Model description

This model is a fine-tuned version of RoBERTa.

Intended uses & limitations

For identifying time expressions in text. This model works in a NER-like manner but only focuses on time expressions.

• You may try an example sentence using the hosted inference API on HuggingFace:

  In Generation VII, Pokémon Sun and Moon were released worldwide for the 3DS on November 18, 2016 and on November 23, 2016 in Europe.

  The JSON output would be like:

  [
    {
      "entity_group": "TIME",
      "score": 0.9959897994995117,
      "word": " November 18",
      "start": 79,
      "end": 90
    },
    {
      "entity_group": "TIME",
      "score": 0.996467113494873,
      "word": " 2016",
      "start": 92,
      "end": 96
    },
    {
      "entity_group": "TIME",
      "score": 0.9942433834075928,
      "word": " November 23, 2016",
      "start": 104,
      "end": 121
    }
  ]
  

Training and evaluation data

TimeBank 1.2

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-05
• train_batch_size: 16
• eval_batch_size: 256
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 3.0

Framework versions

• Transformers 4.25.0.dev0
• Pytorch 1.12.1
• Datasets 2.7.1.dev0
• Tokenizers 0.13.2