framenet_tools.frame_identification package


framenet_tools.frame_identification.frameidentifier module

class framenet_tools.frame_identification.frameidentifier.FrameIdentifier(cM: framenet_tools.config.ConfigManager)

Bases: object

The FrameIdentifier

Manages the neural network and dataset creation needed for training and evaluation.

evaluate(predictions: List[<MagicMock id='139663474718424'>], xs: List[str], reader: framenet_tools.data_handler.reader.DataReader)

Evaluates the model

NOTE: for evaluation purposes use the function evaluate_file instead

  • predictions – The predictions the model made on xs
  • xs – The original fed in data
  • reader – The reader from which xs was derived

evaluate_file(reader: framenet_tools.data_handler.reader.DataReader, predict_fees: bool = False)

Evaluates the model on a given file set

Parameters:reader – The reader to evaluate on
Returns:A Triple of True Positives, False Positives and False Negatives
get_iter(reader: framenet_tools.data_handler.reader.DataReader)

Creates an Iterator for a given DataReader object.

Parameters:reader – The DataReader object
Returns:A Iterator of the dataset
load_model(name: str)

Loads a model from a given file

NOTE: This drops the current model!

Parameters:name – The path of the model to load
prepare_dataset(xs: List[str], ys: List[str], batch_size: int = None)

Prepares the dataset and returns a BucketIterator of the dataset

  • batch_size – The batch_size to which the dataset will be prepared
  • xs – A list of sentences
  • ys – A list of frames corresponding to the given sentences

A BucketIterator of the dataset

query(annotation: framenet_tools.data_handler.annotation.Annotation)

A simple query for retrieving the most likely frame for a given annotation.

NOTE: require are loaded network and a annotation object which has a sentence and fee!

Parameters:annotation – The annotation containing the sentence and the fee.
query_confidence(annotation: framenet_tools.data_handler.annotation.Annotation, n: int = 5)

A deeper query for retrieving a list of likely frames for a given annotation.

NOTE: require are loaded network and a annotation object which has a sentence and fee!

  • annotation – The annotation containing the sentence and the fee.
  • n – The amount of best guesses retrieved.

save_model(name: str)

Saves a model as a file

Parameters:name – The path of the model to save to
train(reader: framenet_tools.data_handler.reader.DataReader, reader_dev: framenet_tools.data_handler.reader.DataReader = None)

Trains the model on the given reader.

NOTE: If no development reader is given, autostopping will be disabled!

  • reader – The DataReader object which contains the training data
  • reader_dev – The DataReader object for evaluation and auto stopping

write_predictions(file: str, out_file: str, fee_only: bool = False)

Prints the predictions of a given file

  • file – The file to predict (either a raw file or annotated file set)
  • out_file – The filename for saving the predictions
  • fee_only – If True, only Frame Evoking Elements are predicted, NOTE: In this case there is no need for either train or load a network

framenet_tools.frame_identification.frameidentifier.get_dataset(reader: framenet_tools.data_handler.reader.DataReader)

Loads the dataset and combines the necessary data

Parameters:reader – The reader that contains the dataset
Returns:xs: A list of sentences appended with its FEE ys: A list of frames corresponding to the given sentences

framenet_tools.frame_identification.frameidnetwork module

class framenet_tools.frame_identification.frameidnetwork.FrameIDNetwork(cM: framenet_tools.config.ConfigManager, embedding_layer: <MagicMock name='mock.Embedding' id='139663474909424'>, num_classes: int)

Bases: object

eval_model(dev_iter: <MagicMock name='mock.Iterator' id='139663474718032'>)

Evaluates the model on the given dataset

UPDATE: again required and integrated for evaluating the accuracy during training. Still not recommended for final evaluation purposes.

NOTE: only works on gold FEEs, therefore deprecated
use f1 evaluation instead
Parameters:dev_iter – The dataset to evaluate on
Returns:The accuracy reached on the given dataset
load_model(path: str)

Loads the model from a given path

Parameters:path – The path from where to load the model
predict(dataset_iter: <MagicMock name='mock.Iterator' id='139663475192216'>)

Uses the model to predict all given input data

Parameters:dataset_iter – The dataset to predict
Returns:A list of predictions
query(x: List[int])

Query a single sentence

Parameters:x – A list of ints representing words according to the embedding dictionary
Returns:The prediction of the frame
save_model(path: str)

Saves the current model at the given path

Parameters:path – The path to save the model at
train_model(dataset_size: int, train_iter: <MagicMock name='mock.Iterator' id='139663474833504'>, dev_iter: <MagicMock name='mock.Iterator' id='139663474756912'> = None)

Trains the model with the given dataset Uses the model specified in net

  • dev_iter – The dev dataset for performance measuring
  • train_iter – The train dataset iterator including all data for training
  • dataset_size – The size of the dataset
  • batch_size – The batch size to use for training

Module contents