A `MultiLabelField` is an extension of the :class:`LabelField` that allows for multiple labels. It is particularly useful in multi-label clas

F1 measure optimised for validation set

If key isn't there, returns default value. Like defaultdict, but it doesn't store the missing keys that were queried.

This module scores and prunes items in a list using a parameterised scoring function and a threshold. Parameters ---------- scor

There are cases where we need to express adjacency relations between elements in two different fields - for instance a TextField and a SpanFi

Compute the fraction of predicted event arguments that are associated with multiple triggers.

For multi-task IE, we want the training instances in a batch to be successive sentences from the same document. Otherwise the coreference lab

Computes relation candidate recall.

Represents a single coreference.

Holds all corefs and represents relations between them.

For multi-task IE, we want the training instances in a batch to be successive sentences from the same document. Otherwise the coreference lab

Computes precision, recall, and micro-averaged F1 for triggers and arguments.

Reads a single JSON-formatted file. This is the same file format as used in the scierc, but is preprocessed

Computes precision, recall, and micro-averaged F1 from a list of predicted and gold labels.

A one-hot encoder class. Only a module in the trivial sense.

Computes precision, recall, and micro-averaged F1 from a list of predicted and gold spans.

Copied (Span Based F1 Measure from allennlp)

An iterator which by default, pads batches with respect to the maximum input lengths `per batch`. Additionally, you can provide a list of fie

TODO(dwadden) document correctly. Parameters ---------- mention_feedforward : ``FeedForward`` This feedforward network is ap

Placeholder class to avoid wasting model parameters on stuff that we don't use.

TODO(dwadden) document me. Parameters ---------- vocab : ``Vocabulary`` text_field_embedder : ``TextFieldEmbedder`` Used

Event extraction for DyGIE.

Right now, counts how many event predictions are valid. I.e. the trigger / argument and argument / ner labels are compatible.

To use when we're co-training on Ontonotes and ACE.

Named entity recognition module of DyGIE model. Parameters ---------- mention_feedforward : ``FeedForward`` This feedforward

Parameters ---------- pretrained_model: ``str`` Either the name of the pretrained model to use (e.g. 'bert-base-uncased'),

Parameters ---------- pretrained_model: ``str`` Either the name of the pretrained model to use (e.g. 'bert-base-uncased'),

Relation extraction module of DyGIE model.

Read in data and do some spot-checks.