Introduction
Rather than offering a "one size fits all" approach to coding, codeit provides a set of tools that you can use according to your individual requirements. The exact approach you take will depend on factors such as time, cost, resources available, the nature of your data, research objectives and so on...
However, the options available can be summarised into three steps: Extract | Refine | Apply.
None of these steps are compulsory - you can pick and choose depending on what makes sense for your particular project.
Extract
During this step, codeit will automatically extract a set of suggested themes or codes from your verbatim data and (optionally) autocode your verbatim data against these items.
This process is completely automated and works out of the box so requires no prior input training data from you (in the technical jargon, this is "unsupervised learning").
The "Extract" process is initiated in a variety of ways, for example by using the themeit tool, codeframe builder tool, or the dedicated brand coding tool.
Refine
Automated AI processes can get you a long way, but are unlikely to produce a "perfect" result on their own.
Of course, a lot depends on your definition of "perfect" or even what you consider "good enough" for the objectives at hand.
Whatever you're trying to achieve, you will probably want the option to refine and curate the automated output.
This is a fundamental aspect of the codeit philosophy - the process needs to be human-led, meaning that real people can take the controls at any time and adjust what the AI has produced. For example, you may want to merge themes, rename themes, apply coding, change coding and so on, based YOUR domain knowledge, research objectives and contextual understanding. This is the bit that humans are good at, so it's important we make it as easy as possible for real people to work with the data and make adjustments.
Apply
At the "Extract" stage, the codeit AI has very little to go on. It has to make a "best guess" at how your data should be categorised and coded. However, once the "Extract" and/or "Refine" stages are complete, the AI does have something very useful to go on - a full codeframe and set of coded examples to learn from.
Using these examples, codeit can seamlessly and automatically build a machine learning model based on your data and the way it's been coded. You are essentially training codeit by example, it learns from what you do.
This machine learning can then be used to autocode new verbatims that require coding subsequently. codeit will be able to do more autocoding more accurately because the model is bespoke to the exact intricacies of your project.
All of which means this can be very useful for autocoding ongoing projects (e.g. tracking studied), repeat projects (e.g. periodic "dips" or pre-post surveys) or simply very large ad-hoc projects.
Mixing and Matching
All three of these steps are optional. You can use all or some of them. For example, you might use themeit to extract your themes and that alone is sufficient your needs. Or you may need to skip the Extract step and instead simply code your data manually building up the themes by hand as you go. Or, if you have some existing training data you could skip the first two steps and simply autocode your data purely by the application of machine learning using your training data.
There are several possibilities, but whichever route you take codeit allows you to handled the process in one simple to use, unified tool.