Work with the Classification Wizard
Hero Platform_'s Classification Wizard is a tool that allows users without AI experience to classify any type and amount of text and then create AI models in an automated Flow.
To build an AI model, the Classification Wizard needs a training dataset Input with a field for example text and a label field.
Create a Text Classification AI Model
- Open AI from the navigation menu and select AI Models.
- Click Create Classification.
- Enter a name for the Classification model.
- Select a dataset Input.
Select the Input field - The sample text.
Select the Output field - The label text.
Mark a radio button for the expected model size.
Click Train to finish and begin training the model.
The different model sizes present different system memory usage profiles. In particular, the length of the input texts (not the dataset size, but the number of characters per sample) affects memory.
The following tables outline the memory limits Automation Hero have tested against for various text lengths. A model was tested for 3 epochs on 10 categories with 100 samples each against various combinations of memory limits and model sizes. The table below reports the maximum tested character-length for which training succeeded.
2GB 3GB 4GB 5GB 6GB 7GB 8GB 9GB 10GB Small >50k >50k >50k >50k >50k >50k >50k >50k >50k Medium 2k 3k 4k 6k 8k 10k 10k 14k Large 1k 1k 1k 2k 2k 2k 2k
Note that these are worst-case scenario estimates for a dataset that is entirely messy. (I.e., Similar to a badly OCR-ed document.)
The Classification model is saved in your AI Model overview.
It may take several minutes depending on the amount of the data to complete the training.
View Classification Training Model Results
View the results of the training model by clicking the settings icon and selecting Show Performance.
- Dataset size: The amount of rows in the total dataset trained.
- Training time: The amount of time it took Hero Platform_ to train the Classification model.
- The correct verses false classifications within the training dataset.
- Hero Platform_ applies the trained model on a portion (20%) of the sample data to check for sample data label accuracy.
- Precision and Recall
- Displays each of the categories (labels) within the dataset as well as an unknown category for data the model did not put into a category.
- Each category displays the results for:
- Precision - The proportion of positive identifications that were actually correct. A model that produces no false positives has a precision of 1.0.
- Recall - The proportion of actual positives that were identified correctly. A model that produces no false negatives has a recall of 1.0.
Use a Classification Model in a Flow
After a Classification model has been trained, it can be used in a Flow.
To use a Classification model in a Flow:
- Open and start creating a Flow in the Flow Studio.
- View the Classification models in the element browser.
- Click and drag a Classification model from the element browser onto the Flow Studio canvas.
- Connect the Classification model using a cable from an element in the Flow.
- Configure the fields for the Classification model.
- Click OK to finish adding the Classification model to the Flow.