This text suggestion tool uses simple
language model to predict expected words after entered text.
Language model used consists of Embedding, GRU cell and Dense layer. It was trained on OpenSubtitles dataset.
LM is evaluated on your device (via TensorFlow.js) so prediction may take some time on slow devices (like smartphones).
Use this link to load small LM for slow devices: use small LM
Warning: Slow CPU backend is used. Only one word will be predicted (at most 10 chars).