Introduction
There are three distinct phases built into the codeit automation workflow: Extract - Refine - Apply.
These three stages are independent and can be mixed and matched according to the specific requirements of your coding project. None of the steps are mandatory, so you can choose which tools make sense for your specific project.
Extract | Refine | Apply
In high-level terms, the three coding stages that codeit offers can be summarised as follows:
| Stage | Description |
|---|---|
| Extract | Use themeit to automatically extract themes from your verbatims and autocode them. |
| Refine | Use codeit's powerful coding tools to efficiently adjust, refine and curate the auto generated output. |
| Apply | codeit's AI seamlessly learns in the background as you perform the "Refine" step above. For larger projects or tracking studies this allows the system to autocode more effectively by applying a machine learning model trained specifically on your project requirements. |
Best Practice: Which methods to choose?
The optimal approach to take depends on a few factors. To help you decide on the best methods to use, refer to this flowchart.
Extract
themeit applies the latest Generative AI technology, to automatically extract a list of suggested themes from your verbatims and autocode your data against these themes. It does this at the push of a button, with no need for prior training on your project or requirements.
themeit will also extract and attach a sentiment score to each of the codes applied to your verbatims. The resulting themes and coding can be examined and analysed using the themeit tool itself.
For more information on themeit and how to use it can be found here.
Refine
After completing the extraction phase, you may find that the autogenerated results are sufficient to meet your needs. However, an unsupervised AI process like themeit cannot give perfect results because it doesn't know anything about your research requirements or objectives. Since you are the expert, it is very likely you can improve the quality of the data by refining it further using your judgement, domain expertise and project knowledge.
At all times, you have full control over the themes and the coding applied to your project. codeit provides a rich, powerful user interface that allows you to fine-tune and refine the themes and how they are applied to your verbatims.
You can make a number of refinements within the themeit interface itself (see here for details) or, if you need to make more detailed, advanced changes you can take advantage of codeit's fully fledged manual coding interface (see here for details.)
Apply
Once you have refined the coding on a project to the level required, codeit will then have a large amount of high-quality examples of how your data should be coded against the themes you have generated. codeit can use this data to build a custom machine learning model using its advanced deep learning system.
This process happens seamlessly in the background without requiring any technical knowledge from users. Once codeit has been trained, it can apply this model to automatically code any new verbatims that are loaded into the project.
This method is more effective and more accurate than the unsupervised "Extract" step above because it can take advantage of the additional guidance given by you during the "Refine" step. The machine learning therefore learns by your example and emulates your coding when autocoding new data.
You can use codeit's guardrails to control the amount of freedom the machine learning has when applying autocoding. Any autocoding that doesn't meet the levels set by these guardrails can be reviewed, checked and approved manually by repeating the "Refine" step above. Repeating this process of refinement means that the machine learning continually improves and becomes even more effective over time.