Text PredictionΒΆ

For supervised learning with text data, AutoGluon provides a simple fit() function that automatically produces high quality text prediction models. In the data, each training example may be a sentence, short paragraph, or be comprised of multiple text fields (e.g. predicting how similar two sentences are). The target values (labels) to predict may be continuous values (regression) or discrete categories (classification). A single call to fit() will train highly accurate neural networks on your provided text dataset, automatically leveraging accuracy-boosting techniques such as fine-tuning a pretrained BERT/ALBERT/ELECTRA model (transfer learning) and hyperparameter optimization.

Quick Start Using FITbeginner.html

Quickly get started with text prediction problems.

Heterogeneous data typesheterogeneous.html

Train neural networks that jointly operate on a mix of text, numerical, or categorical features.