extract the topk span indices Args: candidate_scores: param candidate_starts: candidate_ends: param topk: [num_sentences]

Single document summarization is the task of selecting a subset of the sentences which best represents a summary of the document, with a bala

Compute the F1 score. The F1 score can be interpreted as a weighted average of the precision and recall, where an F1 score reaches its be

adapted from arXiv:1911.07471

r""" Args: scores (~torch.Tensor, ~torch.Tensor): Tuple of two tensors `s_arc` and `s_sib`. `s_arc

r""" Args: sequence (~torch.nn.utils.rnn.PackedSequence): A packed variable length sequence. hx (~torc

Build a CoNLLSentence from a dict. Args: d: A dict storing a list for each field, where each index corresponds to a token.

Build a CoNLLSentence from ``.conllx`` or ``.conllu`` file Args: path: Path to the file. conllu: ``True`` to build :clas

Build a CoNLLSentence from CoNLL-X format str Args: conll (str): CoNLL-X or CoNLL-U format string conllu: ``True`` to bu

Compute span score with FFNN(span embedding) Args: span_emb: tensor of [num_sentences, num_spans, emb_size] config: param

Retrieves all treebank data paths in the given directory. Adopted from https://github.com/Hyperparticle/udify MIT Licence Args: dat

Default data directory depending on the platform and environment variables