Introduction
Before you apply the codeit AI to data in your project, there are a number of settings you can adjust to control which methods are used during the process.
Settings
Within the "Settings" tab you can adjust the following options:
| Options | Description |
|---|---|
| Use Text Matching | Choose whether to apply the Text Matching layer of the codeit AI. This layer autocodes verbatims by looking for identical verbatims in the example coded data. It also attempts to autocode verbatims by matching against the text labels of codes within the codeframe. |
| Use Text Rules | Choose whether to apply the Text Rules layer of the codeit AI. This layer autocodes verbatims by applying pattern matching expressions defined by users on codes within the codeframe. |
| Use AI Autocoding | Choose whether to apply the machine learning layer of the codeit AI. This layer autocodes verbatims by applying codeit's advanced deep learning system. |
AI Settings
If you click on the "AI Settings" button, there are several settings available in the Advanced Autocoding tab depending on your Task Type.
If your task is a "Brand" task containing brand mentions, the following settings are available:
| Setting | Description |
|---|---|
| Typo Matching | Specifies the amount of freedom the codeit AI has when matching mispellings. For example, matching the verbatim "McDonalsd" with "McDonalds". The higher you set Typo Matching, the more misspelled a verbatim can be and still be considered a match. |
| Brand Separators | If your verbatims contain multiple brand mentions in each single response (e.g. "McDonalds, Burger King, KFC" then you can specify the characters that separate the brands. codeit will split each verbatim according to these separators so that they can be coded separately. |
If your task is a "Text" task containing open-end verbatims, the following setting is available:
| Options | Description |
|---|---|
| Retrain Rob | Choose whether to retrain codeit's machine learning model before applying the autocoding. This option is only displayed if sufficient training data is available for this project. This option will bring the model up to date with any new coded examples that have been added since the last time the model was trained. This improves the accuracy of the model by ensuring it has learned from all of the latest data available to it. Switching this option off will speed up the autocoding by skipping the training phase of the process. If your model has previously been trained on a large amount of data and the amount of new data is small, it is probably not worth the additional time to retrain the model because the new training data wont make a material difference to the quantity and accuracy of the autocoding applied. |
| Guardrails | Allows you to define the Guardrails used when applying the AI autocoding. The Guardrails are defined here |