tfa.text.crf_multitag_sequence_score

View source on GitHub

Computes the unnormalized score of all tag sequences matching

tfa.text.crf_multitag_sequence_score(
    inputs: tfa.image.color_ops.TensorLike,
    tag_bitmap: tfa.image.color_ops.TensorLike,
    sequence_lengths: tfa.image.color_ops.TensorLike,
    transition_params: tfa.image.color_ops.TensorLike
) -> tf.Tensor

tag_bitmap.

tag_bitmap enables more than one tag to be considered correct at each time step. This is useful when an observed output at a given time step is consistent with more than one tag, and thus the log likelihood of that observation must take into account all possible consistent tags.

Using one-hot vectors in tag_bitmap gives results identical to crf_sequence_score.

inputs A [batch_size, max_seq_len, num_tags] tensor of unary potentials to use as input to the CRF layer.
tag_bitmap A [batch_size, max_seq_len, num_tags] boolean tensor representing all active tags at each index for which to calculate the unnormalized score.
sequence_lengths A [batch_size] vector of true sequence lengths.
transition_params A [num_tags, num_tags] transition matrix.

sequence_scores A [batch_size] vector of unnormalized sequence scores.