Text Prediction¶
For supervised learning with text data, AutoGluon provides a simple fit() function that automatically produces high quality text prediction models (Transformer neural networks). Each training example may be a sentence, a short paragraph, comprised of multiple text fields (e.g. predicting how similar two sentences are), or may even contain additional numeric/categorical features beyond just text. The target values (labels) to predict may be continuous values (regression) or discrete categories (classification). A single call to predictor.fit() will train highly accurate neural networks on your provided text dataset, automatically leveraging accuracy-boosting techniques such as fine-tuning a pretrained NLP model (transfer learning) and hyperparameter optimization.
How to train high-quality text prediction models in under 5 minutes.
How TextPredictor can be applied to multimodal data tables with a mix of text, numeric, and categorical columns.
How to specify custom hyperparameters of the TextPredictor neural network, and tune them with HPO.