the concept level, giving more granularity and specificity toward understanding tone.
Using this method gives a quick, mostly hands-off approach to text analytics, and is good for many levels of users, whether you're a seasoned pro or you're just getting started. Of course, this approach fits a pretty specific need – getting a basic understanding of the data. If you need something more detailed or focused, you might consider a machine learning approach, which we also offer here at Walker.
Curious about which approach would fit your data best, and how to get started? Comment below or get a hold of us here!